sesanboot en
TRANSCRIPT
SPARC Enterprise – ETERNUSSAN Boot Environment Build Guide
- Solaris (TM) Operating System -
Version 1.4
C120-E392-08EN
Preface
Purpose
This manual explains how to mount a 4Gbps Fibre Channel card (SE0X7F11x/SE0X7F12x) on SPARC
Enterprise and build a SAN Boot environment, in which the OS can be booted from ETERNUS storage
system.
When 8Gbps Fibre Channel card (SE0X7F21x/SE0X7F22x) is used, the construction procedure of thismanual is not necessary. When OS is installed in ETERNUS storage system, the environment isconstructed according to the same procedure as the procedure for installing OS in the internal disk. Referto Solaris online manual (Solaris SAN Configuration and Multipathing Guide) for Solaris OS standardfunction multipath function (MPxIO).
Disk array devices other than ETERNUS are not described at this guide. See Chapter 2.1.1 Required
Hardware.
Intended Audience
This manual is intended for builders and administrators of SAN Boot environments.
Organization
This manual consists of the following chapters:
Chapter 1 Overview
Introductory information about a SPARC Enterprise SAN Boot environment
Chapter 2 Hardware/Software Configuration
Patterns of configuration prerequisite to building a SAN Boot environment
Chapter 3 Precautions
Precautions in building and running a SAN Boot environment
Chapter 4 Building an OS Boot Environment
How to build a SAN Boot environment
Chapter 5 Backing Up and Restoring Boot Disks
How to back up and restore boot disks in a SAN Boot environment
Notation
The methods of notation are used in this manual:
Solaris(TM) 10 Operating System is noted as "Solaris 10."
Actual commands appear in boldface.
# /usr/sbin/FJSVpfca/fc_info -a <Return>
Trademark Notice
Sun, Sun Microsystems, the Sun Logo, Solaris and all Solaris based marks and logos are trademarks or
registered trademarks of Sun Microsystems, Inc. in the U.S. and other countries, and are used under
license.
Fujitsu Limited
January 2010
Edition 1.4 Jan 2010
Notice
This manual shall not be copied without the permission of the publisher.
The contents of this manual are subject to change without notice.
All Rights Reserved, Copyright (C) Fujitsu Limited 2006-2010
Manual Update History
Edition Date published Description
1 May, 23rd, 2007 First release
1.1 Sep, 14rd, 2007 Support ETERNUS 2000 series
Support SPARC Enterprise T5120/T5220
1.2 Oct, 14rd, 2008 Support Solaris 10 10/08
Support SPARC Enterprise
T5140/T5240/T5440/M3000
1.3 Apr, 3rd, 2009 Change the restore procedure
Change the check command procedure of
Custom JumpStart
1.4 Oct, 28rd, 2009 Add the Fibre Channel card boot code update
procedure
Add the precaution 18
Support ETERNUS DX series
Contents
Chapter 1 Overview............................................................................. 1
1.1 Configuration Patterns........................................................................................ 5
1.1.1 Basic configuration................................................................................................................. 5
1.1.2 Disk array device mirroring configuration based on PRIMECLUSTER GDS ....................... 7
1.1.3 Cluster configuration based on PRIMECLUSTER................................................................. 8
1.1.4 ETERNUS Configuration required to use the Advanced Copy function ................................ 9
Chapter 2 Hardware/Software Configuration................................ 14
2.1 Hardware Environment .................................................................................... 14
2.1.1 Required Hardware............................................................................................................... 14
2.1.2 Boot disk configuration ........................................................................................................ 15
2.2 Software Environment...................................................................................... 17
2.2.1 Required software................................................................................................................. 17
2.2.2 Optional software.................................................................................................................. 18
Chapter 3 Precautions ....................................................................... 21
Chapter 4 Building an OS Boot Environment ................................ 25
4.1 Creating a Boot Disk on a Disk Array Device ................................................. 27
4.1.1 Creating a boot disk using a network install server............................................................... 27
4.1.1.1 Creating a network install server ...................................................................................... 284.1.1.2 Configuring a network install server................................................................................. 294.1.1.3 Labeling disks................................................................................................................... 344.1.1.4 Configuring Custom JumpStart ........................................................................................ 364.1.1.5 Configuring the Fibre Channel boot code......................................................................... 414.1.1.6 Executing network installation ......................................................................................... 45
4.1.2 Creating a boot disk by copying an existing boot disk residing on an internal disk.............. 46
4.1.2.1 Getting ready to copy the boot disk to a disk array device ............................................... 474.1.2.2 Creating a boot disk.......................................................................................................... 474.1.2.3 Editing mount table information....................................................................................... 514.1.2.4 Configuring the Fibre Channel boot code......................................................................... 534.1.2.5 Resetting the server .......................................................................................................... 534.1.2.6 Booting from a disk array device...................................................................................... 54
4.2 Making the Path to a Boot Disk Redundant......................................................55
4.2.1 Installing the Enhanced Support Facility...............................................................................55
4.2.2 Configuring the ETERNUS Multipath Driver.......................................................................56
4.2.2.1 Single system (non-cluster system) ...................................................................................564.2.2.2 Cluster system...................................................................................................................64
4.3 Boot Disk Mirroring .........................................................................................74
4.3.1 Mirroring by PRIMECLUSTER GDS ..................................................................................74
4.3.2 Notes on using PRIMECLUSTER........................................................................................76
4.3.2.1 Cluster system building procedure....................................................................................76
Chapter 5 Backing Up and Restoring Boot Disks........................... 77
5.1 Backing Up/Restoring after Booting OS from a Network ................................78
5.1.1 Backup procedure .................................................................................................................78
5.1.2 Restore procedure .................................................................................................................80
5.2 Backing Up/Restoring after Booting the OS from an Internal Disk .................83
5.2.1 Backup procedure .................................................................................................................83
5.2.2 Restore procedure .................................................................................................................85
Appendix A Boot Device Setup Commands....................................... 89
A.1 Command Executable on the OS ......................................................................89
A.2 Command Executable on the OBP ...................................................................94
Appendix B Checking the Fibre Channel Card BootCode Version Number .................................................. 103
B.1 Checking on the OS ........................................................................................103
B.2 Checking on the OBP......................................................................................103
Appendix C Recording SAN Boot Setting Information ................. 105
Appendix D Making Fixes to Setting Files after a Boot Failure.... 107
D.1 If the OS Has Been Installed As Per Section 4.1.1, "Creating a bootdisk using a network install server" ................................................................107
D.2 If the OS Has Been Installed As Per Section 4.1.2, "Creating a bootdisk by copying an existing boot disk residing on an internal disk"...............109
Appendix E Fibre Channel Driver/Boot Code Auto-Target Binding Functions ............................................ 111
E.1 Fibre Channel Driver Auto-Target Binding Function .................................... 111
E.2 Fibre Channel Boot Code Auto-Target Binding Function ............................. 113
Appendix F SAN Boot release procedure........................................ 114
F.1 ETERNUS Multipath Driver.......................................................................... 114
Appendix G Updating the Fibre Channel Card Boot Code ............. 117
1
Chapter 1 Overview
The term SAN Boot refers to having an operating system (OS) or application stored in external SANstorage, not on an internal disk in a server, and starting (that is, booting) the OS or application from there.
This document describes the workflow for building a SAN Boot environment, in which a Fibre Channelcard is mounted on a server to start the OS from ETERNUS storage system (RAID).
Having an OS boot disk on an external disk array device offers the following advantages:
1. Enhanced availability
Use of a high-reliability disk array (RAID) device
Enhanced reliability results from managing the boot disk on a disk array (RAID) device.
Backup/restore work made more efficient
Use of the disk copy feature of a disk array device drastically cuts the period during whichbusiness is stopped for backing up and restoring system volumes. The CPU load incurred whilebacking up and restoring the system volumes is also reduced.
For more details, see Section 1.1.4, "ETERNUS Configuration required to use the AdvancedCopy function."
2
NoteETERNUS SF AdvancedCopy Manager (ACM) or PRIMECLUSTER GDS Snapshot isrequired to use the disk copy feature (Advanced Copy feature) of ETERNUS (disk array device).
2. Greater ease of operation management
System volumes kept under consolidated management
Boot disks that were previously spread among multiple servers can be kept under consolidatedmanagement, as they are contained in a single disk array device.
Development environment generation management
Multiple development environments maintained on a single disk array device can be switched,as required. This eliminates the need to keep a server for each development environment,thereby allowing the number of servers and operational workload to be reduced.
3
3. Better maintainability
Handling of disk failures made simpler
If a disk (system volume) fails, the system administrator notifies a service engineer in charge ofthe fact and lets the engineer replace the disk for the system to recover automatically. Thesystem administrator's workload is thus lightened.
Applying patches made simpler
Use of the disk copy feature of a disk array device trims the period of business stop for backingup system volumes before patches are applied on them. With the OS configured to boot from abackup volume (*1), if a problem occurs after applying the patches, the system can be rolledback to its status in effect prior to applying the patch by rebooting the server and switching theboot volume. For more details, see Section 1.1.4, "ETERNUS Configuration required to use theAdvanced Copy feature."
(*1) PRIMECLUSTER GDS Snapshot provides this functionality with a simple commandoperation.
4
5
1.1 Configuration PatternsBoot the OS from an external disk array (RAID) device using Fibre Channel Cards in the patterns of theFibre Channel connection configuration shown below. Points to watch concerning each configurationpattern are also given.
1.1.1 Basic configuration
1. Using a disk array device from a single server
Implement a multipath configuration pattern based on the ETERNUS multipath disk driver,in which at least two routes of Fibre Channel connections are maintained between theserver and each disk array device.
If disk swaps occur, they could result in degraded application disk access performance.Actions that help prevent disk swaps include adding to the memory mounted on the serveror reducing the rate of memory usage by applications.
If a server that is not equipped with an internal disk is used, an install server is required toinstall the OS and recover the boot disk.
6
2. Using a disk array device from multiple servers
Fabric connection
FC-AL direct connection
Implement a multipath configuration pattern based on the ETERNUS multipath disk driver,in which at least two routes of Fibre Channel connections are maintained between eachserver and the disk array device.
If disk swaps occur, they could result in degraded application disk access performance.Actions that help prevent disk swaps include adding to the memory mounted on each serveror reducing the rate of memory usage by applications.
7
When a server panics, the other servers having their boot disks placed in the same RAIDgroup as the server might suffer degraded boot disk access performance for several tens ofseconds. See "2.1.2 Boot disk configuration."
If a server that is not equipped with an internal disk is used, a dedicated install server forinstalling the OS and recovering the boot disk is required.
1.1.2 Disk array device mirroring configuration basedon PRIMECLUSTER GDS
With the 1Gbps/2Gbps Fibre Channel card (PW008FC3), one card was required per path.With the single-channel 4Gbps Fibre Channel card (SE0X7F11x) and the dual-channel4Gbps Fibre Channel card (SE0X7F12x), in contrast, two Fibre Channel cards can becombined to build a disk array device mirroring configuration based on PRIMECLUSTERGDS, because multiple disk array device configurations in which boot disk is recognizedare enabled.
Implement a multipath configuration pattern based on the ETERNUS multipath disk driver,in which at least two routes of Fibre Channel connections are maintained between eachserver and a disk array device.
If disk swaps occur, they could result in degraded application disk access performance.Actions that help prevent disk swaps include adding to the memory mounted on each serveror reducing the rate of memory usage by applications.
When a server panics, the other servers having their boot disk placed in the same RAIDgroup as the server may suffer degraded boot disk access performance for several tens ofseconds.
8
If a server that is not equipped with an internal disk is used, a dedicated install server forinstalling the OS and recovering the boot disk is required separately.
1.1.3 Cluster configuration based on PRIMECLUSTER
A cluster system based on a SAN Boot environment may also be built.
1. Cluster configuration based on a single disk array device
9
2. Cluster configuration based on multiple disk array devices
1.1.4 ETERNUS Configuration required to use theAdvanced Copy function
On a traditional system on which the OS is booted from an internal disk, business needs to be stopped fora long period of time while backup/restore is executed using tape devices.
10
If backup/restore is executed using the ETERNUS Advanced Copy function (OPC/EC) in a SAN Bootenvironment, business can be carried out without stopping business while disk device copying is inprogress, thus drastically cutting the duration of business stop.
Backup/restore using the Advanced Copy function is operable in two ways: one using ETERNUS SFAdvancedCopy Manager (ACM) and one using PRIMECLUSTER GDS Snapshot (GDS Snapshot).
Using ETERNUS SF AdvancedCopy Manager (ACM)
Use of ACM offers these advantages:
Shorter period of business stop
Higher efficiency in backing up multiple servers (consolidated management)
Using PRIMECLUSTER GDS Snapshot (GDS Snapshot)
Use of GDS Snapshot offers the following advantages:
Shorter period of business stop
Restore made easier on a soft-mirroring configuration based on PRIMECLUSTERGDS
11
12
The table below contains a summary description of the features of ACM and those of GDS Snapshot.
Select the method suited to your system requirements.
NoteWhere system volumes are managed using PRIMECLUSTER GDS, their backup/restore is operable withboth ACM and GDS Snapshot, but use of GDS Snapshot is recommended for system volumes built on asoft-mirroring configuration.
ACM and GDS Snapshot features
: Advantage
ETERNUS SF AdvancedCopyManager (ACM)
PRIMECLUSTER GDS Snapshot
Operational server -
A dedicated server for executing thebackup/restore operations is required inaddition to the server to be backed upand restored.
No separate server is required, becausethe backup/restore operations areexecuted on the server to be backed upand restored.
Backup operation
Shut down the server to be backed uponce, and launch OPC from its backupserver. The server can be rebooted toresume suspended business withoutwaiting for physical copying tocomplete.
-
Reboot the server to be backed up insingle-user mode, and launch OPC.When physical copying completes,reboot the server in multi-user mode toresume suspended business.
Restore operation(mirroring byPRIMECLUSTERGDS notimplemented)
Shut down the server to be restoredonce, and launch OPC. The server canthen be rebooted to resume suspendedbusiness without waiting for physicalcopying to complete.
Suspended business can be resumed byrebooting the server to be restored andswitching the boot volume.
Restore operation(mirroring byPRIMECLUSTERGDS implemented)
-
The mirrored disk needs to bedisconnected and then reconnectedbefore and after the OPC restoreoperation.
-
When OPC physical copyingcompletes, switch back to the originalboot volume to resume suspendedbusiness.
Multi-serverbackup efficiency
Backup volumes on multiple serverscan be placed under consolidatedmanagement from a single backupserver.
-
The backup/restore operations areexecuted on each individual server.
Function forbooting from abackup volume
-
The mount point needs to be changedby editing the vfstab file.
The function can be easily configuredusing command.
13
For more feature details, refer to the ETERNUS SF AdvancedCopy Manager and PRIMECLUSTERGDS manuals.
Instructions on how to verify the progress of OPC physical copying are included in the ETERNUS SFAdvancedCopy Manager and PRIMECLUSTER GDS manuals.
14
Chapter 2 Hardware/SoftwareConfiguration
The hardware configuration and the software configuration described in this chapter are prerequisite tobooting the OS from an external disk array device using Fibre Channel cards.
2.1 Hardware Environment
2.1.1 Required Hardware
In this manual, the following hardware organization and the software configuration are supported.
Type Device name Remarks
Server SPARC EnterpriseT1000/T2000/T5120/T5140/T5220/T5240/T5440/M3000/M4000/M5000/
M8000/M9000
Fibre Channelcard mountedon servers
Single-channel 4 Gbps Fibre Channel card [SE0X7F11x]
Dual-channel 4 Gbps Fibre Channel card [SE0X7F12x]
Disk arraydevice
ETERNUS2000 Model 50/100/200
ETERNUS DX60/DX80
ETERNUS3000 Model 80/100/300/500/700
ETERNUS4000 Model 80/100/300/400/500/600
ETERNUS DX410/DX440
ETERNUS6000 Model 500/700/900/1100
ETERNUS8000 Model 700/800/900/1100/1200
ETERNUS DX8100/DX8400/DX8700
Fibre Channelswitch
ETERNUS SN200 Series
If a server that is not equipped with an internal disk is used, an install server is required separately toinstall and recover the OS.
15
2.1.2 Boot disk configuration
SAN Boot has a system disk placed on the ETERNUS disk array. In addition to system disks, a varietyof disk volumes are placed on the disk array device. The way these disk volumes are placed on the diskarray device could affect the performance of access to system disks residing on other servers and userdata disks such as databases. To keep disk access performance unaffected, take the following precautionin implementing the disk configuration:
Do not place in the same RAID group a system disk area and areas (system and data disks)that are accessible from other servers.
16
If any other kind of disk configuration is used, the following problem may occur:
If multiple system disks are placed in the same RAID group, the occurrence of disk swapscould result in degraded disk access performance for volumes residing in the same RAIDgroup.
17
Placing a system disk and a shared data area in the same RAID group would degradeaccess performance for the shared data area by several tens of seconds during a memorydump access session triggered by a server panic.
2.2 Software EnvironmentThis section describes the required software.
2.2.1 Required software
The following software components are required:
Software Version Remarks
Solaris(TM) operating system Solaris(TM) 10 operating system Solaris 10 11/06 or higher
It is necessary to applySolaris 10 8/07 or higherwhen SPARC Enterprise
T5120/T5140/T5220 is used.
It is necessary to applySolaris 10 5/08 or higherwhen SPARC EnterpriseT5240/T5440 is used.
It is necessary to apply
18
Solaris 10 10/08 or higherwhen SPARC EnterpriseM3000 is used.
FUJITSU PCI Fibre Channeldriver
4.0 or higher
ETERNUS Multipath Driver 2.0.1 or higher Patch 914267-04 or laterneeds to be installed.
It is necessary to apply sincepatch 914267-05 whenETERNUS2000 is used.
It is necessary to apply sincepatch 914267-09 whenETERNUS DX60/DX80 isused.
It is necessary to apply sincepatch 914267-11 whenETERNUS DX4xx/DX8xxxis used.
When EMPD 2.0.1 is used inSolaris 10 10/08, then pleaseapply patch 914267-07.
2.2.2 Optional software
PRIMECLUSTER GDS is required to implement system volume mirroring between disk arraydevices.
Software Version Remarks
PRIMECLUSTER GDS 4.2 orhigher
Patch 914423-03 or later needs to be installed.
It is necessary to apply since patch 914423-05when ETERNUS2000 is used.
It is necessary to apply since patch 914423-10when ETERNUS DX60/DX80/DX4xx/DX8xxx isused.
It is necessary to apply since patch 914423-08when create an alternate boot environment.
One of the following PRIMECLUSTER products is required to build a cluster system:
19
Software Version Remarks
PRIMECLUSTER Enterprise Edition
PRIMECLUSTER HA Server
PRIMECLUSTER Lite Pack
4.2 orhigher
This product is bundled with PRIMECLUSTERGDS.
Patches 901201-06 or later, 914325-03 or later,914468-01 or later and 914499-01 or later need tobe installed.
ETERNUS SF AdvancedCopy Manager is required to execute backup/restore using ETERNUSAdvanced Copy function.
Software Version Remarks
ETERNUS SF AdvancedCopyManager
13.0 orhigher
Both agents and the manager are Solaris 10-ready.
ETERNUS SF AdvancedCopyManager tape agent license
13.0 orhigher
Required to back up to tape with ETERNUS SFAdvancedCopy Manager.
One copy is required for each tape agent installed.
Tape agents are Solaris 10-ready.
ETERNUS SF AdvancedCopy Manager tapeserver option is required to be installed on tapeserver.NOTE: The tape server is only Solaris 8 or Solaris9-ready.
For more details on ETERNUS SF AdvancedCopy Manager, refer to the ETERNUS SFAdvancedCopy Manager manual.
PRIMECLUSTER GDS Snapshot is required to create snapshots of a system volume, or createan alternate boot environment using the ETERNUS Advanced Copy function or thePRIMECLUSTER GDS copy function.
Software Version Remarks
PRIMECLUSTER GDS Snapshot 4.2 orhigher
A PRIMECLUSTER product bundled withPRIMECLUSTER GDS is required to use thisproduct.
Patch 914457-02 or later needs to be installed.
It is necessary to apply since patch 914457-03when ETERNUS DX60/DX80/DX4xx/DX8xxx isused.
20
It is necessary to apply since patch 914457-03when using OS since Solaris 10 10/08 is used orthe patch (137137-09 or later) suitable sinceSolaris 10 10/08 is used.
For more details on PRIMECLUSTER GDS Snapshot, refer to "PRIMECLUSTER GlobalDisk Service Guide."
21
Chapter 3 Precautions
1. The single-channel 4 Gbps Fibre Channel card (SE0X7F11x) and dual-channel 4 Gbps FibreChannel card (SE0X7F12x) support a boot code to allow the OS to be booted from a diskarray device connected to the Fibre Channel card. The Fibre Channel cards come with thisboot code disabled by default. The boot code needs to be enabled before the OS can be bootedusing a file channel card.
Enabling or disabling the boot function on either port of the dual-channel 4Gbps Fibre Channelcard (SE0X7F12x) will automatically impart the same setting to the other port. It is notpossible to make a change to the setting of either port alone and not both.
For information about enabling and disabling the boot function on Fibre Channel cards, seeAppendix A.2 1 fjpfca-set-bootfunction.
2. Be sure to install the OS on the boot disk using one of the methods described in Chapter 4.
It is not possible to copy any boot disk that has been used on another host by using the ddcommand or using ETERNUS EC (Equivalent Copy) or OPC (One Point Copy) and use it.
Boot disks cannot be created in any procedure other than that described in Chapter 4.
3. Keep a record of the boot configuration information set on the Fibre Channel card.
The boot configuration information is required when the Fibre Channel card is replaced,because it needs to be reproduced on the new Fibre Channel card.
For information on the kinds of information to be recorded, see Appendix C, "Recording SANBoot Setting Information."
4. Fujitsu recommends not placing in the same RAID group the boot disks for different hosts.
For more details, see Section 2.1.2, "Boot disk configuration."
22
5. Do not create a huge file or a large number of files in /tmp (tmpfs).
When creating files in /tmp (tmpfs), take care to ensure that the size of space used by /tmp(tmpfs) does not exceed the installed memory size.
If the size of space used by /tmp (tmpfs) should exceed the installed memory size as a result ofhaving created a huge file or a large number of files in /tmp (tmpfs), the system could slowdown due to a lack of sufficient memory available to it.
This precaution also applies when an internal disk is used as a boot disk.
6. When using PRIMECLUSTER or PRIMECLUSTER GDS to build a system implementation,see Section 4.3.2, "Notes on using PRIMECLUSTER," for additional tips.
7. The access path settings based on WWN (World Wide Port Name) described below arerecommended for the ETERNUS disk array device and the ETERNUS SN200 Fibre Channelswitch.
These settings remove the need to reconfigure the following devices for using the resourcemanagement software, Systemwalker Resource Coordinator, facilitating the transition processand speed up the transition.
ETERNUS disk array device
Configure as host table or enable the host affinity feature with regard to the ETERNUS FC-CAport, and set the WWN of the Fibre Channel card used as a host World Wide Name. Refer toeach Server Connection Guide for the setting of the ETERNUS disk array device.
ETERNUS SN200 Fibre Channel switch
Using the WWN of the Fibre Channel card and that of each disk array device, configure a one-to-one WWN zoning plan, which establishes zoning using the WWN of the host HBA port andthat of the FC-CA port.
8. Disks bearing the EFI (Extensible Firmware Interface) disk label cannot be used as a boot disk.
The EFI disk label supports disks larger than 1Tbyte on a system running the 64-bit Solariskernel. The OS cannot be booted from a disk bearing the EFI disk label, though.
23
9. The warning message shown below is displayed when a multipath is created. This messagemay be ignored.
This message reports that the disk array device has received a SCSI RESET that is issued atmultipath build time and does not relate to any disk array device or server operation.
This message will be displayed only when a new multipath is built or additions are made to afleet of disk array devices. If message monitoring is implemented, disable the monitoringprocess temporarily or simply ignore the output message when it is displayed.
WARNING: /pci@1,700000/fibre-channel@0/sd@10,0 (sd805) :
Error for Command: write (10) Error Level: Retryable
Requested Block: 5651696 Error Block: 5651696
Vendor: FUJITSU Serial Number: 0000080115
Sense Key: Unit Attention
ASC: 0x29 (bus device reset message occurred) , ASCQ: 0x3, FRU: 0x1
10. If a system disk is mirrored using PRIMECLUSTER GDS, the following message may bedisplayed at boot time, but simply ignore it, because there is no actual problem with the system:
NOTICE: "forceload: drv/<driver name> appears more than once in /etc/system.
* One of mplb, mplbt and sd appears in place of <driver name>.
This message is displayed if the setting forceload is defined in duplicate in /etc/system file. Toleave this message hidden, delete later occurrences of the setting of forceload.
forceload: drv/mplb
~
forceload: drv/mplb Delete this line.
24
11. In configuring the Fibre Channel driver, the link speed (transmission line speed) setting can beset to automatic selection to facilitate connectivity, but the expected link speed (especially forthe 4Gbps) may not be attained depending on the connection timing. Enter the link speedsetting in /kernel/drv/fjpfca.conf.
Example: Set fjpfca0 to a link speed of 4 Gbps.
port=
"fjpfca0:nport:sp4";
For instructions how to set a link speed, refer to "FUJITSU PCI Fibre Channel Guide."
12. Be sure to configure LUN 0(Host Logical Unit Number 0) on ETERNUS. The LUN 0 is usedto recognize ETERNUS by Fibre Channel boot code.
13. When you install the Solaris 10 10/08 or higher by the installation server, the installation servershould work on Solaris 10 10/08 environment or higher or on Solaris 10 environment that137137-09 or later patch are applied; otherwise driver packages cannot be installed.
The Sun Microsystems Inc. is not recommending the construction of the ZFS file systemenvironment on the disk array device. Refer to "Solaris ZFS Administration Guide" for details.
14. If you backup and restoring data in ZFS environment, you should use Solaris 10 10/08environment or on Solaris 10 environment that 137137-09 or later patch are applied; otherwisethere is a possibility that the import of the system volume cannot be done.
15. In the SAN Boot environment by the ZFS file system, the mirroring of the system disk byPRIMECLUSTER GDS cannot be done.
16. When you restore the boot disk, you should use the boot block on the restore destination devicefor the boot block creating.
17. In boot code V12L30 installed on fibre channel card (SE0X7F11F,SE0X7F12F), the OSbooting process might hang up for the following conditions.
When connecting it with 8G port of the SN200 series
When booting it after about 25 days or more pass from the power supply turning on of theserver device
This problem is corrected in boot code V12L40. Referring to "Appendix G Updating the FibreChannel Card Boot Code", update the boot code.
When hanging up occurs, turn the power off, and then turn the power on, and boots again.
18. The composition of the system volume larger than 512Gbyte is not supported.
25
Chapter 4 Building an OS BootEnvironment
This chapter describes the following topic:
Building an OS boot environment from a disk array device
Before performing this procedure, set up a disk array device to make available a lun in which to create aboot disk.
If a server that is not equipped with an internal disk is used, only the method described in Section 4.1.1,"Creating a boot disk using a network install server," can be used. Since the FUJITSU PCI Fibre Channeldriver is not contained in Solaris OS, OS boot environment cannot be built using CD/DVD of Solaris OS.
Boot environments cannot be built in any procedure other than that described in this guide.
In Section 4.1.1, "Creating a boot disk using a network install server," the method of identifying the diskarray device to become a boot disk can be automated. This method facilitates the job of configuring theFibre Channel driver and the Fibre Channel boot code. For more details, see Section 4.1.1.2,"Configuring a network install server."
If a disk array device is identified through automatic setting, set zoning with the FC Switch to disable aFibre Channel card from connecting to multiple disk array devices. In an environment in which a FibreChannel card is allowed to connect to multiple disk array devices, implement disk array deviceidentification manually (manual setting).
If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, see Appendix G,"Updating the Fibre Channel Card Boot Code," update the boot code.
For information on building a cluster environment, see Section 4.3.2, "Notes on usingPRIMECLUSTER."
The workflow for building an OS boot environment is as follows:
26
Install the OS as instructed in Section 4.1, "Creating a Boot Disk on a Disk Array Device," and create aboot disk on a disk array device.
Then, boot the OS in single-path mode as instructed in Section 4.1.2.6, "Booting from a disk arraydevice."
Lastly, define a multipath configuration as per Section 4.2, "Making the Path to a Boot Disk Redundant."
27
4.1 Creating a Boot Disk on a Disk Array DeviceThere are two ways to create a boot disk on a disk array device, as follows:
1. Creating a boot disk using a network install server
2. Creating a boot disk by copying an existing boot disk residing on an internal disk (only if theserver is equipped with an internal disk)
4.1.1 Creating a boot disk using a network installserver
Executing a network install requires an install server, in addition to the host (install machine) that uses adisk array device as a boot device.
Work with the install machine from the install machine console. Work with the install server from aterminal, which is marked as "(INSTALL SERVER)" in the examples appearing in this guide.
28
4.1.1.1 Creating a network install server
Configure an install server to execute a network install. For more details on the work of creating aninstall server, refer to " Solaris x x/x Release and Installation Collection" at docs.sun.com.
If there are multiple hosts that use a disk array device as a boot device each, an OS image needs to becreated for each host. The hosts may share a single install image, however, in these situations:
1. Hosts use AL direct Fibre Channel connection, sharing the same values of target ID and maxthrottle.
2. Hosts use automatically configured disk array devices on a FC switch connection, sharing thesame values of target ID and max throttle.
Note) Multiple hosts which are the same architecture type can share the same OS install image,but the same OS install image can not be used for the other architecture type. SPARC EnterpriseM3000/M4000/M5000/M8000/M9000 (which are sun4u) can share the same OS install image.SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440 (sun4v) can share thesame OS install image.
A sample flow for creating an image of Solaris10 OS is described below.
In this flow, /export/install/Solaris10_hostname is used as the name of a sample directory in which anOS image is created.
1. Become a superuser on the install server.
(INSTALL SERVER) % su - <RETURN>
Password: password
2. Make a directory in which to create an OS image.
(INSTALL SERVER) # mkdir /export/install <RETURN>
(INSTALL SERVER) # cd /export/install <RETURN>
(INSTALL SERVER) # mkdir Solaris10_hostname <RETURN>
Include the host name hostname of the install machine in the created directory name to allowmanagement by host.
3. Mount the Solaris 10 Operating System DVD-ROM.
(INSTALL SERVER)# cd /cdrom/cdrom0/s0/Solaris_10/Tools <RETURN>
(INSTALL SERVER)# ./setup_install_server /export/install/Solaris10_hostname<RETURN>
4. When copying of the Solaris 10 Operating System completes, demount the DVD-ROM.
29
(INSTALL SERVER)# cd / <RETURN>
(INSTALL SERVER)# eject cdrom <RETURN>
4.1.1.2 Configuring a network install server1. Register the IP/MAC address of an installation target machine on the network install server.
○ Register the IP address of an installation target machine
Edit the /etc/hosts file with a text editor.
When an IP address is "192.168.1.1", it becomes the following.
192.168.1.1 hostname
○Register the MAC address of an installation target machine
Edit the /etc/ethers file with a text editor.
When an IP address is "0:80:17:28:1:f8", it becomes the following.
0:80:17:28:1:f8 hostname
2. For boot the installation target machine from the network, execute add_install_client commandon the network install server.
An add_install_client parameter changes with machine models.
SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440
(INSTALL SERVER) # cd /export/install/Solaris10_hostname/Solaris_10/Tools/ <RETURN>(INSTALL SERVER) # ./add_install_client hostname sun4v <RETURN>
SPARC Enterprise M3000/M4000/M5000/M8000/M9000
(INSTALL SERVER) # cd /export/install/Solaris10_hostname/Solaris_10/Tools/ <RETURN>(INSTALL SERVER) # ./add_install_client hostname sun4u <RETURN>
3. Install the Fibre Channel driver in the OS install image on the network install server.
This step makes the disk array device connected to the Fibre Channel card identifiable. Mount"FUJITSU PCI Fibre Channel 4.0" on the CD-ROM drive in the network install server and dothe following:
The process of installing the Fibre Channel driver varies with each target machine model.
Solaris 10 5/08 or older
SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440
(INSTALL SERVER) # cd /cdrom/cdrom0 <RETURN>
(INSTALL SERVER) # bin/pfcapkgadd.sh -R/export/install/Solaris10_hostname/Tools/Boot/ -p sun4v <RETURN>
SPARC Enterprise M3000/M4000/M5000/M8000/M9000
(INSTALL SERVER) # cd /cdrom/cdrom0 <RETURN>
30
(INSTALL SERVER) # bin/pfcapkgadd.sh -R/export/install/Solaris10_hostname/Tools/Boot/ -p sun4u <RETURN>
Solaris 10 10/08 or higher
(1) Create the working directory for unpacking miniroot.
(INSTALL SERVER) # mkdir /tmp/work <RETURN>
(2) Unpack the miniroot to the work directory using the root_archive(1M) command.
If /tmp/work/tmp/AdDrEm.lck file is not exist, following procedure ignoring.
(INSTALL SERVER) # /boot/solaris/bin/root_archive unpackmedia/export/install/Solaris10_hostname /tmp/work <RETURN>
(INSTALL SERVER) # rm /tmp/work/tmp/AdDrEm.lck <RETURN>
- The following messages might be displayed when root_archive command is executed. Youmight ignore these messages.
umount: /tmp/mnt29984 busy
rmdir: directory "/tmp/mnt29984": Directory is a mount point or in use
lofiadm: could not unmap file /export/install/Solaris10_hostname/boot/sparc.miniroot:Device busy
rmdir: directory "/tmp/mnt29984": Directory is a mount point or in use
(3) Install the Fibre Channel driver in the working directory.
SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440
(INSTALL SERVER) # cd /cdrom/cdrom0 <RETURN>
(INSTALL SERVER) # bin/pfcapkgadd.sh -R /tmp/work/ -p sun4v <RETURN>
SPARC Enterprise M3000/M4000/M5000/M8000/M9000
(INSTALL SERVER) # cd /cdrom/cdrom0 <RETURN>
(INSTALL SERVER) # bin/pfcapkgadd.sh -R /tmp/work/ -p sun4u <RETURN>
(4) Pack the working directory.
(INSTALL SERVER) # mkdir -p /tmp/media/Solaris_10 <RETURN>
(INSTALL SERVER) # /tmp/work/boot/solaris/bin/root_archive packmedia /tmp/media/tmp/work <RETURN>
The error messages might be displayed when root_archive command is executed. You mightignore these messages.
31
(5) Copy the file in /tmp/media directory to the installation image on the install server.
Target device path name for "umount -f" and "lofiadm -d" commands can confirm by "df -k"command.
(INSTALL SERVER) # cd /tmp/media <RETURN>
(INSTALL SERVER) # find boot Solaris_10/Tools/Boot | cpio -pdum/export/install/Solaris10_hostname <RETURN>
(INSTALL SERVER) # umount -f /dev/lofi/1 <RETURN>
(INSTALL SERVER) # lofiadm -d /dev/lofi/1 <RETURN>
The error messages might be displayed when root_archive command is executed. You mightignore these messages.
4. Verify the correspondence between the Fibre Channel card location and driver instancenumber.
Boot the installation target machine from the network in single-user mode, with the -s optionspecified.
ok boot net -s <RETURN>
5. Execute the following to verify the correspondence between the Fibre Channel card devicepath and the driver instance:
# grep fjpfca /tmp/root/etc/path_to_inst <RETURN>
"/pci@1,700000/fibre-channel@0" 0 "fjpfca"
"/pci@2,600000/fibre-channel@0" 1 "fjpfca"
Each line of this command listing contains device path, instance number and instance name inthis order.
In the example above, the drive instance of the Fibre Channel card mounted at device path"/pci@1,700000/XXXX@0" is fjpfca0.
The correspondence between device paths and server slot positions is described in the relevantserver’s user's guide. If any other server is used or if the relevant user's guide is not available forreference, check the correspondence in the following way:
6. Flash the LED on the Fibre Channel card associated with the driver instance. The flashingLED lets you identify the Fibre Channel card location for a single-channel 4Gbps FibreChannel card (SE0X7F11x), or the Fibre Channel card location and the port positionassociated with the driver instance for a dual-channel 4Gbps Fibre Channel card
32
(SE0X7F12x). The LED at fjpfca0 can be flashed as instructed below. The LINK LED willflash for 3 minutes.
# /usr/sbin/FJSVpfca/fc_adm -l fjpfca0 <RETURN>
To stop the flashing of the LED, enter Ctrl-c (press the c key while holding down Ctrl key).
For information on using fc_adm, refer to "FUJITSU PCI Fibre Channel Guide."
7. Shutdown the OS on the installation target machine and return to OBP.
# /usr/sbin/shutdown -g0 -i0 -y <RETURN>
8. Configure the definition files that are used for booting the OS at network install time.
When Solaris 10 10/08 or higher is installed, the following procedures are executedbeforehand.
Solaris 10 10/08 or higher
Unpack the miniroot to the work directory using the root_archive(1M) command.
(INSTALL SERVER) # /boot/solaris/bin/root_archive unpackmedia/export/install/Solaris10_hostname /tmp/work <RETURN>
- The following messages might be displayed when root_archive command is executed. Youmight ignore these messages.
umount: /tmp/mnt29984 busy
rmdir: directory "/tmp/mnt29984": Directory is a mount point or in use
lofiadm: could not unmap file /export/install/Solaris10_hostname/boot/sparc.miniroot:Device busy
rmdir: directory "/tmp/mnt29984": Directory is a mount point or in use
The files require the following setting on the network install server:
Fibre Channel driver setting file
Solaris 10 5/08 or older
{OS install path on the install server}/Tools/Boot/kernel/drv/fjpfca.conf
Solaris 10 10/08 or higher
/tmp/work/kernel/drv/fjpfca.conf
Configure the Fibre Channel driver setting file (fjpfca.conf) to make the disk arraydevice on which to create a boot disk identifiable to the Fibre Channel driver. On adirect connection (FC-AL), the specification of port, fcp-auto-bind-function and fcp-bind-target is not mandatory. For more information on configuring fjpfca.conf, referto "FUJITSU PCI Fibre Channel Guide."
33
There are two ways to make a disk array device identifiable on a fabric connection(using the FC Switch): automatic setting and manual setting. Automatic setting formaking a disk array device identifiable, when selected, offers the following benefits:
1. If multiple hosts use a disk array device as a boot device each, they can stillshare and use the Solaris OS install image on the install server.
2. The Fibre Channel driver is made easier to configure.
These two methods of identifying a disk array device are described below.
a. [Automatic setting]
Example: Set a fabric connection as a topology, a link speed of 4 Gbps andautomatic setting as method of disk array device identification for fjpfca0.
port="fjpfca0:nport:sp4";
fcp-auto-bind-function=1;
port Define a connection topology type and a link speed.
fcp-auto-bind-function Configure a disk array device to be identifiedautomatically.
For more information about the automatic identification function, see AppendixE, "Fibre Channel Driver/Boot Code Auto-Target Binding Functions."
If a disk array device is identified through automatic setting, set zoning withthe FC Switch to disable a Fibre Channel card from connecting to multiple diskarray devices. In an environment in which a Fibre Channel card is allowed toconnect to multiple disk array devices, allow disk array devices to be identifiedthrough manual setting.
For information on how to set zoning with the FC Switch, refer to the relevantFC switch’s manual.
b. [Manual setting]
Example: Set a fabric connection as a FC switch topology, a link speed of 4Gbps and binding of a disk array device with target ID 16 for fjpfca0.
port="fjpfca0:nport:sp4";
fcp-bind-target="fjpfca0t16:0x210000c0004101d9";
port Define a connection topology type and a link speed.
fcp-bind-target Specify the target WWN.
Target driver setting file
Solaris 10 5/08 or older
{OS install path on the install server}/Tools/Boot/kernel/drv/sd.conf
Solaris 10 10/08 or higher
/tmp/work/kernel/drv/sd.conf
34
Configure the target driver setting file (sd.conf) to make the logical unit (LU) ofthe disk array device on which to create a boot disk identifiable. Define only theboot disk on the disk array device. If an identifiable logical unit of the disk arrayunit is already defined, the entry may be bypassed.
Example: Identify target ID 16 and logical unit 0.
name="sd" class="scsi" target=16 lun=0;
When Solaris 10 10/08 or higher is installed, the following procedures areexecuted beforehand.
(1) Pack the working directory.
(INSTALL SERVER) # mkdir -p /tmp/media/Solaris_10 <RETURN>
(INSTALL SERVER) # /tmp/work/boot/solaris/bin/root_archive packmedia/tmp/media /tmp/work <RETURN>
The error messages might be displayed when root_archive command is executed.You might ignore these messages.
(2) Copy the file in /tmp/media directory to the installation image on the install server.
Target device path name for "umount -f" and "lofiadm -d" commands can confirm by"df -k" command.
(INSTALL SERVER) # cd /tmp/media <RETURN>
(INSTALL SERVER) # find boot Solaris_10/Tools/Boot | cpio -pdum/export/install/Solaris10_hostname <RETURN>
(INSTALL SERVER) # umount -f /dev/lofi/1 <RETURN>
(INSTALL SERVER) # lofiadm -d /dev/lofi/1 <RETURN>
The error messages might be displayed when root_archive command is executed.You might ignore these messages.
9. If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, seeAppendix G, "Updating the Fibre Channel Card Boot Code," update the boot code.
4.1.1.3 Labeling disks1. Boot the installation target machine from the network in single user mode
ok boot net -s <RETURN>
2. Create a disk label in the lun that is used as a boot disk by carrying out the format (1M)command, and then check the size of the lun.
35
# format <RETURN>
Searching for disks...done
AVAILABLE DISK SELECTIONS:0. c7t16d0 <FUJITSU-ETERNUS-4000 cyl 1038 alt 2 hd 64 sec 256>
/pci@1,700000/fibre-channel@0/sd@10,0Specify disk (enter its number): 0<RETURN>selecting c7t16d0[disk formatted]Disk not labeled. Label it now? y <RETURN>FORMAT MENU:
disk - select a disktype - select (define) a disk typepartition - select (define) a partition tablecurrent - describe the current diskformat - format and analyze the diskrepair - repair a defective sectorlabel - write label to the diskanalyze - surface analysisdefect - defect list managementbackup - search for backup labelsverify - read and display labelssave - save new disk/partition definitionsinquiry - show vendor, product and revisionvolname - set 8-character volume name!<cmd> - execute <cmd>, then returnquit
format> partition <RETURN>
PARTITION MENU:0 - change '0' partition1 - change '1' partition2 - change '2' partition3 - change '3' partition4 - change '4' partition5 - change '5' partition6 - change '6' partition7 - change '7' partition
select - select a predefined tablemodify - modify a predefined partition tablename- name the current tableprint - display the current tablelabel - write partition map and label to the disk!<cmd> - execute <cmd>, then returnquit
partition> print <RETURN>Current partition table (original):Total disk cylinders available: 4254 + 2 (reserved cylinders)
Part Tag Flag Cylinders Size Blocks0 root wm 0 – 15 128.00MB (16/0/0) 2621441 swap wu 16 – 31 128.00MB (16/0/0) 2621442 backup wu 0 – 4253 33.23GB (4254/0/0) 696975363unassigned wm 0 0 (0/0/0) 0
36
4unassigned wm 0 0 (0/0/0) 05unassigned wm 0 0 (0/0/0) 06 usr wm 32 – 4253 32.98GB (4222/0/0) 691732487unassigned wm 0 0 (0/0/0) 0
format> quit <RETURN>
3. Shutdown the OS on the installation target machine and return to OBP.
# /usr/sbin/shutdown -g0 -i0 -y <RETURN>
4.1.1.4 Configuring Custom JumpStart
Configure Solaris custom jump start to install a driver package. If Custom JumpStart is used, a driverpackage is automatically installed and configured at the same time as Solaris is installed.
Perform this procedure on the install server.
1. Make a Custom JumpStart directory.
Make and share a jumpstart directory on the install server.
(INSTALL SERVER)# mkdir /jumpstart <RETURN>
(INSTALL SERVER)# share -F nfs -o ro,anon=0 /jumpstart <RETURN>
2. Copy a driver package, patch and install file.
Copy a driver package, patch and install file to the jumpstart directory on the install server.
Copy the CD image of FUJITSU PCI Fibre Channel to the jumpstart directory on the installserver.
(INSTALL SERVER)# mkdir /jumpstart/FJPFCA <RETURN>
(INSTALL SERVER)# cd /cdrom/cdrom0 <RETURN>
(INSTALL SERVER)# find . | cpio -pumd /jumpstart/FJPFCA <RETURN>
When FUJITSU PCI GigabitEthernet/FUJITSU ULTRA LVD SCSI Host Bus Adapter Driverinstall, the following execute.
FUJITSU PCI GigabitEthernet 3.0 Update1 or higher
37
Copy the CD image of FUJITSU PCI GigabitEthernet 3.0 Update1 or higher to thejumpstart directory on the install server.
(INSTALL SERVER)# mkdir /jumpstart/fjgi <RETURN>
(INSTALL SERVER)# cp -p /cdrom/cdrom0/install /jumpstart/fjgi/. <RETURN>
(INSTALL SERVER)# cp -p /cdrom/cdrom0/admin /jumpstart/fjgi/. <RETURN>
(INSTALL SERVER)# cp -pr /cdrom/cdrom0/FJSVgid_3.0/10/* /jumpstart/fjgi/.<RETURN>
FUJITSU PCI GigabitEthernet 4.0 or higher
Copy the CD image of FUJITSU PCI GigabitEthernet 4.0 or higher to the jumpstartdirectory on the install server.
(INSTALL SERVER)# mkdir /jumpstart/fjgi <RETURN>
(INSTALL SERVER)# cp -p /cdrom/cdrom0/install_v4 /jumpstart/fjgi/. <RETURN>
(INSTALL SERVER)# cp -p /cdrom/cdrom0/admin /jumpstart/fjgi/. <RETURN>
(INSTALL SERVER)# cp -pr /cdrom/cdrom0/FJSVgid_4.0/10/* /jumpstart/fjgi/.<RETURN>
FUJITSU ULTRA LVD SCSI Host Bus Adapter Driver
Copy the CD image of FUJITSU ULTRA LVD SCSI Host Bus Adapter Driver to thejumpstart directory on the install server.
(INSTALL SERVER)# mkdir /jumpstart/fjulsa <RETURN>
(INSTALL SERVER)# cp -p /cdrom/cdrom0/install /jumpstart/fjulsa/. <RETURN>
(INSTALL SERVER)# cp -p /cdrom/cdrom0/admin /jumpstart/fjulsa/. <RETURN>
(INSTALL SERVER)# cp -pr /cdrom/cdrom0/ultra_lvd_driver/10/*/jumpstart/fjulsa/. <RETURN>
3. Copy a JumpStart sample
Copy the JumpStart sample file from the OS install image.
(INSTALL SERVER)# cp -r/export/install/Solaris10_hostname/Solaris_10/Misc/jumpstart_sample/* /jumpstart<RETURN>
4. Edit the profile
Edit the /jumpstart/profile file with a text editor.
38
Create a profile to meet the installation target machine configuration. Create one as instructedin "Solaris Installation Guide: Custom JumpStart and Advanced Installations."
The setting procedure of profile are different by the file system of the system disk.
Profile sample setting : UFS file system
install_type initial_install # The install_type parameter is required. SpecifyInitial_install.
system_type server #Specifies server as system_type.
partitioning explicit # Specifies explicit as partitioning.
cluster SUNWCXall #Specifies SUNWCXall (Entire Software Group Plus OEM Support)as an installed OS cluster.
filesys c7t16d0s1 4096 swap #Assigns a swap file system space of 4096 MB toc7t16d0s1.
filesys c7t16d0s0 free / #Assigns the rest of the disk space to c7t16d0s0.
Profile sample setting : ZFS file system
install_type initial_install # The install_type parameter is required. Specify Initial_install.
system_type server # Specifies server as system_type.
partitioning explicit # Specifies explicit as partitioning.
cluster SUNWCXall # Specifies SUNWCXall (Entire Software Group Plus OEMSupport) as an installed OS cluster.
pool newpool auto auto auto c7t16d0s0 # The size is automatically allocated in c7t16d0s0and newpool is made. The size of swap and dump made in newpool is automatically allocated.
bootenv installbe bename sxce_xx # The boot file system is made by name(newpool/ROOT/sxce_xx) of sxce_xx.
5. Copy a finish script sample
Copy a sample of the finish script from the FJPFCA directory to the /jumpstart directory asfinish.
(INSTALL SERVER)# cp/jumpstart/FJPFCA/FJPFCA4.0/tool/FJPFCA_jumpstart_finish.sample/jumpstart/finish <RETURN>
6. Edit the finish script
Edit /jumpstart/finish with a text editor. Edit the following parameters:
JUMPSTART_HOST Specifies the host name or IP address of the install server.
39
JUMPSTART_DIR Specifies the directory in which the JumpStart setting file is stored.Edit this parameter only when a directory other than /jumpstart is used.
When FUJITSU PCI GigabitEthernet/FUJITSU ULTRA LVD SCSI Host Bus Adapter Driverinstall , add the following contents. In the following examples, it adds it under"PF_ARCH=`uname -m`".
FUJITSU PCI GigabitEthernet 3.0 Update1 or higher
${MNT}/fjgi/install -R /a -d ${MNT}/fjgi -p "$PF_ARCH"
FUJITSU PCI GigabitEthernet 4.0 or higher
${MNT}/fjgi/install_v4 -R /a -d ${MNT}/fjgi -p "$PF_ARCH"
FUJITSU ULTRA LVD SCSI Host Bus Adapter Driver
${MNT}/fjulsa/install -R /a -d ${MNT}/fjulsa -p "$PF_ARCH"
Finish script sample setting
#!/bin/sh
### Edit here ###
JUMPSTART_HOST=
JUMPSTART_DIR=/jumpstart
### End of edit ###
### MAIN ###
MNT=/a/mnt
mount -F nfs ${JUMPSTART_HOST}:${JUMPSTART_DIR} ${MNT}
PF_ARCH=`uname -m`
${MNT}/fjgi/install -R /a -d ${MNT}/fjgi -p "$PF_ARCH"
${MNT}/fjulsa/install -R /a -d ${MNT}/fjulsa -p "$PF_ARCH"
${MNT}/FJPFCA/bin/pfcapkgadd.sh -R /a -p "$PF_ARCH"
# Copy fjpfca.conf
if [ -f /kernel/drv/fjpfca.conf ]
then
echo "copying fjpfca.conf "
cp /kernel/drv/fjpfca.conf /a/kernel/drv/fjpfca.conf
40
COPY_STATUS="$?"
if [ "$?" != "0" ]
then
echo "ERROR: fjpfca.conf copy failed."
fi
else
echo "NOTICE: /kernel/drv/fjpfca.conf does not exists."
fi
## Copy sd.conf
if [ -f /kernel/drv/sd.conf ]
then
echo "copying sd.conf "
cp /kernel/drv/sd.conf /a/kernel/drv/sd.conf
COPY_STATUS="$?"
if [ "$?" != "0" ]
then
echo "ERROR: sd.conf copy failed."
fi
else
echo "NOTICE: /kernel/drv/sd.conf does not exists."
fi
umount ${MNT}
7. Edit the rules file
Edit the /jumpstart/rules file with a text editor. Specify a profile and finish script used for eachhost in the rules file.
The rules file comes with a number of sample settings by default. Comment them out, becausethey are not required.
Append the following to the rules file:
hostname <install machine hostname> - profile finish
8. Check and enable the rules file
Execute the check command to enable the rules file.
41
Solaris 10 5/08 or older
(INSTALL SERVER)# cd /jumpstart <RETURN>
(INSTALL SERVER)# /jumpstart/check -p /export/install/Solaris10_hostname -r rules<RETURN>
Solaris 10 10/08 or higher
(INSTALL SERVER)# cd /jumpstart <RETURN>
(INSTALL SERVER)# /jumpstart/check -p /tmp/media -r rules <RETURN>
If the check command is executed and the following error messages are displayed, thecheck command execute again after the following procedures.
Error message:
ERROR: /tmp/media is not a valid Solaris 2.x CD image
(INSTALL SERVER)# cd /tmp/media/Solaris_10/Tools/Boot <RETURN>
(INSTALL SERVER)# bzcat lu.cpio.bz2 | cpio -idum <RETURN>
(INSTALL SERVER)# ls usr/sbin/install.d/chkprobe <RETURN>
usr/sbin/install.d/chkprobe
(INSTALL SERVER)# cd /jumpstart <RETURN>
(INSTALL SERVER)# /jumpstart/check -p /tmp/media -r rules <RETURN>
9. Enter network boot settings.
The setting varies with each installation target machine model.
SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440
(INSTALL SERVER)# /export/install/Solaris10_hostname/Solaris_10/Tools/add_install_client -i install machine IP address -e install machine mac address -s installserver host name:/export/install/Solaris10_hostname -c install server host name:/jumpstartinstall machine host name sun4v <RETURN>
SPARC Enterprise M3000/M4000/M5000/M8000/M9000
(INSTALL SERVER)# /export/install/Solaris10_hostname/Solaris_10/Tools/add_install_client -i install machine IP address -e install machine mac address -s installserver host name:/export/install/Solaris10_hostname -c install server host name:/jumpstartinstall machine host name sun4u <RETURN>
4.1.1.5 Configuring the Fibre Channel boot code
Configure the Fibre Channel boot code required to boot the OS.
42
Perform this procedure on the install machine.
1. If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, execute thefollowing command:
ok setenv auto-boot? false <RETURN>
ok reset-all <RETURN>
If you use SPARC Enterprise M3000/M4000/M5000/M8000/M9000, set the server modeswitch to service mode and execute the following command:
ok reset-all <RETURN>
2. Make sure that Fibre Channel card is identified on the OBP. Check the physical path name ofthe slot in which the Fibre Channel card is mounted.
Example of having a single-channel 4 Gbps Fibre Channel card (SE0X7F11x) and a dual-channel 4 Gbps Fibre Channel card (SE0X7F12x) mounted on a server
ok show-devs <RETURN>
/pci@1,700000
/pci@2,600000
**
/openprom
/chosen
/packages
/pci@1,700000/fibre-channel@0 *physical path name of single-channel 4Gbps FibreChannel card
/pci@2,600000/fibre-channel@0 *physical path name of dual-channel 4Gbps FibreChannel card port0
/pci@2,600000/fibre-channel@0,1 *physical path name of dual-channel 4Gbps FibreChannel card port1
/mc@0,0/bank@0,c0000000
/mc@0,0/bank@0,80000000
3. With the boot code enabled on the Fibre Channel card used for booting the OS, restart theserver. Move to the Fibre Channel card physical path (/pci@1,700000/fibre-channel@0)confirmed in Step 2 before executing a setting command. There is no need to enable the bootcode on those Fibre Channel cards that do not use the boot feature.
ok cd /pci@1,700000/fibre-channel@0 <RETURN> Move to the Fibre Channel cardphysical path
ok ENABLE fjpfca-set-bootfunction <RETURN> Enable the boot feature.
ok reset-all <RETURN> Initialize the server.
43
Perform this step on all cards used for booting the OS.
4. After restarting the server, view information about the disk array devices connected to it.
Example: fabric connection
ok cd /pci@1,700000/fibre-channel@0 <RETURN>
ok PROBE fjpfca-info <RETURN>
Target - DID 10500 210000e00040101d9 FUJITSU-E4000-0000
Target - DID 10600 210000e00040101da FUJITSU-E4000-0000
Targets residing on a fabric connection may appear as "-," but this is of no concern.
5. Configure the disk array device to make it identifiable with the Fibre Channel boot code.
There are two ways to make a disk array device identifiable on a fabric connection (using theFC Switch): automatic setting and manual setting. Automatic setting for making a disk arraydevice identifiable, when selected, offers the following benefits:
(1) If multiple hosts use a disk array device as a boot device each, they still can share and use theSolaris OS install image on the install server.
(2) The Fibre Channel driver is made easier to configure.
These two methods of identifying a disk array device are described below.
a. [Automatic setting]
Remove the disk array device setting in the Fibre Channel boot code. Execute thefollowing command:
ok cd /pci@1,700000/fibre-channel@0 <RETURN> Move to the Fibre Channel cardphysical path
ok ENABLE fjpfca-all-target-cancel <RETURN>
fjpfca-all-target-cancel: Delete bind target parameter ...
If a disk array device is identified through automatic setting, set zoning with the FCSwitch to disable a Fibre Channel card from connecting to multiple disk array devices. Inan environment in which a Fibre Channel card is allowed to connect to multiple disk arraydevices, allow disk array devices to be identified through manual setting.
b. [Manual setting]
In the Fibre Channel switch (fabric connection) environment, execute the fjpfca-bind-target command to define a disk array device that is identifiable to the Fibre Channel bootcode.
The WWPN displayed in Step 4 or the DID value is required at this time. There is noneed to execute the fjpfca-bind-target command in the FC-AL environment, because thedisk array device is configured automatically.
Move to the Fibre Channel card physical path (/pci@1,700000/fibre-channel@0) confirmedin Step 2 before carrying out a configuration command.
44
o Definition by WWPN
ok cd /pci@1,700000/fibre-channel@0 <RETURN> Move to the Fibre Channel cardphysical path
ok 10 target-wwpn 210000e0004101d9 fjpfca-bind-target <RETURN>
fjpfca-bind-target: Change bind target parameter
o Definition by DID
ok cd /pci@1,700000/fibre-channel@0 <RETURN> Move to the Fibre Channel cardphysical path
ok 10 target-did 10500 fjpfca-bind-target <RETURN>
fjpfca-bind-target: Change bind target parameter .
* The setting is not required in the FC-AL (Private loop) environment.
6. Other settings
The connection topology, link speed, and server and disk array device power supply interlockwait time can be modified as required. The connection topology and link speed are set to "auto"(automatic setting) by default.
For more details, see Appendix A, " Boot Device Setup Commands."
For example, if automatic setting is not used on a connection with a Fibre Channel switch readyfor 2 Gbps, enter the following:
ok 2g fjpfca-set-linkspeed <RETURN>
ok nport fjpfca-set-topology <RETURN>
Confirm the settings.
ok fjpfca-output-prop <RETURN>
boot function: ENABLE
topology : N_Port
link-speed : 2G
boot wait time: DISABLE (interval time: DISABLE/ boot wait msg: DISABLE)
bind-target: Target_ID=16,WWN=0x210000c0004101d9
For example, if a power supply interlocking boot wait time of 1200 seconds (20 minutes) is seton a direct connection with a disk array device ready for 2 Gbps, enter the following:
ok 2g fjpfca-set-linkspeed <RETURN>
ok al fjpfca-set-topology <RETURN>
ok d# 1200 fjpfca-set-boot-wait-time <RETURN>
45
Confirm the settings.
ok fjpfca-output-prop <RETURN>
boot function: ENABLE
topology : AL
link-speed : 2G
boot wait time: 1200 sec (interval time: DISABLE/ boot wait msg: DISABLE
bind-target: Target_ID=16,WWN=0x210000c0004101d9
7. Execute the following command to reset:
ok reset-all <RETURN>
4.1.1.6 Executing network installation
Execute the following command on the OBP of the installation target host:
ok boot net - install <RETURN>
Subsequently, proceed with the install process as directed by onscreen guidance.
When the installation of the OS has completed and a prompt is displayed, proceed to the next procedure.
If Solaris10 is installed on a server that supports a graphics card to use a bitmapped display, right-clickanywhere on the screen to display a menu, from which open a terminal and proceed further. The message"Click <Reboot> to continue" is displayed, but ignore it for now.
When the network install completes, the OS will automatically boot itself from the disk array device.
46
4.1.2 Creating a boot disk by copying an existing bootdisk residing on an internal disk
This section explains how to copy a boot disk that has already been created on an internal disk orelsewhere and use it to create a boot disk on a disk array device. Before getting started, make sure thatthe server has been started from the OS stored on the internal disk and a connection with the disk arraydevice has been configured via a Fibre Channel card.
A single-path connection with the disk array device will do for now. Convert the connection to amultipath implementation according to Section 4.2, "Making the Path to a Boot Disk Redundant."
Perform this procedure entirely from the install machine console.
This procedure does not allow a boot disk to be created on a target device that has been connected by theauto-target bind feature of the Fibre Channel driver. Be sure to configure a target device with fcp-bind-target to create a boot disk on it.
For more information about configuring fcp-bind-target, refer to "FUJITSU PCI Fibre Channel Guide."
47
4.1.2.1 Getting ready to copy the boot disk to a disk arraydevice
1. Execute the format command or the like to confirm the disk array device on which to create aboot disk.
# format <RETURN>
Searching for disks...done
AVAILABLE DISK SELECTIONS
0. c7t16d0 <FUJITSU-ETERNUS-4000 cyl 1038 alt 2 hd 64 sec 256>
/pci@1,700000/fibre-channel@0/sd@10,0
Check for the availability of sufficient disk space at the boot disk creation destination andsufficient associated partition space when compared with the source to copy from. If a slice hasnot been created in the lun at the boot disk creation destination, execute the format command tocreate one.
2. If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, seeAppendix G, "Updating the Fibre Channel Card Boot Code," update the boot code.
4.1.2.2 Creating a boot diskCreating a boot disk, the procedure is different depending on the composition of the following file
systems.
Procedure for copying internal disk (UFS file system) to disk array device (UFS file system).
Procedure for copying internal disk (UFS file system) to disk array device (ZFS file system).
Procedure for copying internal disk (ZFS file system) to disk array device (ZFS file system).
The environment that can be made is different according to the operating system.
Operating System Method of creating boot disk
Solaris 10 5/08 or older Procedure for copying internal disk (UFS file system) to disk array
device (UFS file system).
Solaris 10 5/08 or older and
patch 137137-09 or later
Procedure for copying internal disk (UFS file system) to disk arraydevice (UFS file system).
Procedure for copying internal disk (UFS file system) to disk arraydevice (ZFS file system).
Solaris 10 10/08 or higher
Procedure for copying internal disk (UFS file system) to disk arraydevice (UFS file system).
Procedure for copying internal disk (UFS file system) to disk arraydevice (ZFS file system).
48
Procedure for copying internal disk (ZFS file system) to disk arraydevice (ZFS file system).
It explains each procedure as follows.
Procedure for copying internal disk (UFS file system) to disk array device (UFS file system).
1. Migrate the system to the obp environment.
# /usr/sbin/shutdown -y -i0 <RETURN>
Launch the system in single-user mode.
ok boot -s <RETURN>
2. Write a boot block.
Write a boot block to the system block (using a path that has been verified by carrying out theformat command or the like).
# installboot /usr/platform/`uname -m`/lib/fs/ufs/bootblk /dev/rdsk/c7t16d0s0 <RETURN>
3. Create a file system.
# newfs -v /dev/rdsk/c7t16d0s0 <RETURN>
4. Execute the mount command (mount the LUN for the boot disk on the identified disk arraydevice).
# mount -F ufs /dev/dsk/c7t16d0s0 /mnt <RETURN>
5. Copy the boot disk to the disk array device.
# ufsdump 0f - /dev/rdsk/c0t0d0s0 | ( cd /mnt; ufsrestore rf -) <RETURN>
The command above, when carried out, copies data located in directories other than /mnt toRAID.
6. Copy other partitions.
If /var and /opt have been defined as separate partitions, repeat Steps 3 to 5 above for eachpartition.
49
Procedure for copying internal disk (UFS file system) to disk array device (ZFS file system).
1. Migrate the system to the obp environment.
# /usr/sbin/shutdown -y -i0 <RETURN>
Launch the system in single-user mode.
ok boot -s <RETURN>
2. Create the ZFS file system on a disk array device.
# zpool create rootpool c7t16d0s0 <RETURN>
# zfs create rootpool/rootfs <RETURN>
# zfs create rootpool/rootfs/s10_1008 <RETURN>
# zfs create -V 2G rootpool/swap <RETURN>
# zfs create -V 2G rootpool/dump <RETURN>
# zfs set mountpoint=legacy rootpool/rootfs/s10_1008 <RETURN>
By way of the example, the file system used to the root(/) make a name of rootfs, the swap areais made by 2GB and the dump areas are made by 2GB. The mount point of rootpool/rootfs set tolegacy.
3. Write a boot block.
Write a boot block to the system block (using a path that has been verified by carrying out theformat command or the like).
# installboot -F zfs /usr/platform/`uname -i`/lib/fs/zfs/bootblk /dev/rdsk/c7t16d0s0<RETURN>
4. Execute the mount command.
Mount the ZFS file system made on the disk array device according to Step 2.
# mount -F zfs rootpool/rootfs/s10_1008 /mnt <RETURN>
5. Copy the boot disk to the disk array device.
# ufsdump 0f - /dev/rdsk/c0t0d0s0 | ( cd /mnt; ufsrestore rf -) <RETURN>
6. Copy other partitions.
If /var and /opt have been defined as separate partitions, repeat Steps 4 to 5 above for eachpartition.
50
Procedure for copying internal disk (ZFS file system) to disk array device (ZFS file system).
1. Migrate the system to the obp environment.
# /usr/sbin/shutdown -y -i0 <RETURN>
Launch the system in single-user mode.
ok boot -s <RETURN>
2. Create the ZFS file system on a disk array device.
# zpool create rootpool c7t16d0s0 <RETURN>
# zfs create rootpool/rootfs <RETURN>
# zfs create -V 2G rootpool/swap <RETURN>
# zfs create -V 2G rootpool/dump <RETURN>
# zfs set mountpoint=legacy rootpool/rootfs <RETURN>
By way of the example, the file system used to the root(/) make a name of rootfs, the swap areais made by 2GB and the dump areas are made by 2GB. The mount point of rootpool/rootfs setto legacy.
3. Write a boot block.
Write a boot block to the system block (using a path that has been verified by carrying out theformat command or the like).
# installboot -F zfs /usr/platform/`uname -i`/lib/fs/zfs/bootblk /dev/rdsk/c7t16d0s0<RETURN>
4. Create a snapshot.
Create the snapshot of root(/) of the internal disk.
# zfs list <RETURN>
NAME USED AVAIL REFER MOUNTPOINT
rootpool 142K 78.7G 19K /rootpool
rootpool/dump 2G 78.7G 16K -
rootpool/rootfs 18K 78.7G 18K legacy
rootpool/swap 2G 78.7G 16K -
rpool 5.84G 61.1G 94K /rpool
rpool/ROOT 4.81G 61.1G 18K legacy
51
rpool/ROOT/s10_1008 4.81G 61.1G 4.81G /
rpool/dump 512M 61.1G 512M -
rpool/export 32.0M 61.1G 20K /export
rpool/export/home 32.0M 61.1G 32.0M /export/home
rpool/swap 512M 61.6G 10.8M -
# zfs snapshot rpool/ROOT/s10_1008@snapshot <RETURN>
5. Copy the boot disk to the disk array device.
# mkdir /backup <RETURN>
# zfs send rpool/ROOT/s10_1008@snapshot > /backup/s10_1008.img <RETURN>
# zfs receive rootpool/rootfs/s10_1008 < /backup/s10_1008.img <RETURN>
6. Copy other partitions.
If /var have been defined as separate partitions, repeat Steps 4 to 5 above for each partition.
7. Execute the mount command.
# mount -F zfs rootpool/rootfs/s10_1008 /mnt <RETURN>
4.1.2.3 Editing mount table information
Enter the access path to the disk array device on which a boot disk has been created in /mnt/etc/vfstab.The boot disk LUN being mounted on /mnt. In the UFS file system and the ZFS file system, the settingmethod is different.
In the UFS file system, all the access paths is added, comment out access paths that are of no use.
Example: SAN Boot environment by UFS file system
#device device mount FS fsck mount mount
#to mount to fsck point type pass at boot options
#
#/dev/dsk/c0t0d0s3 - - swap - no -
#/dev/dsk/c0t0d0s0 /dev/rdsk/c0t0d0s0 / ufs 1 no-
#/dev/dsk/c0t0d0s1 /dev/rdsk/c0t0d0s1 /var ufs 1 no-
52
fd - /dev/fd fd - no -
/proc - /proc proc - no -
/dev/dsk/c7t16d0s3 - - swap - no -
/dev/dsk/c7t16d0s0 /dev/rdsk/c7t16d0s0 / ufs 1 no-
/dev/dsk/c7t16d0s1 /dev/rdsk/c7t16d0s1 /var ufs 1 no-
..
In the ZFS file system, only swap is added, comment out access paths that are of no use.
Example: SAN Boot environment by ZFS file system(the root device is not set.)
#device device mount FS fsck mount mount
#to mount to fsck point type pass at boot options
#
#/dev/dsk/c0t0d0s3 - - swap - no -
#/dev/dsk/c0t0d0s0 /dev/rdsk/c0t0d0s0 / ufs 1 no -
#/dev/dsk/c0t0d0s1 /dev/rdsk/c0t0d0s1 /var ufs 1 no -
fd - /dev/fd fd - no -
/proc - /proc proc - no -
/dev/zvol/dsk/rootpool/swap - - swap - no -
If a disk array device has a LU other than its boot disk defined in sd.conf, remove the definition from/mnt/kernel/drv/sd.conf.
The following procedures are executed continuously only when the SAN Boot environment by the ZFSfile system is constructed.
1. The mountpoint property is changed to the root path (/).
# zfs set mountpoint=/ rootpool/rootfs/s10_1008 <RETURN>
- It advances to the following procedure ignoring it though the following messages aredisplayed when the mountpoint property is set to the root path (/).
cannot mount '/': directory is not empty
property may be set but unable to remount filesystem
53
2. Bootfs setting.
# zpool set bootfs=rootpool/rootfs/s10_1008 rootpool <RETURN>
4.1.2.4 Configuring the Fibre Channel boot code
Configure the Fibre Channel boot code to allow the OS to be booted from a disk array device.
Example: Set on fjpfca0.
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -b ENABLE <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -c /kernel/drv/fjpfca.conf <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -v <RETURN>
boot_function : ENABLE
topology : N_Port
link-speed : 4G
boot wait time : DISABLE ( interval time : DISABLE , boot wait msg : DISABLE )
bind-target: Target_ID=16,WWN=0x210000c0004101d9
4.1.2.5 Resetting the server
If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, execute the followingcommands depending on the system status.
The OS is booting
execute the following command:
# /usr/sbin/eeprom auto-boot?=false <RETURN>
# /usr/sbin/shutdown -i0 -g0 -y <RETURN>
After migrating to the obp environment, execute the following command:
ok reset-all <RETURN>
In the obp environment
Execute the following command:
ok setenv auto-boot? false <RETURN>
ok reset-all <RETURN>
54
The server is turned off
Turn on the server and execute “The OS is booting” or “In the obp environment” depending onthe system status.
If you use SPARC Enterprise M3000/M4000/M5000/M8000/M9000 set the server mode switch toservice mode and execute the following commands depending on the system status.
The OS is booting
Set the server mode switch to service mode and execute the following command:
# /usr/sbin/shutdown -i0 -g0 -y <RETURN>
After migrating to the obp environment, execute the following command:
ok reset-all <RETURN>
In the obp environment
Set the server mode switch to service mode and then execute the following command:
ok reset-all <RETURN>
The server is turned off
Set the server mode switch to service mode and then turn on the server.
4.1.2.6 Booting from a disk array device
Specify a disk array device from which boot the OS.
ok boot /pci@1,700000/fibre-channel@0/disk@10,0 <RETURN> (*1)
Boot device: /pci@1,700000/fibre-channel@0/disk@10,0 File and args: kernel/sparcv9/unix
(*1) The value (in the example above, "10,0") that follows "disk" specified by boot denotes thetarget_id/LUN. It must be the same as the value of target_id/LUN of the disk array device that isidentified by the Fibre Channel driver after the OS boots. In the FC-AL environment, specify thevalue of target_id that is displayed when fjpfca-info is executed.
The values of target_id and LUN need be specified in hexadecimal at boot time.
55
4.2 Making the Path to a Boot Disk RedundantThis section explains how to make the path to a boot device based on the ETERNUS Multipath Driverredundant.
4.2.1 Installing the Enhanced Support Facility
Install Enhanced Support Facility if it is yet to be installed. If it is already installed, proceed to the nextprocedure.
1. Specify the disk array device and boot the OS in single-user mode.
ok boot /pci@1,700000/fibre-channel@0/disk@10,0 -s <RETURN>
Boot device: /pci@1,700000/fibre-channel@0/disk@10,0 File and args: kernel/sparcv9/unix
2. Install Enhanced Support Facility as instructed in "Enhanced Support Facility Install Guide."
3. Restart the host in the following way:
# /usr/sbin/shutdown -i0 -g0 -y <RETURN>
4. After migrating to the obp environment, execute the following command:
ok reset-all <RETURN>
5. Specify the disk array device and boot the OS.
ok boot /pci@1,700000/fibre-channel@0/disk@10,0 <RETURN>
Boot device: /pci@1,700000/fibre-channel@0/disk@10,0 File and args: kernel/sparcv9/unix
56
4.2.2 Configuring the ETERNUS Multipath Driver
This section explains how to configure the ETERNUS Multipath Driver to make the path to a boot deviceredundant.
About target ID of the equipment used as a boot device, it recommends setting it as the same value withtwo Fibre Channel cards which constitute a multipath.
4.2.2.1 Single system (non-cluster system)1. Specify a disk array device and boot the OS from it to install the ETERNUS Multipath Driver.
Launch the host from the boot disk on the disk array device. Then, install ETERNUSMultipath Driver as instructed in the "ETERNUS Multipath Diver Install Guide." When theinstall completes, respond with "y" at the following prompt to let the grmpdautoconf commandexecute automatically to proceed to the multipath building process in Step2.
Do you want to make a multipath configuration now?
If the ETERNUS Multipath Driver package has already been installed, execute grmpdautoconfto proceed to the multipath building process in Step 2.
# /usr/sbin/grmpdautoconf <RETURN>
2. Execute grmpdautoconf to build a multipath.
Work with grmpdautoconf interactively. For more details, refer to "ETERNUS MultipathDriver User's Guide." During this interactive session, make the following choices:
Select "Manual selection" m in response to the automatic/manual path selection prompt.
Select an access path automatically or manually?
** If automatic selection is selected, all access paths marked "New" are registered with thesystem.
** Select automatic selection if an access path has been properly selected at ETERNUS andswitch setup.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and device installation.
Manual selection ---> 'm'
Automatic selection ---> 'a'
Quit ---> 'q'
Enter a key. [m,a,q] m <RETURN>
57
Different manual path selection screens are displayed depending on which method of diskarray device identification by the Fibre Channel driver has been selected, automatic settingor manual setting.
a. [Automatic setting]
Select a start-up path on the manual path selection screen.
Adapter Switch ETERNUSStatus
instance WWN WWN product
-----+-------------------------------------+-----+------------------------------------------------------+-----
[ ] 1 fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0New
[ ] 2 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.
** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] 1 2 <RETURN>
Adapter Switch ETERNUSStatus
instance WWN WWN product
-----+-------------------------------------+-----+------------------------------------------------------+-----
[*] 1 fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0New
[*] 2 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
58
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] x <RETURN>
The setting of the path (disk array device) selected above is reflected in the Fibre Channeldriver setting file (/kernel/drv/fjpfca.conf). Once this step completes, the setting of the diskarray device that is identified by the Fibre Channel driver is fixed, so the following setting in theFibre Channel driver setting file can be removed:
fcp-auto-bind-function=1;
b. [Manual setting]
The wwn entered in fjpfca.conf appears as "Exist" or "AL." Leave all other paths unselected,selecting "Confirmed (x)" for them.
Adapter Switch ETERNUSStatus
instance WWN WWN product
-----+----------------------------------+-----+----------------------------------------------------+-----
fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0Exist
[ ] 1 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.
59
** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] 1 <RETURN>
Adapter Switch ETERNUSStatus
instance WWN WWN product
-----+----------------------------------+-----+------------------------------------------------------+------
fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0Exist
[*] 1 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.
** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] x <RETURN>
60
When SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440 is being used, itis necessary to choose the mode of an access path. Select mplb mode at the prompt for accesspath mode selection. The ETERNUS Multipath Driver supports two different multipathaccess modes: Solaris standard mode, in which the multipath is accessed from a Solarisstandard special file, and mplb mode, in which the multipath is accessed from a mplb specialfile as in the past.
In the Solaris10 environment, select mplb mode for the access path mode. Solaris standardmode does not work in the SAN Boot environment.
When SPARC Enterprise M3000/M4000/M5000/M8000/M9000 is being used, it is notnecessary to choose access path mode.
Choose an access special file between the following:
Solaris standard special file (/dev/[r]dsk/c*t*d*s*)
mplb special file (/dev/FJSVmplb/[r]dsk/mplb*s*)
/dev/[r]dsk/c*t*d*s* ---> 's'
/dev/FJSVmplb/[r]dsk/mplb*s* ---> 'm'
Enter a key. [s, m] m <RETURN>
3. Check the device path name of the boot device. The grmpdautoconf command carried out inStep 2 displays a combination of a multipath management special file and a selected accessspecial file. Use the output of the ls command to identify the boot disk and the physical devicepath name of each configuration path. Physical device path names thus found are used in Steps6 and 9.
# ls -l <Boot disk slice 0 > <RETURN>
# ls -l < Configuration path slice 2 > <RETURN>
Assume that grmpdautoconf has delivered the following output listing:
*** Phase 1: read mplb.conf ***
*** Phase 2: read /dev ***
*** Phase 3: read /devices ***
*** Phase 4: compare mplb.conf and /devices ***
Path : Action : Element path : LUN : Storage
mplb0 : new : c2t16d0s2 c13t16d0s2 : 0 : E40004641- 130011 :
mplb1 : new : c2t16d1s2 c13t16d1s2 : 1 : E40004641- 130011 :
mplb2 : new : c2t16d2s2 c13t16d2s2 : 2 : E40004641- 130011 :
The boot disk and the paths that configure it up are as follows:
61
Boot disk /dev/FJSVmplb/rdsk/mplb0s0
Configuration path /dev/rdsk/c2t16d0s2
/dev/rdsk/c13t16d0s2
Execute the ls command to check the device path names.
# ls -l /dev/FJSVmplb/rdsk/mplb0s0 <RETURN>
lrwxrwxrwx 1 root root 36 Aug 29 12:05 /dev/FJSVmplb/rdsk/mplb0s0 -> (line wrapping)
../../../devices/pseudo/mplb@0:a,raw <RETURN>
^^^^^^^^^^^^^^^
# ls -l /dev/rdsk/c2t16d0s2 <RETURN>
lrwxrwxrwx 1 root root 58 Aug 29 17:13 /dev/rdsk/c2t16d0s2 -> (line wrapping)
../../devices/pci@1,700000/fibre-channel@0/mplbt@10,0:c,raw <RETURN>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
# ls -l /dev/rdsk/c13t16d0s2 <RETURN>
lrwxrwxrwx 1 root root 58 Aug 29 17:13 /dev/rdsk/c13t16d0s2 -> (line wrapping)
../../devices/pci@2,600000/fibre-channel@0/mplbt@10,0:c,raw <RETURN>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
4. Set a Fibre Channel driver link speed.
Edit the Fibre Channel driver setting file (/kernel/drv/fjpfca.conf) to set a Fibre Channel linkspeed.
In configuring the Fibre Channel driver, the link speed setting can be set to automatic selectionfor the sake of easier connectivity. The expected link speed may not be attained depending onthe connection status. Therefore, set the highest transmission rate available in the environment.
Example: Set fjpfca0 to a link speed of 4 Gbps.
port=
"fjpfca0:nport:sp4";
For more information about configuring fjpfca.conf, refer to "FUJITSU PCI Fibre ChannelGuide."
5. Load the Fibre Channel boot codes that are used to access the boot disk with a disk arraydevice boot setting.
Note
62
The setting to the Fibre Channel boot code used by boot is unnecessary now.
Example: Set on fjpfca1.
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca1 -b ENABLE <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca1 -c /kernel/drv/fjpfca.conf <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca1 -v <RETURN>
boot_function : ENABLE
topology : N_Port
link-speed : 4G
boot wait time : DISABLE ( interval time : DISABLE , boot wait msg : DISABLE )
bind-target: Target_ID=16,WWN=0x210000e0004101d9
6. Correct the system settings to meet the multipath implementation.
SAN Boot environment by the UFS file systema. Root device setting (/etc/system)
Edit /etc/system file to set rootdev and forceload. For the rootdev setting, set the boot diskphysical device name as found in Step 3 above, excluding "../../devices" at the beginning and",raw" at the end.
When the setting concerning forceload to each driver exists in /etc/system file, an additionalsetting need not be done.
rootdev: /pseudo/mplb@0:a
forceload: drv/mplbt
forceload: drv/mplb
forceload: drv/sd
b. Mount information setting (/etc/vfstab)
Edit /etc/vfstab file to rewrite each entry to a path name after the multipath implementation.
/dev/FJSVmplb/dsk/mplb0s0 /dev/FJSVmplb/rdsk/mplb0s0 / ufs 1 no -
/dev/FJSVmplb/dsk/mplb0s3 - - swap - no –
c. An access setup of a boot disk (/kernel/drv/sd.conf).
When target ID of the equipment used as a boot device has differed with two Fibre Channelcards which constitute a multipath, target ID about all the paths of a boot disk may not be writtento sd.conf. In that case, the following setup is needed.
○ Edit /kernel/drv/sd.conf
The definition of target ID of the path used as a boot disk is added.
63
Example: The definition of "Target ID = 18" is added.
name=”sd” class=”scsi” target=18 lun=0;
○ Reconstruction of sd driver
# touch /reconfigure <RETURN>or
# update_drv -f sd <RETURN>
SAN Boot environment by the ZFS file systema. Root device setting (/etc/system)
Edit /etc/system file to set forceload.
forceload: drv/mplb
forceload: drv/sd
b. An access setup of a boot disk (/kernel/drv/sd.conf).
When target ID of the equipment used as a boot device has differed with two Fibre Channelcards which constitute a multipath, target ID about all the paths of a boot disk may not be writtento sd.conf. In that case, the following setup is needed.
○ Edit /kernel/drv/sd.conf
The definition of target ID of the path used as a boot disk is added.
Example: The definition of "Target ID = 18" is added.
name=”sd” class=”scsi” target=18 lun=0;
○ Reconstruction of sd driver
# touch /reconfigure <RETURN>or
# update_drv -f sd <RETURN>
7. Configure a dump device as required.
SAN Boot environment by the UFS file system
# dumpadm -d /dev/FJSVmplb/dsk/mplb0s3 <RETURN>
SAN Boot environment by the ZFS file system
# dumpadm -d /dev/zvol/dsk/rootpool/dump <RETURN>
8. Stop the unit and reset the obp environment.
64
# /usr/sbin/shutdown -y -i0 -g0 <RETURN>
ok reset-all <RETURN>
9. Configure a boot device.
Configure a boot device on all redundant paths to the boot disk on the OBP. Take the physicaldevice name of each configuration path found in Step 3, excluding "../../devices" at thebeginning and ":*,raw" at the end, and set it with "mplbt" being replaced with "disk."
ok nvalias raid1 /pci@1,700000/fibre-channel@0/disk@10,0 <RETURN>
ok nvalias raid2 /pci@2,600000/fibre-channel@0/disk@10,0 <RETURN>
ok setenv boot-device raid1 raid2 <RETURN>
10. If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, execute thefollowing command;
ok setenv auto-boot? true <RETURN>
If you use SPARC Enterprise M3000/M4000/M5000/M8000/M9000, set the server modeswitchback to AUTO.
11. Start the host.
ok boot <RETURN>
4.2.2.2 Cluster system
Perform this procedure to build a cluster system using PRIMECLUSTER.
1. Specify a disk array device and boot the OS from it to install the ETERNUS Multipath Driver.
Launch the host from the boot disk on the disk array device. Then, install ETERNUSMultipath Driver as instructed in "ETERNUS Multipath Driver Install Guide." When the installcompletes, respond with "y" at the following prompt to let the grmpdautoconf command executeautomatically to proceed to the multipath building process in Step 2.
Do you want to make a multipath configuration now ?
If the ETERNUS Multipath Driver package has already been installed, execute grmpdautoconfto proceed to the multipath building process in Step 2.
# /usr/sbin/grmpdautoconf <RETURN>
2. Execute grmpdautoconf to build a multipath.
65
Work with grmpdautoconf interactively. For more details, refer to "ETERNUS MultipathDriver User's Guide." During this interactive session, make the following choices:
Select "Manual selection" m in response to the automatic/manual path selection prompt.
Select an access path automatically or manually?
** If automatic selection is selected, all access paths marked "New" are registered with thesystem.
** Select automatic selection if an access path has been properly selected at ETERNUS andswitch setup.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and device installation.
Manual selection ---> 'm'
Automatic selection ---> 'a'
Quit ---> 'q'
Enter a key. [m,a,q] m <RETURN>
Different manual path selection screens are displayed depending on which method of diskarray device identification has been selected in the Fibre Channel driver, automatic settingor manual setting.
a. [Automatic setting]
Select a startup path on the manual path selection screen.
Adapter Switch ETERNUSStatus
instance WWN WWN product
-----+----------------------------------+-----+--------------------------------------------------+-----
[ ] 1 fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0New
[ ] 2 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.
** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
66
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] 1 2 <RETURN>
Adapter Switch ETERNUS Status
instance WWN WWN product
-----+----------------------------------+-----+-------------------------------------------------+-----
[*] 1 fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0 New
[*] 2 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.
** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] x <RETURN>
The setting of the path (disk array device) selected above is reflected in the Fibre Channeldriver setting file (/kernel/drv/fjpfca.conf). Once this step completes, the setting of the diskarray device that is identified by the Fibre Channel driver is fixed, so the following setting in theFibre Channel driver setting file can be removed:
fcp-auto-bind-function=1;
b. [Manual setting]
67
The wwn entered in fjpfca.conf appears as "Exist" or "AL." Leave all other paths unselected,selecting "Confirmed (x)" for them.
Adapter Switch ETERNUSStatus
instance WWN WWN product
---+-----------------------------------+-----+------------------------------------------------------+-----
fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0Exist
[ ] 1 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.
** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] 1 <RETURN>
Adapter Switch ETERNUSStatus
instance WWN WWN product
-----+----------------------------------+-----+------------------------------------------------------+-----
fjpfca0 100000000e24ac06 1 210000e0004101d9 E4000 CM1CA0P0Exist
[*] 1 fjpfca1 100000000e244737 3 210000e0004101da E4000 CM0CA0P0New
Among the paths marked "New," enter the number of the access path you wantregistered with the system.
** The access path list is redisplayed after the entry.
68
** Selected paths are marked with "*."
** If an invalid number is selected, enter that number again to deselect it.
** All paths marked "Exist" are eligible for LUN installation.
** All devices residing on an AL connection are eligible for LUN and deviceinstallation.
Enter path numbers ---> Enter numerals (multiple path numbers can be enteredseparated from each
other with a comma)
Conclude entry ---> 'x'
Quit ---> 'q'
Enter a key. [Path number ,x,q] x <RETURN>
Select y in response to the prompt asking whether to use PRIMECLUSTER.
Is the disk array device used as a PRIMECLUSTER/SafeCLUSTER shared disk?
**If PRIMECLUSTER/SafeCLUSTER is used, use the multipath setup feature provided bythe relevant product.
** The value of maxthrottle also needs to be reviewed.
Yes ---> 'y' (Processing ends)
No ---> 'n'
Enter a key. [y,n] y <RETURN>
Processing up to sd has ended successfully.
*** IMPORTANT NOTICE ***
Installation of ETERNUS Multipart Driver Package was successful.
3. Next, execute the mplbconfig command.
# /usr/sbin/mplbconfig -o /tmp/mplb-file1 <RETURN>
*** Phase 1: Loading mplb.conf ***
*** Phase 2: Checking the device file in /dev ***
*** Phase 3: Checking the device fie in /devices ***
*** Phase 4: Checking mplb.conf against the configuration in /devices ***
=== Multipath configuration plan ===
69
Existing instance : 0
New instance : 2
Add a path : 0 (instance)
Delete a path : 0 (instance)
4. Delete all lines except for the system volume. Edit /tmp/mplb-file1 using a vi editor or the liketo delete all except for the path to a system disk.
With a cluster system, the instance number of each local multipathdisk, such as a boot disk,must not be defined in duplicate among the nodes that configure the cluster. Change instancenumbers (designated by X in mplbX) between 0 and 2047 to avoid duplication with other nodes.
*** mplb config file ***
Path : Process : Configuration path : LUN : Device information
mplb0 : new : c2t16d0s2 c13t16d0s2 : 0 : E40004641- 130011
5. Apply the edited file, converting the system volume into a multipath implementation.
# /usr/sbin/mplbconfig -f /tmp/mplb-file1 <RETURN>
*** Phase 1: Loading mplb.conf ***
*** Phase 2: Checking the device file in /dev ***
*** Phase 3: Checking the device file in /devices ***
*** Phase 4: Checking mplb.conf against the configuration in /devices ***
*** Phase 5: Updating mplb.conf ***
=== Multipath configuration plan ===
Existing instance: 0
New instance: 1
Add a path : 0 (instance)
Delete a path : 0 (instance)
6. Check the device path name of the boot device. The grmpdautoconf command carried out inStep 4 displays a combination of a multipath management special file and a selected accessspecial file. Use the output of the ls command to identify the boot disk and the physical devicepath name of each configuration path. Physical device path names thus found are used in Steps9 and 12.
# ls -l <Boot disk slice 0 > <RETURN>
# ls -l <Configuration path slice 2 > <RETURN>
70
Assume that grmpdautoconf has delivered the following output listing:
*** Phase 1: read mplb.conf ***
*** Phase 2: read /dev ***
*** Phase 3: read /devices ***
*** Phase 4: compare mplb.conf and /devices ***
Path : Action : Element path : LUN : Storage
mplb0 : new : c2t16d0s2 c13t16d0s2 : 0 : E40004641- 130011 :
mplb1 : new : c2t16d1s2 c13t16d1s2 : 1 : E40004641- 130011 :
mplb2 : new : c2t16d2s2 c13t16d2s2 : 2 : E40004641- 130011 :
The booth disk and the paths that configure it up are as follows:
Boot disk /dev/FJSVmplb/rdsk/mplb0s0
Configuration path /dev/rdsk/c2t16d0s2
/dev/rdsk/c13t16d0s2
Execute the ls command to check the device path names.
# ls -l /dev/FJSVmplb/rdsk/mplb0s0 <RETURN>
lrwxrwxrwx 1 root root 36 Aug 29 12:05 /dev/FJSVmplb/rdsk/mplb0s0 -> (line wrapping)
../../../devices/pseudo/mplb@0:a,raw <RETURN>
^^^^^^^^^^^^^^^
# ls -l /dev/rdsk/c2t16d0s2 <RETURN>
lrwxrwxrwx 1 root root 58 Aug 29 17:13 /dev/rdsk/c2t16d0s2 -> (line wrapping)
../../devices/pci@1,700000/fibre-channel@0/mplbt@10,0:c,raw <RETURN>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
# ls -l /dev/rdsk/c13t16d0s2 <RETURN>
lrwxrwxrwx 1 root root 58 Aug 29 17:13 /dev/rdsk/c13t16d0s2 -> (line wrapping)
../../devices/pci@2,600000/fibre-channel@0/mplbt@10,0:c,raw <RETURN>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
7. Set a Fibre Channel driver link speed.
71
Edit the Fibre Channel driver setting file (/kernel/drv/fjpfca.conf) to set a Fibre Channel driverlink speed.
In configuring the Fibre Channel driver, the link speed setting can be set to automatic selectionfor the sake of easier connectivity. The expected link speed may not be attained depending onthe connection status. Therefore, set the highest transmission rate available in the environment.
Example: Set fjpfca0 to a link speed of 4 Gbps.
port=
"fjpfca0:nport:sp4";
For more information about configuring fjpfca.conf, refer to the "FUJITSU PCI Fibre ChannelGuide."
8. Load the Fibre Channel cards that are used to access the boot disk with a disk array deviceboot setting.
Note
The setting to the Fibre Channel boot code used by boot is unnecessary now.
Example: Set on fjpfca1.
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca1 -b ENABLE <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca1 -c /kernel/drv/fjpfca.conf <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca1 -v <RETURN>
boot_function : ENABLE
topology : N_Port
link-speed : 4G
boot wait time : DISABLE ( interval time : DISABLE , boot wait msg : DISABLE )
bind-target: Target_ID=16,WWN=0x210000e0004101d9
9. Correct the system settings to meet the multipath implementation.
SAN Boot environment by the UFS file systema. Root device setting (/etc/system)
Edit /etc/system file to set rootdev and forceload. For the rootdev setting, set the boot diskphysical device name as found in Step 6 above, excluding "../../devices" at the beginning and",raw" at the end.
When the setting concerning forceload to each driver exists in /etc/system file, an additionalsetting need not be done.
rootdev: /pseudo/mplb@0:a
forceload: drv/mplbt
forceload: drv/mplb
72
forceload: drv/sd
b. Mount information setting (/etc/vfstab)
Edit /etc/vfstab file to rewrite each entry to a path name after the multipath implementation.
/dev/FJSVmplb/dsk/mplb0s0 /dev/FJSVmplb/rdsk/mplb0s0 / ufs 1 no -
/dev/FJSVmplb/dsk/mplb0s3 - - swap - no -
c. An access setup of a boot disk (/kernel/drv/sd.conf).
When target ID of the equipment used as a boot device has differed with two Fibre Channelcards which constitute a multipath, target ID about all the paths of a boot disk may not be writtento sd.conf. In that case, the following setup is needed.
○ Edit /kernel/drv/sd.conf
The definition of target ID of the path used as a boot disk is added.
Example: The definition of "Target ID = 18" is added.
name=”sd” class=”scsi” target=18 lun=0;
○ Reconstruction of sd driver
# touch /reconfigure <RETURN>or
# update_drv -f sd <RETURN>
SAN Boot environment by the ZFS file systema. Root device setting (/etc/system)
Edit /etc/system file to set forceload.
forceload: drv/mplb
forceload: drv/sd
b. An access setup of a boot disk (/kernel/drv/sd.conf).
When target ID of the equipment used as a boot device has differed with two Fibre Channelcards which constitute a multipath, target ID about all the paths of a boot disk may not be writtento sd.conf. In that case, the following setup is needed.
○ Edit /kernel/drv/sd.conf
The definition of target ID of the path used as a boot disk is added.
Example: The definition of "Target ID = 18" is added.
name=”sd” class=”scsi” target=18 lun=0;
○ Reconstruction of sd driver
# touch /reconfigure <RETURN>
73
or# update_drv -f sd <RETURN>
10. Configure a dump device as required.
SAN Boot environment by the UFS file system
# dumpadm -d /dev/FJSVmplb/dsk/mplb0s3 <RETURN>
SAN Boot environment by the ZFS file system
# dumpadm -d /dev/zvol/dsk/rootpool/dump <RETURN>
11. Stop the equipment and reset the obp environment.
# /usr/sbin/shutdown -y -i0 -g0 <RETURN>
ok reset-all <RETURN>
12. Configure a boot device.
Configure a boot device on all redundant paths to the boot device on the OBP. Take thephysical device name of each configuration path found in Step 6, excluding "../../devices" at thebeginning and ":*,raw" at the end, and set it with "mplbt" being replaced with "disk."
ok nvalias raid1 /pci@1,700000/fibre-channel@0/disk@10,0 <RETURN>
ok nvalias raid2 /pci@2,600000/fibre-channel@0/disk@10,0 <RETURN>
ok setenv boot-device raid1 raid2 <RETURN>
13. If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, execute thefollowing command;
ok setenv auto-boot? true <RETURN>
If you use SPARC Enterprise M3000/M4000/M5000/M8000/M9000, set the server modeswitchback to AUTO.
14. Start the host.
ok boot <RETURN>
74
4.3 Boot Disk MirroringThis section explains how to mirror between multipath implementations of two boot disks while the OS isbooted from one of them.
4.3.1 Mirroring by PRIMECLUSTER GDS
1. Verify that zoning has been implemented by the FC Switch as shown above.
2. Verify that the connection of the Fibre Channel card to the disk array device at the mirroringdestination has been configured properly in the OBP environment.
ok cd /pci@1,700000/fibre-channel@0 <RETURN>
ok PROBE fjpfca-info <RETURN>
Target -- DID 10500 210000e00040101d9 FUJITSU-E4000-0000
Target -- DID 10600 210000e00040101da FUJITSU-E4000-0000
3. Boot the OS and verify that the connection of the Fibre Channel driver to the disk array deviceat the mirroring destination has been configured properly.
# /usr/sbin/FJSVpfca/fc_info -p <RETURN>
adapter=fjpfca#0 :
port_id=0x010500 tid=0 wwn=210000e00040101d9 adapter=fjpfca#1 connected
class=class3
port_id=0x01060 tid=0 wwn=210000e00040101da adapter=fjpfca#1 connected
class=class3
75
4. If a Multipath Driver is yet to be set on the disk array device at the mirroring destination, set it.Referring to "ETERNUS Multipath Driver User's Guide," set a Multipath Driver and create amultipath disk on the disk array device (ETERNUS #2) at the mirroring destination.
5. If using the ETERNUS Multipath Driver, add a definition of the mirroring destination diskto target driver setting file /kernel/drv/sd.conf.
Example: Identify target ID 16, logical unit 1.
name=”sd” class=”scsi” target=16 lun=1;
6. Execute the install with reference to the PRIMECLUSTER GDS manual.
7. Refer to “PRIMECLUSTER Global Disk Services Guide” and mirror the disk at the mirroringsource and destination with each other.
If a system disk is mirrored using PRIMECLUSTER GDS, the message below may bedisplayed at boot time. This message may be ignored.
NOTICE: “forceload: drv/<driver name> appears more than once in /etc/system.
* One of mplb, mplbt and sd appears in place of <driver name>.
This message is displayed if the setting forceload is defined in duplicate in /etc/system file. Toleave this message hidden, delete later occurrences of the setting of forceload.
forceload: drv/mplb
~
forceload: drv/mplb Delete this line.
8. If a snapshot of a system volume is created using PRIMECLUSTER GDS Snapshot, performStep 1 through 5 above for the snapshot disk the same way as for the disk at the mirroringdestination. Then, configure the snapshot as instructed in “PRIMECLUSTER Global DiskService Guide.”
76
4.3.2 Notes on using PRIMECLUSTER
4.3.2.1 Cluster system building procedure
Build a cluster system in the procedural steps described below.
77
Chapter 5 Backing Up and Restoring BootDisks
Boot disks can be backed up in this environment described in this guide by following the procedural stepsexplained below.
Boot the OS from an internal disk and back up and restore it by file system or by lun.
Boot the OS from a network and back up and restore the boot disk by file system or by lun.
Back up and restore up the boot disk using ETERNUS EC (Equivalent Copy) or OPC (OnePoint Copy).* ETERNUS SF AdvancedCopy Manager is required.
In the environment explained in this guide, the method of booting the Solaris OS from CD/DVD andbacking up the boot disk cannot be used. This chapter concerns the procedures: "Boot the OS from aninternal disk and back up and restore it by file system or by lun" and "Boot the OS from a network andback up and restore the boot disk by file system or by lun." For information on the procedure "Back upand restore the boot disk using ETERNUS EC (Equivalent Copy) or OPC (One Point Copy) withETERNUS SF AdvancedCopy Manager," refer to the ETERNUS SF AdvancedCopy Manager manual.
For information on the procedures for backing up and restoring a system disk that is mirrored byPRIMECLUSTER GDS, refer to the "PRIMECLUSTER Global Disk Services Guide." Although themethod of booting the Solaris OS from CD/DVD and backing up and restoring a system disk is coveredhere, the same method can also be used when the OS has been booted from a network or internal disk. IfPRIMECLUSTER GDS Snapshot is used, the OS can be booted from a boot disk on a disk array deviceand then the boot disk can be backed up and restored using the ETERNUS Advanced Copy function orPRIMECLUSTER GDS copy function.
The type of tape device used may dictate certain precautions may apply in the procedure of configuringor implementing the backup and restore operations. Refer in advance to the instruction manual pertainingto the type of tape device used to ensure that boot disks are backed and restored as instructed.
In an environment in which a system disk has optional software installed on it that has a module runningas part of a kernel, such as a driver or file system, additional precautions may apply. Refer to the manualfor the optional software and follow its instructions.
Refer to "Solaris ZFS Administration Guide" for details of back up and restore of the ZFS file systemenvironment.
This chapter assumes that Solaris is installed on disk device c7t16d0 on a disk array device.
78
5.1 Backing Up/Restoring after Booting OSfrom a Network
If a boot disk residing on a disk array device has been created by installing the OS from a network, followthe procedures explained in this section to back up and restore the boot disk.
5.1.1 Backup procedure1. Boot the OS from a network in single-user mode with the -s option specified.
ok boot net -s <Return>
2. Back up the boot disk. The boot disk can be backed up by file system or by lun.
a. Back up by file system
UFS file system environment(1) The procedure for backing up a boot disk by file system using the ufsdump(1M)
command is explained below. Disk partition information, such as slice size, is notbacked up and needs to be recorded beforehand using the prtvtoc(1M) command orformat(1M) command.
# prtvtoc /dev/rdsk/c7t16d0s2 <Return>
or
# format /dev/rdsk/c7t16d0s2 <Return>
format> partition <Return>
partition> print <Return>
(2) Back up a boot disk using the ufsdump(1M) command. In this example,/dev/dsk/c7t16d0s0 is used as a boot disk, and tape device /dev/rmt/0 is used.
# ufsdump 0cf /dev/rmt/0 /dev/rdsk/c7t16d0s0 <Return>
ZFS file system environment(1) The procedure for backing up a boot disk by file system using the zfs(1M) command is
explained below. Disk partition information, such as slice size, is not backed up andneeds to be recorded beforehand using the prtvtoc(1M) command or format(1M)command.
# prtvtoc /dev/rdsk/c7t16d0s2 <Return>
or
# format /dev/rdsk/c7t16d0s2 <Return>
format> partition <Return>
79
partition> print <Return>
(2) Back up a boot disk using the zpool(1M) command.
It advances to the following procedure ignoring it though the error message might bedisplayed when the pool was imported.
# zpool import <Return>
pool: rpool
id: 4856116377389642800
state: ONLINE
action: The pool can be imported using its name or numeric identifier.
config:
rpool ONLINE
c7t16d0s0 ONLINE
# zpool import 4856116377389642800 <Return> Specify the ID that confirmed inzpool import.
# zfs list <Return>
NAME USED AVAIL REFER MOUNTPOINT
rpool 6.33G 13.2G 94K /rpool
rpool/ROOT 4.83G 13.2G 18K legacy
rpool/ROOT/s10_1008 4.83G 13.2G 4.76G /
rpool/dump 1.00G 13.2G 1.00G -
rpool/export 38K 13.2G 20K /export
rpool/export/home 18K 13.2G 18K /export/home
rpool/swap 512M 13.7G 10.0M -
#
(3) Creating of snapshot.
# zfs snapshot rpool/ROOT/s10_1008@snapshot <Return>
When failing in creating the snapshot, the following procedures are executed. And,the snapshot is created again.
# zfs set mountpoint=legacy rpool/ROOT/s10_1008 <Return>
# mount -F zfs rpool/ROOT/s10_1008 /mnt <Return>
# umount /mnt <Return>
(4) Back up a boot disk using the zfs(1M) command. In this example,rpool/ROOT/s10_1008 is used as a boot path, and tape device /dev/rmt/0 is used.
80
# zfs send rpool/ROOT/s10_1008@snapshot > /dev/rmt/0 <Retrun>
b. Backup by disk
(1) Back up a boot disk using the dd(1M) command. In this example, tape device/dev/rmt/0 is used.
# dd if=/dev/rdsk/c7t16d0s2 of=/dev/rmt/0 bs=64k <Return>
In this format, if= is followed by the name of the disk to be backed up, such as/dev/rdsk/c0t0d0s2, specified in the character type (/dev/rdsk/...). Remember tospecify s2, a slice that designates a disk as a whole.
The dd(1M) command might not be able to be backed up according to the size ofLUN because it doesn't correspond to the multi volume.
5.1.2 Restore procedure1. Boot the OS from a network in single-user mode with the -s option specified.
ok boot net -s <Return>
2. Restore the boot disk. Restore it in the same unit in which it has been backed up.
a. Restore by file system
UFS file system environment(1) If the boot disk has a new lun defined or has its lun derived from another use, create a
disk slice and a disk label using the format(1M) command. Reference the disk partitioninformation that has been recorded at backup for the size of the slice created and otherdetails.
# format <Return>
For information on creating a disk slice and a disk label using the format(1M)command, browse through the online manual.
(2) Create a new file system using the newfs(1M) command.
# newfs /dev/rdsk/c7t16d0s0 <Return>
Here, specify the slice name of the restore destination device in the character type(/dev/rdsk/...).
(3) Mount the boot disk. In this example, /dev/dsk/c7t16d0s0 is used as a boot disk.
# mount -F ufs /dev/dsk/c7t16d0s0 /mnt <Return>
81
(4) Move to the mounted directory.
# cd /mnt <Return>
(5) Restore the boot disk using the ufsrestore(1M) command. In this example, tape device/dev/rmt/0 is used. For example, the boot disk might be restored from another LU inthe disk array device.
# ufsrestore rf /dev/rmt/0 <Return>
(6) Create a boot block using the installboot(1M) command. For information on creating aboot block using installboot(1M) command, browse through the online manual.
Use the boot block on the restore destination device for the boot block creating.
Here, specify slice 0 of the restore destination device in the character type(/dev/rdsk/...). In this example, /dev/dsk/c7t16d0s0 is used as a boot disk.
# installboot /mnt/usr/platform/`uname -i`/lib/fs/ufs/bootblk /dev/rdsk/c7t16d0s0<Return>
(7) Move to the root directory and unmount the boot disk.
# cd / <Return>
# umount /mnt <Return>
(8) Check the file system for consistency using the fsck(1M) command.
# fsck /dev/rdsk/c7t16d0s0 <Return>
Here, specify the slice name the restore destination device in the character type(/dev/rdsk/...).
ZFS file system environment(1) If the boot disk has a new lun defined or has its lun derived from another use, create a
disk slice and a disk label using the format(1M) command. Reference the disk partitioninformation that has been recorded at backup for the size of the slice created and otherdetails.
# format <Return>
For information on creating a disk slice and a disk label using the format(1M)command, browse through the online manual.
(2) Create the ZFS file system.
82
# zpool create rpool c7t16d0s0 <Return>
# zfs create rpool/ROOT <Return>
(3) Restore the boot disk using the zfs(1M) command.
# zfs receive rpool/ROOT/s10_1008@snapshot < /dev/rmt/0 <Return>
(4) The mountpoint property set to the legacy.
# zfs set mountpoint=legacy rpool/ROOT/s10_1008 <Return>
(5) Mount the restore destination device.
# mount -F zfs rpool/ROOT/s10_1008 /mnt <Return>
(6) Create a boot block using the installboot(1M) command. For information on creating aboot block using installboot(1M) command, browse through the online manual.
Use the boot block on the restore destination device for the boot block creating.
Here, specify slice 0 of the restore destination device in the character type(/dev/rdsk/...). In this example, /dev/dsk/c7t16d0s0 is used as a boot disk.
# installboot -F zfs /mnt/usr/platform/`uname -i`/lib/fs/zfs/bootblk/dev/rdsk/c7t16d0s0 <Return>
(7) The mountpoint property set to the root path.It advances to the following procedure ignoring it though the error message isdisplayed when the mountpoint property is set the root path.
# zfs set mountpoint=/ rpool/ROOT/s10_1008 <Return>
(8) Bootfs setting.
# zpool set bootfs=rpool/ROOT/s10_1008 rpool <Return>
b. Restore by disk
(1) Restore the boot disk using the dd(1M) command. In this example, tape device/dev/rmt/0 is used.
# dd if=/dev/rmt/0 of=/dev/rdsk/c7t16d0s2 bs=64k <Return>
Specify the name of the disk to be backed up in the character type (/dev/rdsk/...).Remember to specify s2, a slice that designates a disk as a whole.
83
5.2 Backing Up/Restoring after Booting the OSfrom an Internal Disk
If a boot disk residing on a disk array device has been created by copying the OS from an internal disk,follow the procedures explained in this section to back up and restore the boot disk.
5.2.1 Backup procedure1. Boot the OS from an internal disk in single-user mode with the -s option specified.
ok boot <internal disk> -s <Return>
2. Back up the boot disk. The boot disk can be backed up by file system or by lun.
a. Back up by file system
UFS file system environment(1) The procedure for backing up a boot disk by file system using the ufsdump(1M)
command is explained below. Disk partition information, such as slice size, is notbacked up and needs to be recorded beforehand using the prtvtoc(1M) command orformat(1M) command.
# prtvtoc /dev/rdsk/c7t16d0s2 <Return>
or
# format /dev/rdsk/c7t16d0s2 <Return>
format> partition <Return>
partition> print <Return>
(2) Back up the boot disk using the ufsdump(1M) command. In this example,/dev/dsk/c7t16d0s0 is used as a boot disk, and tape device /dev/rmt/0 is used.
# ufsdump 0ucf /dev/rmt/0 /dev/rdsk/c7t16d0s0 <Return>
ZFS file system environment(1) The procedure for backing up a boot disk by file system using the zfs(1M) command
is explained below. Disk partition information, such as slice size, is not backed upand needs to be recorded beforehand using the prtvtoc(1M) command or format(1M)command.
# prtvtoc /dev/rdsk/c7t16d0s2 <Return>
or
# format /dev/rdsk/c7t16d0s2 <Return>
84
format> partition <Return>
partition> print <Return>
(2) Back up a boot disk using the zpool(1M) command.
It advances to the following procedure ignoring it though the error message mightbe displayed when the pool was imported.
# zpool import <Return>
pool: rpool
id: 4856116377389642800
state: ONLINE
action: The pool can be imported using its name or numeric identifier.
config:
rpool ONLINE
c7t16d0s0 ONLINE
# zpool import 4856116377389642800 <Return> ID confirmed with zpool importis specified.
# zfs list <Return>
NAME USED AVAIL REFER MOUNTPOINT
rpool 6.33G 13.2G 94K /rpool
rpool/ROOT 4.83G 13.2G 18K legacy
rpool/ROOT/s10_1008 4.83G 13.2G 4.76G /
rpool/dump 1.00G 13.2G 1.00G -
rpool/export 38K 13.2G 20K /export
rpool/export/home 18K 13.2G 18K /export/home
rpool/swap 512M 13.7G 10.0M -
#
(3) Creating of snapshot.
# zfs snapshot rpool/ROOT/s10_1008@snapshot <Return>
When failing in creating the snapshot, the following procedures are executed. And,the snapshot is created again.
# zfs set mountpoint=legacy rpool/ROOT/s10_1008 <Return>
# mount -F zfs rpool/ROOT/s10_1008 /mnt <Return>
# umount /mnt <Return>
85
(4) Back up a boot disk using the zfs(1M) command. In this example,rpool/ROOT/s10_1008 is used as a boot path, and tape device /dev/rmt/0 is used.
# zfs send rpool/ROOT/s10_1008@snapshot > /dev/rmt/0 <Retrun>
b. Back up by disk
(1) Back up the boot disk using the dd(1M) command. In this example, tape device/dev/rmt/0 is used. The boot disk may also be backed up to another LU on the diskarray device.
# dd if=/dev/rdsk/c7t16d0s2 of=/dev/rmt/0 bs=64k <Return>
In this format, if= is followed by the name of the disk to be backed up, such as/dev/rdsk/c0t0d0s2, specified in the character type (/dev/rdsk/...). Remember tospecify s2, a slice that designates a disk as a whole.
The dd(1M) command might not be able to be backed up according to the size ofLUN because it doesn't correspond to the multi volume.
5.2.2 Restore procedure1. Boot the OS from an internal disk in single-user mode with the -s option specified.
ok boot <internal disk> -s <Return>
2. Restore the boot disk in the same unit in which it has been backed up.
a. Restore by file system
UFS file system environment(1) If the boot disk has a new lun defined or has its lun derived from another use, create a
disk slice and a disk label using the format(1M) command. Reference the disk partitioninformation that has been recorded at backup for the size of the slice created and otherdetails.
# format <Return>
For information on creating a disk slice and a disk label using the format(1M)command, browse through the online manual.
(2) Create a new file system using the newfs(1M) command.
# newfs /dev/rdsk/c7t16d0s0 <Return>
Specify the name of the slice in which the boot disk is restored in the character type(/dev/rdsk/...).
(3) Mount the boot disk.
86
In this example, /dev/dsk/c7t16d0s0 is used as a boot disk.
# mount -F ufs /dev/dsk/c7t16d0s0 /mnt <Return>
(4) Move to the mounted directory.
# cd /mnt <Return>
(5) Restore the boot disk using the ufsrestore(1M) command. In this example, tape device/dev/rmt/0 is used.
# ufsrestore rf /dev/rmt/0 <Return>
(6) Create a boot block using the installboot(1M) command. For information on creating aboot block using installboot(1M) command, browse through the online manual.
Use the boot block on the restore destination device for the boot block creating.
Here, specify slice 0 of the restore destination device in the character type(/dev/rdsk/...). In this example, /dev/dsk/c7t16d0s0 is used as a boot disk.
# installboot /mnt/usr/platform/`uname -i`/lib/fs/ufs/bootblk /dev/rdsk/c7t16d0s0<Return>
(7) Move to the root directory and unmount the boot disk.
# cd / <Return>
# umount /mnt <Return>
(8) Check the file system for consistency using the fsck(1M) command.
# fsck /dev/rdsk/c7t16d0s0 <Return>
Specify the name of the slice in which the boot disk is restored in the character type(/dev/rdsk/...).
ZFS file system environment(1) If the boot disk has a new lun defined or has its lun derived from another use, create a
disk slice and a disk label using the format(1M) command. Reference the diskpartition information that has been recorded at backup for the size of the slice createdand other details.
# format <Return>
87
For information on creating a disk slice and a disk label using the format(1M)command, browse through the online manual.
(2) Create the ZFS file system.
# zpool create rpool c7t16d0s0 <Return>
# zfs create rpool/ROOT <Return>
(3) Restore the boot disk using the zfs(1M) command.
# zfs receive rpool/ROOT/s10_1008@snapshot < /dev/rmt/0 <Return>
(4) The mountpoint property set to the legacy.
# zfs set mountpoint=legacy rpool/ROOT/s10_1008 <Return>
(5) Mount the restore destination device.
# mount -F zfs rpool/ROOT/s10_1008 /mnt <Return>
(6) Create a boot block using the installboot(1M) command. For information on creating aboot block using installboot(1M) command, browse through the online manual.
Use the boot block on the restore destination device for the boot block creating.
Here, specify slice 0 of the restore destination device in the character type(/dev/rdsk/...). In this example, /dev/dsk/c7t16d0s0 is used as a boot disk.
# installboot -F zfs /mnt/usr/platform/`uname -i`/lib/fs/zfs/bootblk/dev/rdsk/c7t16d0s0 <Return>
(7) The mountpoint property set to the root path.It advances to the following procedure ignoring it though the error message isdisplayed when the mountpoint property is set the root path.
# zfs set mountpoint=/ rpool/ROOT/s10_1008 <Return>
(8) Bootfs setting.
# zpool set bootfs=rpool/ROOT/s10_1008 rpool <Return>
b. Restore by disk
(1) Restore the boot disk using the dd(1M) command. In this example, tape device/dev/rmt/0 is used. The boot disk may also be restored from another LU on the diskarray unit that has been backed up beforehand.
88
# dd if=/dev/rmt/0 of=/dev/rdsk/c7t16d0s2 bs=64k <Return>
Here, specify the name of the disk to be backed up in the character type(/dev/rdsk/...). Remember to specify s2, a slice that designates a disk as a whole.
89
Appendix A Boot Device Setup Commands
This appendix focuses on the commands used to configure the boot code on the Fibre Channel card.These commands are executable on either the OS or the OBP.
The commands introduced here work only on the single-channel 4Gbps Fibre Channel card (SE0X7F11x)and dual-channel 4Gbps Fibre Channel card (SE0X7F12x).
A.1 Command Executable on the OSPerform this procedure with the OS being booted and with FUJITSU PCI Fibre Channel 4.0 or any laterpackage installed.
1 fc_hbaprp
Name
fc_hbaprp
Format
/usr/sbin/FJSVpfca/fc_hbaprp -i adpname -f tgt_id -P WWN
-f tgt_id -I PORT_ID
-d tgt_id
-D [-y]
-w boot-wait-time
-l linkspeed
-t topology
-v
-s savefile
-r|-R filename
-c conffile
-C [-y]
-b ENABLE|DISABLE
90
Function
Configures the boot code on the Fibre Channel card.
Operands
The settings that work on the boot code on the Fibre Channel card are listed below.
All these settings need to be accompanied by the specification of -i adpname. Specify theinstance name of the Fibre Channel driver as adpname.
-i adpname -f tgt_id -P WWN
-i adpname -f tgt_id -I PORT_ID
Configure a target device. Up to 10 entries can be registered.
The following values can be set:
tgt_id Specifies Target_ID of the target device with a decimal number.
WWN Specifies WWPN of the target device with a hexadecimal number (boot devicespecification by WWPN).
PORT_ID Specifies Port_ID(DID) of the target device with a hexadecimal number (boot devicespecification by Port_ID).
-i adpname -d tgt_id
Erases the setting of the target device. The following value can be set:
tgt_id Specifies Target_ID of the target device with a decimal number.
-i adpname -D [-y]
Erases all settings of the target device.
If -y is not attached, a message asking if you are sure you want to erase the settings is displayed.
If -y is attached, the settings are erased unconditionally.
-i adpname -w boot-wait-time
Sets a boot wait time in seconds. The following value can be set:
boot-wait-time Specifies a boot wait time with a decimal number. Either 0 second (no boot-wait-time) or a value between 180 and 86,400 seconds can be set.
-i adpname -l linkspeed
Sets a link speed. One of the following values can be set:
1G|1g : Sets 1 Gbps.
2G|2g : Sets 2 Gbps.
91
4G|4g : Sets 4 Gbps.
AUTO|auto : Sets a link speed with AUTO.
-i adpname -t topology
Sets a topology. One of the following values can be set:
NPORT|nport : Makes an NPORT connection.
Connects to the Fibre Channel switch.
AL|al : Sets the FC-AL topology.
AUTO|auto : Sets a topology automatically.
-i adpname -v
Displays the settings as listed below.
Item Value Description
boot function DISABLE/ENABLE Enables or disables the boot function.
Target_ID Example) 0 (hexadecimal) Bound Target_ID
Target WWN Example) 210000e0004101d9(hexadecimal)
Bound target
Target DID Example) 010111 (hexadecimal) Bound DID
topology AL/N_Port/AUTO Set topology. AUTO refers to an automaticallyset topology.
link-speed 1G/2G/4G/AUTO Set link speed. AUTO refers to an automaticallyset link speed.
boot wait time DISABLE or numeric value(decimal)
Set boot wait time. Value is specified in seconds.DISABLE overrides a boot wait time.
interval time DISABLE This item is not available and cannot be changed.
boot wait msg DISABLE This item is not available and cannot be changed.
92
-i adpname -s savefile
Saves the settings as listed below.
Item Value Description
boot function DISABLE/ENABLE Enables or disables the boot function.
Target_ID Example) 0 (hexadecimal) Bound Target_ID
Target WWN Example) 210000e0004101d9(hexadecimal)
Bound target
Target DID Example) 010111 (hexadecimal) Bound DID
topology AL/N_Port/AUTO Set topology. AUTO refers to an automaticallyset topology.
link-speed 1G/2G/4G/AUTO Set link speed. AUTO refers to an automaticallyset link speed.
boot wait time DISABLE or numeric value(decimal)
Set boot wait time. Value is specified in seconds.DISABLE overrides a boot wait time.
interval time DISABLE This item is not available and cannot be changed.
boot wait msg DISABLE This item is not available and cannot be changed.
-i adpname -r|-R filename
Updates the boot code on the Fibre Channel card with the settings saved with -s.
If -r is specified, the boot code on the Fibre Channel card is updated with all settings, except forthe boot function.
If -R is specified, the boot code on the Fibre Chanel card is updated with all settings, includingthe boot function.
-i adpname -c conffile
Updates the boot code on the Fibre Channel card with the settings of the driver setting file(/kernel/drv/fjpfca.conf).
-i adpname -C [-y]
Erases all settings, except for enabling/disabling of the boot function.
If -y is not attached, a message asking if you are sure you want to erase the settings is displayed.
If -y is attached, the settings are erased unconditionally.
93
-i adpname -b ENABLE|DISABLE
Enables or disables the Fibre Channel card boot function. One of the following values can beset:
ENABLE : Enables the boot function.
DISABLE : Disables the boot function.
Note
Enabling or disabling the boot function on either port of the dual-channel 4Gbps Fibre Channelcard (SE0X7F12x) will automatically impart the same setting to the other port. It is not possibleto make a change to the setting of either port alone and not both.
Example
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -f 0 -P 0x210000e0001014d9 <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -f 1 -I 0x10c00 <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -d 0 <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -D <RETURN>
delete all bind registration ? [y(Y),n(N) ] y <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -w 180 <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -l 4g <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -t nport <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -v <RETURN>
boot function : ENABLE
topology : N_Port
link-speed : 4G
boot wait time : 180 ( interval time : DISABLE , boot wait msg : DISABLE )
bind-target: Target_ID=0,WWPN=0x210000e0001014d9
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -s savefile <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -r savefile <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -c /kernel/drv/fjpfca.conf <RETURN>
# /usr/sbin/FJSVpfca/fc_hbaprp -i fjpfca0 -b ENABLE <RETURN>
94
A.2 Command Executable on the OBPBefore starting the procedure, set the to maintenance mode and restart the server. If you use SPARCEnterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440, execute the following command:
ok setenv auto-boot? false <RETURN>
ok reset-all <RETURN>
If you use SPARC Enterprise M3000/M4000/M5000/M8000/M9000, set the server mode switch toservice mode and execute the following command:
ok reset-all <RETURN>
If an attempt is made to configure the Fibre Channel card in any other condition, it could hang up. Whenit does, turn the power to the server off, and then turn it back on.
To carry out this command, you need to move to the node of the Fibre Channel card on which toconfigure.
Example: single-channel 4Gbps Fibre Channel card (SE0X7F11x) and dual-channel 4Gbps FibreChannel card (SE0X7F12x) mounted on a server
ok show-devs <RETURN>
/pci@1,700000
/pci@2,600000
**
/openprom
/chosen
/packages
/pci@1,700000/fibre-channel@0 *physical path name of the single channel 4 Gbps FibreChannel card
/pci@2,600000/fibre-channel@0 *physical path name of the dual-channel 4Gbps FibreChannel card port0
/pci@2,600000/fibre-channel@0,1 physical path name of the dual-channel 4Gbps FibreChannel card port1
ok cd /pci@1,700000/fibre-channel@0 <RETURN>
ok
1 fjpfca-set-bootfunction
Name
fjpfca-set-bootfunction
95
Format
ENABLE | DISABLE fjpfca-set-bootfunction
Function
Enables or disables the Fibre Channel card boot function.
Note)The server must be restarted or the reset-all command must be carried out after thiscommand has been carried out. In configuring on more than one card as well, remember to carryout reset-all for every card mounted.
Operands
One of the following values can be set:
ENABLE : Enable the boot function.
DISABLE : Disables the boot function.
Note
Enabling or disabling the boot function on either port of the dual-channel 4Gbps Fibre Channelcard (SE0X7F12x) will automatically impart the same setting to the other port. It is not possibleto make a change to the setting of either port alone and not both.
Example
ok ENABLE fjpfca-set-bootfunction <RETURN>
ok reset-all <RETURN>
..
ok DISABLE fjpfca-set-bootfunction <RETURN>
ok reset-all <RETURN>..
2 fjpfca-output-prop
Name
fjpfca-output-prop
Format
fjpfca-output-prop
96
Function
Displays the settings stored in ROM on the Fibre Channel card.
Item Value Description
boot function DISABLE/ENABLE Enables or disables the boot function.
Target_ID Example) 0 (hexadecimal) Bound Target_ID
Target WWN Example) 210000e0004101d9(hexadecimal)
Bound target
Target DID Example) 010111 (hexadecimal) Bound DID
topology AL/N_Port/AUTO Set topology. AUTO refers to anautomatically set topology.
link-speed 1G/2G/4G/AUTO Set link speed. AUTO refers to anautomatically set link speed.
boot wait time DISABLE or numeric value(decimal)
Set boot wait time. Value is specifiedin seconds.DISABLE overrides a boot wait time.
interval time DISABLE This item is not available and cannotbe changed.
boot wait msg DISABLE This item is not available and cannotbe changed.
Example
ok fjpfca-output-prop <RETURN>
boot function : ENABLE
topology : AUTO
link-speed : AUTO
boot wait time : DISABLE ( interval time : DISABLE , boot wait msg : DISABLE )
bind-target: Target_ID=0,WWPN=0x210000e0001014d9
97
3 fjpfca-set-linkspeed
Name
fjpfca-set-linkspeed
Format
1g | 2g | 4g | auto fjpfca-set-linkspeed
Function
Sets a link speed.
Operands
One of the following values can be set:
1g : Sets 1 Gbps.
2g : Sets 2 Gbps.
4 g : Sets 4 Gbps.
auto: Sets a link speed automatically.
Example
ok 1g fjpfca-set-linkspeed <RETURN>
ok 2g fjpfca-set-linkspeed <RETURN>
ok 4g fjpfca-set-linkspeed <RETURN>
ok auto fjpfca-set-linkspeed <RETURN>
Default value
auto
Note
This command is operable only if the boot function is enabled.
4 fjpfca-set-topology
Name
fjpfca-set-topology
Format
nport | al | auto fjpfca-set-topology
Function
98
Sets a topology.
Operands
nport: Makes an NPORT connection.Connects to the Fibre Channel switch.
al: Sets the FC-AL topology.
auto: Sets a topology automatically.
Example
ok nport fjpfca-set-topology <RETURN>
ok al fjpfca-set-topology <RETURN>
ok auto fjpfca-set-topology <RETURN>
Default value
auto
Note
This command is operable only if the boot function is enabled.
5 fjpfca-bind-target
Name
fjpfca-bind-target
Format
value1 target-alpa | target-did |target-wwpn value2 fjpfca-bind-target
Function
Configure a target device. Up to 10 entries can be registered.
Operands
Specify a target device to be connected with either Port_ID (DID) or WWPN.
value1: Specifies Target_ID of the target device with a hexadecimal number.
target-wwpn: Specifies WWPN of the target device (boot device specification by WWPN).
target-alpa: Specifies Port_ID (DID) of the target device (boot device specification byPort_ID).
target-did: Specifies Port_ID (DID) of the target device (boot device specification byPort_ID).
99
value2: Specifies Port_ID (DID or WWPN with a hexadecimal number.
Example
ok 0 target-wwpn 210000e0004101d9 fjpfca-bind-target <RETURN> * WWN specification
ok 1 target-alpa 11206 fjpfca-bind-target <RETURN> * Port_ID(DID) specification
ok 2 target-did 11000 fjpfca-bind-target <RETURN> * Port_ID(DID) specification
Note
This command is operable only if the boot function is enabled.
6 led-flash
Name
led-flash
Format
[sec-time] led-flash
Function
Flashes the LED on the Fibre Channel card (for 10 seconds by default and for 60 seconds at thelongest).
Use this command to verify the location of the Fibre Channel card and where the ports arepositioned (on a dual-channel 4Gbps Fibre Channel card).
Operands
Specify the duration of time in which the LED flashes.
A number preceded by d# is assumed to be a decimal number. All other numbers are assumedto be hexadecimal.
Example
ok led-flash Flashes the LED for 10 seconds.
ok d# 10 led-flash Flashes the LED for 10 seconds (period specified with a decimal number).
ok 3c led-flash Flashes the LED for 60 seconds (period specified with a hexadecimal number)
Note
This command is operable only if the boot function is enabled.
100
7 fjpfca-set-boot-wait-time
Name
fjpfca-set-boot-wait-time
Format
wait-time | DISABLE fjpfca-set-boot-wait-time
Function
If a power supply interlock control is implemented between the server and a disk array device,it is necessary to let the disk array device start up before launching a boot sequence. Use thefjpfca-set-boot-wait-time command to delay the launch of the boot sequence for a specifiedperiod of time.
The Fibre Channel card is monitoring the status of the disk array device even while the OSwaits to be booted, ready to start booting automatically as soon as it confirms that the disk arraydevice is seen started up, even before the specified period of time expires.
The period of time for which the boot sequence is delayed can be set in seconds between 180and 86,400 seconds.
The command comes with this mode disabled by default.
Referring to the manual supplied with the disk array device, set (for the boot wait time) theamount of time that it takes for the system to enter the READY state after the POWER switch ispressed.
Operands
Specify a wait time (in seconds) at boot time with a hexadecimal number.
A number preceded by d# is assumed a decimal number.
DISABLE overrides a boot wait time.
Example
ok d# 1200 fjpfca-set-boot-wait-time Sets 1200 seconds
ok b4 fjpfca-set-boot-wait-time Sets 180 seconds
ok DISABLE fjpfca-set-boot-wait-time Overrides a boot wait time
Default value
DISABLE
Note
This command is operable only if the boot function is enabled.
101
8 fjpfca-info
Name
fjpfca-info
Format
STATUS | PROBE fjpfca-info
Function
Displays connectivity information about the Fibre Channel card mounted.
Operands
One of the following values can be set:
STATUS Displays the Link status of the Fibre Channel card.Indicates whether the target device set up on the Fibre Channel card is connectable.
PROBE Displays a list of target devices connectable from the Fibre Channel card.
Example
ok STATUS fjpfca-info <RETURN>
Link_status=up topology=Nport port_id=0x010000 wwpn=1000000b5d65c00a(0)
port_id=0x010100 tid=0 wwpn=210000e00004101d9 connected(0)
ok PROBE fjpfca-info <RETURN>
Target -- DID 10100 WWPN 210000e0004101d9 FUJITSU-E4000-0000
Note
This command is operable only if the boot function is enabled.
9 fjpfca-target-cancel
Name
fjpfca-target-cancel
Format
tgt_id fjpfca-target-cancel
Function
Erases the settings of a target device.
102
Operands
tgt_id Specifies Target_ID of the target device with a hexadecimal number.
Example
ok 0 fjpfca-target-cancel <RETURN>
Note
This command is operable only if the boot function is enabled.
10 fjpfca-all-target-cancel
Name
fjpfca-all-target-cancel
Format
fjpfca-all-target-cancel
Function
Erases all settings of the target device.
Example
ok fjpfca-all-target-cancel <RETURN>
delete all bind registration ? [ y(Y),n(N) ] y
Note
This command is operable only if the boot function is enabled.
103
Appendix B Checking the Fibre ChannelCard Boot Code VersionNumber
This appendix explains how to confirm the boot code (firmware) version number of the Fibre Channelcard.
There are two ways to perform this confirmation: confirming the code by checking the OS and confirmingthe code by checking OBP.
B.1 Checking on the OSPerform this procedure with the OS already booted and with FUJITSU PCI Fibre Channel 4.0 or any laterpackage installed.
View /var/adm/messages to check the boot code version number from the following display:
scsi: [ID 243001 kern.info] /pci@1,700000/fibre-channel@0 (fjpfca0):
INFO : FUJITSU PCI Fibre Channel FCode Version : v12l30, boot_function=ENABLE;
B.2 Checking on the OBPPerform this procedure on the OBP. Before starting the procedure, set the to maintenance mode andrestart the server. If you use SPARC Enterprise T1000/T2000/T5120/T5140/T5220/T5240/T5440,execute the following command:
ok setenv auto-boot? false <RETURN>
ok reset-all <RETURN>
If you use SPARC Enterprise M3000/M4000/M5000/M8000/M9000, set the server mode switch toservice mode and execute the following command:
ok reset-all <RETURN>
Move to the node of the Fibre Channel card whose boot code version number is to be confirmed, andexecute the .properties command.
104
Read the value of fjpfca_fcode_vl.
ok cd /pci@1,700000/fibre-channel@0 <RETURN>
ok .properties <RETURN>
status okay
fru PCI slot(PCI#08)
component-name PCI#08
assigned-addresses 81001814 00000000 00000700 00000000 00000100
(Omission)
fjpfca_fcode_vl v12l30
(Omission)
ok
105
Appendix C Recording SAN Boot SettingInformation
Record and save to any other document the values listed in the "SAN Boot Setting Information" table thathave been set at installation. Whenever a Fibre Channel card in use is replaced, reconfigure the same setof values of this "SAN Boot Setting Information" on the replacement card as well.
The reconfiguration sequence takes different courses depending on which of the following cases applies:
1. Active replacement of the card during OS running on an alternate path.
2. Cold replacement of the card when OS is available on alternate path.
3. Cold replacement of the card when OS can not boot on any paths.
When a card is replaced in Cases 1 and 2 above, the SAN boot setting information can be reconfigureinto the new card from the Fibre Channel driver environment definition files using the fc_hbaprpcommand. This "SAN Boot Setting Information" is required when replacing a card in Case 3. In thiscase, carry out a command executable on the OBP to reconfigure the SAN boot setting information.
Keep a record of this setting information with regard to all Fibre Channel cards used for rebooting the OS.
106
"SAN Boot Setting Information"
No. Item Explanation Actual value (record)
1 Device path name Name of a physical device path on theOS (/pci@XXXX/yyyy@z)
2 Slot position Mounted slot position
3 Boot function (boot function) Specify whether to enable or disable theboot function.
4 Topology information(topology)
Specify a topology (nport to connect tothe switch, al to make a directconnection)
5 Link speed (link speed) Use to fix a speed of transmission line(1G/2G/4G/auto).
6 Boot delay function (boot waittime)
Specify whether to enable or disable theboot delay function and a boot waittime.
target_id :1
wwn | did :
target_id :2
wwn | did :
target_id :3
wwn | did :
target_id :4
wwn | did :
target_id :5
wwn | did :
target_id :6
wwn | did :
target_id :7
wwn | did :
target_id :8
wwn | did :
target_id :9
wwn | did :
target_id :
7 Target bind information(Target_ID/TargetWWN|Target DID)
Specify disk array device bindinformation (for all registered entries).
10wwn | did :
107
Appendix D Making Fixes to Setting Filesafter a Boot Failure
Errors in any SAN Boot setting file (such as sd.conf, mplb.conf or /etc/system) could prevent the OSfrom starting up after a boot failure. When this occurs, start the OS in single-user mode and mount thesystem disk on a disk array device to make fixes to the setting file as instructed in this appendix.
The way the OS is started depends on which of the following ways has been used to install the OS on thesystem disk on a disk array device.
System disk installation as per Section 4.1.1, "Creating a boot disk using a network install server"
System disk installation as per Section 4.1.2, "Creating a boot disk by copying an existing boot diskresiding on an internal disk"
D.1 If the OS Has Been Installed As Per Section4.1.1, "Creating a boot disk using a networkinstall server"
1. Initialize the obp environment.
ok reset-all <RETURN>
2. Boot from the network. Start the OS in single-user mode.
ok boot net -s <RETURN>
3. Mount the system disk on a disk array device.
For the disk to be mounted, specify the device that has been set as an OS installation device inSection "4.1.1.4, Configuring Custom JumpStart"
Environment installed by UFS file system
# mount -F ufs /dev/dsk/c7t16d0s0 /mnt <RETURN>
Environment installed by ZFS file systemIt advances to the following procedure ignoring it though the error message might be displayedwhen importing pool.
# zpool import <RETURN>
pool: raid_pool
id: 9153334525621735888
state: ONLINE
108
action: The pool can be imported using its name or numeric identifier.
config:
raid_pool ONLINE
c7t16d0s0 ONLINE
# zpool import 9153334525621735888 <RETURN> Specify the ID that confirmed in zpoolimport.
# zfs list <RETURN>
NAME USED AVAIL REFER MOUNTPOINT
raid_pool 5.98G 92.5G 93K /raid_pool
raid_pool/ROOT 4.98G 92.5G 18K legacy
raid_pool/ROOT/s10_1008 4.98G 92.5G 4.98G /
raid_pool/dump 512M 92.5G 512M -
raid_pool/export 38K 92.5G 20K /export
raid_pool/export/home 18K 92.5G 18K /export/home
raid_pool/swap 512M 92.9G 88.0M -
# zfs set mountpoint=legacy raid_pool/ROOT/s10_1008 <RETURN>
# mount -F zfs raid_pool/ROOT/s10_1008 /mnt <RETURN>
4. Fix the setting files in the /mnt directory.
Example)If /etc/vfstab is the cause of the failure and requires change, fix /mnt/etc/vfstab.
5. When the fix is completed, unmount the system disk and migrate to the obp environment asinstructed below.
UFS file system
# cd / <RETURN>
# umount /mnt <RETURN>
ZFS file systemIt advances to the following procedure ignoring it though the error message is displayed whenthe mountpoint property is set the root path.
# cd / <RETURN>
# umount /mnt <RETURN>
# zfs set mountpoint=/ raid_pool/ROOT/s10_1008 <RETURN>
6. Perform a retry from the boot sequence afterwards.
109
D.2 If the OS Has Been Installed As Per Section4.1.2, "Creating a boot disk by copying anexisting boot disk residing on an internaldisk"
1. Initialize the obp environment.
ok reset-all <RETURN>
2. Boot from the internal disk. Start the OS in single-user mode.
ok boot disk0 -s <RETURN>
^^^^^^
^^^^^^ denotes an internal disk specification.
3. Mount the system disk on a disk array device.
For the disk to be mounted, specify the disk that has been verified in Section "4.1.2.1 Gettingready to copy the boot disk to a disk array device."
Environment installed by UFS file system
# mount -F ufs /dev/dsk/c7t16d0s0 /mnt <RETURN>
Environment installed by ZFS file systemIt advances to the following procedure ignoring it though the error message might be displayedwhen importing pool.
# zpool import <RETURN>
pool: raid_pool
id: 9153334525621735888
state: ONLINE
action: The pool can be imported using its name or numeric identifier.
config:
raid_pool ONLINE
c7t16d0s0 ONLINE
# zpool import 9153334525621735888 <RETURN> Specify the ID that confirmed in zpoolimport.
# zfs list <RETURN>
NAME USED AVAIL REFER MOUNTPOINT
raid_pool 5.98G 92.5G 93K /raid_pool
raid_pool/ROOT 4.98G 92.5G 18K legacy
110
raid_pool/ROOT/s10_1008 4.98G 92.5G 4.98G /
raid_pool/dump 512M 92.5G 512M -
raid_pool/export 38K 92.5G 20K /export
raid_pool/export/home 18K 92.5G 18K /export/home
raid_pool/swap 512M 92.9G 88.0M -
# zfs set mountpoint=legacy raid_pool/ROOT/s10_1008 <RETURN>
# mount -F zfs raid_pool/ROOT/s10_1008 /mnt <RETURN>
4. Fix the setting files in /mnt directory.
Example)If /etc/vfstab is the cause of the failure and requires change, fix /mnt/etc/vfstab.
5. When the fix is completed, unmount the system disk and migrate to the obp environment asinstructed below.
UFS file system
# cd / <RETURN>
# umount /mnt <RETURN>
ZFS file systemIt advances to the following procedure ignoring it though the error message is displayed whenthe mountpoint property is set the root path.
# cd / <RETURN>
# umount /mnt <RETURN>
# zfs set mountpoint=/ raid_pool/ROOT/s10_1008 <RETURN>
6. Perform a retry from the boot sequence afterwards.
111
Appendix E Fibre Channel Driver/BootCode Auto-Target BindingFunctions
This appendix describes the Fibre Channel driver/boot code auto-target binding functions.
E.1 Fibre Channel Driver Auto-Target BindingFunction
The Fibre Channel driver auto-target binding function enables the Fibre Channel driver to automaticallyconnect a server to target devices without requiring their definitions in fcp-bind-target in fjpfca.conf.
The Fibre Channel driver connects all target devices attached to a fabric device to the lowest availableTarget_ID in ascending order of WWNs.
1. Fibre Channel driver auto-target bind example
Server
FC Card
FC Switch
ETERNUS #2
CM0 CM1
ETERNUS #1
CM0 CM1
ETERNUS #1 CM0(WWN) : 0x210000e0004101d9CM1(WWN) : 0x230000e0004101d9
ETERNUS #2 CM0(WWN) : 0x210000e0004101daCM1(WWN) : 0x230000e0004101da
Target_ID:0Target_ID:3
Target_ID:2 Target_ID:1
112
In the example shown above, the Fibre Channel driver automatically connects target disk array devices inascending order of WWNs as follows:
Connect ETERNUS #1 CM0(WWN=0x210000e0004101d9) to Target_ID:0
Connect ETERNUS #1 CM1(WWN=0x230000e0004101d9) to Target_ID:2
Connect ETERNUS #2 CM0(WWN=0x210000e0004101da) to Target_ID:1
Connect ETERNUS #2 CM1(WWN=0x230000e0004101da) to Target_ID:3
Prerequisites for implementing this automatic connectivity are:
1. "fcp-auto-binding function=1;" is entered in fjpfca.conf.
2. SAN Boot environment built on a fabric connection
The Fibre Channel driver auto-target binding function is designed with primary emphasis on building aSAN Boot environment effortlessly, and its use is recommended only for the purpose of building anenvironment.
If the Fibre Channel driver auto-target binding function is used in normal operations, it might fail tocomplete intended target device connections owing to the effect of the failure of target devices or the like.The use of fcp-bind-target to make target device connections is recommended for normal operations.
For information about marking target device connections using fcp-bind-target, refer to the "FUJITSUPCI Fibre Channel Guide."
113
E.2 Fibre Channel Boot Code Auto-TargetBinding Function
This section describes the Fibre Channel boot code auto-target binding function.
The Fibre Channel boot code auto-target binding function lets the Fibre Channel boot code detect targetdevices automatically without requiring their definitions in fjpfca-bind-target, allowing only those targetdevices having the lowest Port-ID assigned to them to be connected to a server for SAN Boot.
1. Fibre Channel boot code auto-target bind example
Server
FC Card
ETERNUS #1
CM0 CM1
Port_ID:10100 Port_ID:10400
Port_ID:10200
Port_ID:10300
ETERNUS #2
CM0 CM1
FC Switch
Port_ID:10000
Connected as targetdevices to implementSAN Boot
In the example shown above, target devices located at ETERNUS#1 CM0 with the lowest assignedPort_ID are connected to implement SAN Boot.
Prerequisites to implementing this automatic connectivity are:
1. No target definitions are given in fjpfca-bind-target
2. SAN Boot environment built on a fabric connection
Because the Fibre Channel boot code auto-target binding function focuses on effortless implementation ofSAN Boot, it use is recommended only in environment construction and in environments in which FCswitch-based zoning is implemented. If the Fibre Channel boot code auto-target binding function is usedelsewhere, it might fail to complete intended target device connections owing to the effect of the failureof target devices or the like. The use of fjpfca-bind-target to manually configure target deviceconnections is recommended in incompatible environments. For information about making target deviceconnections using fjpfca-bind-target, see Appendix A, "Boot Device Setup Commands."
If SAN Boot is implemented with the boot code auto-target binding function, the target deviceinformation is imparted to the Fibre Channel driver, enabling it to make target device connectionsautomatically (fcode-auto-bind function). For more information about the fcode-auto-bind function, referto "FUJITSU PCI Fibre Channel Guide."
114
Appendix F SAN Boot release procedure
If you release the multipath definition of the boot disk, it does the following procedures.
It becomes impossible might do boot when it releases by the methods other than the procedure or itmakes a mistake in the procedure.
After the mirroring is released, the following procedures are executed when mirroring it withPRIMECLUSTER GDS.
F.1 ETERNUS Multipath DriverIt is ETERNUS multipath driver's release procedure.
All multipath are released.
1. In the SAN Boot environment by the UFS file system, edit /etc/vfstab file to rewrite mountpoint.
In the SAN Boot environment by the ZFS file system, it advances to Step 2.
/dev/FJSVmplb/dsk/mplb0s0 /dev/FJSVmplb/rdsk/mplb0s0
↓ ↓
/dev/dsk/c2t16d0s0 /dev/rdsk/c2t16d0s0
Former mount point is either path constituting multipath displayed by the iompadm command.
# /usr/opt/FJSViomp/bin/iompadm info /dev/FJSVmplb/fiomp/adm0
IOMP: /dev/FJSVmplb/fiomp/adm0
Element:
/dev/rdsk/c2t16d0s2 online active block "good status with active
[E30004641- 130011-CM01-CA01-PORT36] (mplbt0)"
/dev/rdsk/c3t16d0s2 online standby block "good status with standby
[E30004641- 130011-CM00-CA00-PORT32] (mplbt32)"
2. Release the multipath.
# mplbconfig -r
Cannot unload module: mplb
Will be unloaded upon reboot.
Forcing update of mplb.conf.
115
3. Edit /kernel/drv/mplbt.conf file to delete all the definitions.
Example)The following delete.
name="mplbt" parent="fjpfca" target=16 lun=0;
…
4. Edit /kernel/drv/mplbh.conf file and make it default setting.
Example)Delete all the following definitions of the mplbh.conf file.
mplbh-path-0="pci10cf,1178-0-10" mplbh-path-1="pci10cf,1178-1-10"
mplbh-disk-name="E30004641- 130011-0010";
mplbh-detect-disk-num=1;
mplbh-detect-disk-0="E30004641- 130011-0010";
mplbh-used-path-num=2;
mplbh-used-path-0="pci10cf,1178-0-10";
mplbh-used-path-1="pci10cf,1178-1-10";
It adds when there is no ";" at the end of the following lines.
name="mplbh" parent="mplbx" instance=X;
5. Edit the /kernel/drv/sd.conf file to delete the definition of mplb.
Delete the following lines.
# Start eternusmpd configuration -- do NOT alter or delete this line
name="sd" parent="mplbh" target=0 lun=0;
…
# End eternusmpd configuration -- do NOT alter or delete this line
6. Delete the following definitions added to /etc/system file.
● SAN Boot environment by the UFS file systemDelete the following lines.
rootdev: /pseudo/mplb@0:a
forceload: drv/mplbt
forceload: drv/mplb
forceload: drv/sd
● SAN Boot environment by the ZFS file systemDelete the following lines.
116
forceload: drv/mplbt
forceload: drv/mplb
forceload: drv/sd
7. Stop the service.
# svcadm disable -t svc:/system/fjsvmplb:default
8. Make a special file of sd. Ignore the error message.
# update_drv -f sd
Cannot unload module: sd
Will be unloaded upon reboot.
Forcing update of sd.conf.
9. Restart the server.
# touch /reconfigure
# reboot
10. When the dump device changed, it returns. (Only the SAN Boot environment by the UFS filesystem)
# dumpadm -d /dev/dsk/c2t16d0s3
117
Appendix G Updating the Fibre ChannelCard Boot Code
This appendix explains how to update the boot code (firmware) of the Fibre Channel card.
It explains the procedure for updating the boot code to V12L40 as an example. To confirm the boot codeversion please see Appendix B, "Checking the Fibre Channel Card Boot Code Version Number."
1. Apply the driver patch
The FJPFC driver patch necessary for the update of the boot code is applied to the system volume formaintenance. Apply the 914583-07 or later patch.
The system volume for maintenance indicates the system volume on the server for the network boot orthe system volume on the internal disk.
When the system volume for maintenance is on the internal disk
(1-1) Applies patch to the internal disk.
# patchadd 914583-07 <RETURN>
(1-2) Boot the server with a system volume for maintenance.
# shutdown -g0 -y -i6 <RETURN>
When the system volume for maintenance is on the network install server
- Solaris 10 5/08 or older
(1-1) Apply the patch in the OS install image on the install server.
(INSTALL SERVER) # patchadd -C/export/install/Solaris10_hostname/Solaris_10/Tools/Boot/ 914583-07 <RETURN>
(1-2) Start the OS from the system volume for maintenance.
ok boot net -s <RETURN>
- Solaris 10 10/08 or higher
(1-1) Create the working directory for unpacking miniroot.
(INSTALL SERVER) # mkdir /tmp/work <RETURN>
118
(1-2) Unpack the miniroot to the work directory using the root_archive(1M) command.
If /tmp/work/tmp/AdDrEm.lck file is not exist, following procedure ignoring.
(INSTALL SERVER) # /boot/solaris/bin/root_archive unpackmedia/export/install/Solaris10_hostname /tmp/work <RETURN>
(INSTALL SERVER) # rm /tmp/work/tmp/AdDrEm.lck <RETURN>
Note) The following messages might be displayed when root_archive command is executed. Youmight ignore these messages.
umount: /tmp/mnt29984 busy
rmdir: directory "/tmp/mnt29984": Directory is a mount point or in use
lofiadm: could not unmap file /export/install/Solaris10_hostname/boot/sparc.miniroot:Device busy
rmdir: directory "/tmp/mnt29984": Directory is a mount point or in use
(1-3) Apply the patch in the working directory.
(INSTALL SERVER) # patchadd -C /tmp/work 914583-07 <RETURN>
(1-4) Pack the working directory.
(INSTALL SERVER) # mkdir -p /tmp/media/Solaris_10 <RETURN>
(INSTALL SERVER) # /tmp/work/boot/solaris/bin/root_archive packmedia /tmp/media/tmp/work <RETURN>
The error messages might be displayed when root_archive command is executed. You mightignore these messages.
(1-5) Copy the file in /tmp/media directory to the installation image on the install server.
Target device path name for "umount -f" and "lofiadm -d" commands can confirm by "df -k"command.
(INSTALL SERVER) # cd /tmp/media <RETURN>
(INSTALL SERVER) # find boot Solaris_10/Tools/Boot | cpio -pdum/export/install/Solaris10_hostname <RETURN>
(INSTALL SERVER) # umount -f /dev/lofi/1 <RETURN>
(INSTALL SERVER) # lofiadm -d /dev/lofi/1 <RETURN>
The error messages might be displayed when root_archive command is executed. You mightignore these messages.
(1-6) Start the OS from the system volume for maintenance.
ok boot net -s <RETURN>
119
2. Update the boot code.
After OS in the system volume for maintenance starts, execute the update command.
# /usr/sbin/FJSVpfca/fc_fcode_update <RETURN>
This command updates the boot code to all cards mounted in the server. In the case of the latest bootcord, it is not updated.
Example
# /usr/sbin/FJSVpfca/fc_fcode_update <RETURN>
Are you ready to udpate FCode? [y,n] yes
Overwrite FCode to fjpfca0 (V12L30 -> V12L40).. Success
fjpfca1: Same version of FCode has already installed. (current version :V12L40)
Note)
- The setting of the card is kept by update.
- Do not execute it while execute I/O.
Umount is done beforehand when using it as a filesystem.
- There is no problem even if the optical cable is connected.
- If the following messages are displayed when the command is executed, the Fibre Channelcard might be defective. In this case, please exchange the cards for a new one.
WARNING: /pci@%x,%x/fibre-channel@%x (fjpfca%d):
ROMAccess failed.