© 2009 ibm corporation tws for z/os end-to-end & z centric implementation best practices

16
© 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

Upload: marshall-taylor

Post on 16-Dec-2015

235 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2009 IBM Corporation

TWS for z/OS

end-to-end & Z centric

Implementation Best Practices

Page 2: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation2

The goals Provide some best practices and suggestions related to the end-to-end features provided by

TWS for z/OS:– End-to-end with fault tolerant capabilities (aka e2e)– End-to-end with Z centric capabilities (aka Z centric)

They are intended to:– Improve the performances – Provide suggestions for a more effective usage of the feature

The Agenda

Overview

Z centric– Feature exploitation – Best practices

e2e– Feature exploitation– Best practices

Page 3: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

IBM Tivoli Workload Automation end-to-end features

... From a single point of control...

Single point of control to minimize the administrative oversight and time

Full impact view from the point of service delivery, for better efficiency, effectiveness and alignment of IT to business

Provide flexibility to establish the single point of control from any end-point

Heterogeneous workloads must be seen as homogeneous, and managed from a single point of control through a unified paradigm

End-to-end workload automation

3

Page 4: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

End-to-end with fault tolerant capabilities

z/OS z/OS SysplexSysplex

---------- ISPF ----------

Open Open systemssystems

TWS for z/OSTopology Server

End-to-e

nd

Domain Manager

Fault Tolerant Agent

Extended Agent

4

It is the Plan-based End-to-end, this means:

Hierarchical topology Optional multi-level configuration Communication layer, USS Server Network Fault Tolerance Distribution of plan to agents Monitoring allowed at agent level

TWS for z/OS controller

Page 5: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

End-to-end with Z centric capabilities

z/OS z/OS SysplexSysplex

---------- ISPF ----------

Open Open systemssystems

Z-centric

Extended Agent

Z-centric agent

5

The TWS controller acts as choreographer, this means:

Flat topology Simplified deployment and configuration No communication layer, no Server Fully centralized and homogeneous control Direct control over distributed workload

TWS for z/OSTopology Server TWS for z/OS

controller

Page 6: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

Cross dependencies

6

They allows to manage heterogeneous systems by:

– Defining in one scheduling environment dependencies on batch activities that are managed by another scheduling environment

– Controlling the status of these dependencies by navigating from a unique user interface across the different scheduling environments

– It is works both for TWS and TWS for z/OS

Cross dep.

TWS 1TWS for z/OS

controllerTWS 2

TWS MDM

TWS 3TWS for z/OS

controller Cross dep.

Page 7: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

The z centric feature can be easily activated, bycustomizing the controller PARM member, by:

– Setting the HTTPOPTS parameters– Adding in the ROUTOPTS statement the

HTTP/HTTPS parameters

It’s very easy to add, delete or modifyz centric WSs.

– There is a MODIFY commanddynamically reloading thedestination definitions: /F TW1A,RFRDEST

The same Z centric agent can run workload for various TWS for z/OS controllers at the same time

– This greatly simplify the infrastructure– Reduces the maintenance effort

7

Z centric – feature exploitation (1)

Z centricAgent

TWS for z/OS controllers

Page 8: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

It is possible to define more Z centric WSs having the same destination, this:

– Allows to globally act on a subset of the overall workload ran by a certain server– can be very useful if the same server runs workload related to different LOBs.

The Z centric WSs grant the same flexibility of the z/OS WSs, this means:

– Open time intervals;– Alternate WSs;– Parallel servers

It is possible to use the TWS for z/OSvariables to tailor in a “centralized way” theworkload on many distributed servers.

– In the example on the right the suppliedvariable related to the extended op. name,is used to parameterize the remote file name of an FTP job

8

Z centric – feature exploitation (2)

Page 9: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

The z centric agents support the filewatch feature

– It’s an executable able to perform advanced file discovery – File creation, deletion, edition. An example:

filewatch -condition wcr -filename C:\ftpdir\ftp.file -int 30 -deadline 0

It is can be very usefully integrated with the FTP job executor toautomate file discovery and transfer scenarios.

Just create an application running z centric jobs where:

1. The first one runs the filewatch executable2. Its successor run an FTP job.3. Subsequent jobs perform the file content

elaboration

9

Z centric – feature exploitation (3)

Z centricAgent

TWS for z/OS controller

file

DATASETfile elaboration

FTPFilewatchfile discovery

Page 10: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation10

Z centric – best practices The settings defined in the EQQUX001 override the other settings. A typical error scenario:

– If a job ends in error with ext. status OSUB– It may be the user submitting the job is that defined in the exit and not the right one

The TWS controller has to resolve the IP addrs of the z centric agents and vice versa. A typical error scenario:

– The TWS for z/OS user interface shows a jobs in “started” status– The job has been really submitted on the server hosting the z centric agent– In this case setting the HTTPOPTS HOSTNAME keyword can solve the issue.

Consider that the TWS controller tries to connecta Z centric agent only when it has to run the firstdaily job.

– Check on the real status of the agent by scheduling a TSO WSSTAT command, i.e.

WSSTAT SUBSYS(TW1A) WSNAME(ZAGT) STATUS(A)

– This can be very useful for agents running workload during the night

Page 11: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

If an FTA is NOT a backup DM than always set CPUFULLSTAT(OFF), this:

– Greatly reduces the network traffic– Reduces the number of events the active

DMs have to manage

Always use the mailman servers by setting the CPUSERVER keyword when defining an FTA or a standard agent. This:

– Increases the event handling speed performed by the DMs

– Make the e2e network more robust in case of network problems

Always keep the agents running jobs linkedby predecessor-successor dependencies in the same domain, this:

– Reduces the network traffic– Increases the fault tolerance level

11

E2E – feature exploitation (1)

FTA2

DM

FTA1

TWS domainTWS domain

Page 12: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation

When designing e2e applications consider they will be “mirrored” in Symphony Job streams. Take in consideration the main “mirroring” rules:

– The jobs defined on FTWS are present in theSymphony file

– Just the direct predecessors of those jobs (even if they are not scheduled on FTWS) are present inthe Symphony file as well.

Designing the applications so to have smallerSymphony Job streams grant various positive effects:

– Reduce the DP batch duration thanks to the shorter time needed to create the Symphony file

– Increase the performances when e2e applicationsare dynamically added in the CP

– Increase the performance of the TWS agents especially when they runs not centralized scripts

12

E2E – feature exploitation (2)

Current Plan

MasterDomain Manager

OPCMASTER

Domain ManagerDOM1

FTAFTW1

Symphony

Symphony

FTAFTW2

Symphony

Page 13: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation13

E2E – feature exploitation (3)

10

5

15

20

25

255

30

35

255

15

App2

App3

App1

10

10

App4

App5

10

5

15

2030

App1

255

15

App2

App3

It can be useful adding a dummy predecessoron NON REP WS (see op. 15)

The Symphony file can contain unconnected operations, so creating applications whosejobs are connected just by a dummy successorcan make smaller Symphony job streams(see ops. 30 and 20)

z/OS Plan

Symphony

Page 14: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation14

E2E – best practices (1)

Consider to increase the size of the datasets EQQTWSCS, EQQTWSIN EQQTWSOU

– Event loss could occur if thy are note well sized– Especially if a lot of dynamic additions are performed

Take in consideration the evtzise cmd to increase the max reachable size of the event files (such as Mailbox.msg or Intercom.msg)

– This works both for the files present in the e2e server work directory and for the TWS agents

– The manual “Scheduling End-to-end with Fault Tolerance Capabilities” documents this command usage.

In case of server maintenance take in considerationto set the CPUREC parameter CPUAUTOLNK(OFF)

– It makes “manual” the TWS agent initialization – No time is wasted in trying to initialize it– A Symphony renew is sufficient to activate it

Page 15: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation15

E2E – best practices (2)

The workload throughput of the agent can be globally managed by using the CPUREC CPULIMIT parameter.

– This is very useful if a server hosting a TWS agent seems to be too much stressed– By setting it to 0 it’s possible to keep the agent active avoiding the job submission– The parameter can be changed:

• by submitting a symphony renew• In dynamic way, by using the TWS agent admin. CLI conman

(i.e.: “conman lc FTA1; 0” )

If different kind of workloads have to be run on the same server (i.e. they refer to different LOBs)

– It’s NOT needed the installation of moreagents (this could stress the server)

– Some agents could be “simulated” by using local UNIX extended agents

– This is possible just on UNIX servers

Page 16: © 2009 IBM Corporation TWS for z/OS end-to-end & Z centric Implementation Best Practices

© 2011 IBM Corporation16