Organization and Management of ATLAS Nightly Builds F. Luehringa, E. Obreshkovb, D.Quarriec, G. Rybkined, A. Undruse

University of Indiana, USAa, DESY, Germanyb , LBNL, USAc ,LAL, Franced ,BNL, USAe

Organization and Management of ATLAS Nightly Builds F. Luehringa, E. Obreshkovb, D.Quarriec, G. Rybkined, A. Undruse

University of Indiana, USAa, DESY, Germanyb , LBNL, USAc ,LAL, Franced ,BNL, USAe

Abstract

25 branches of automated multi-platform software nightly builds is a major component in ATLAS collaborative software organization and code approval scheme. The successful nightly releases are transformed into stable releases used for data processing worldwide. ATLAS nightly builds are managed by the fully automated NICOS framework [1]. The modular structure of NICOS allows for an easy integration of third-party build and validation tools. The ATN [2] test tool is embedded within the nightly system and provides the first results even before the full compilations completion. Several ATLAS test frameworks are synchronized with NICOS jobs and run larger production jobs with the nightly releases. NICOS web pages dynamically provide information about the progress and results of the builds. For faster feedback the e-mail notifications about nightly build problems are automatically distributed to responsible developers.

Abstract

25 branches of automated multi-platform software nightly builds is a major component in ATLAS collaborative software organization and code approval scheme. The successful nightly releases are transformed into stable releases used for data processing worldwide. ATLAS nightly builds are managed by the fully automated NICOS framework [1]. The modular structure of NICOS allows for an easy integration of third-party build and validation tools. The ATN [2] test tool is embedded within the nightly system and provides the first results even before the full compilations completion. Several ATLAS test frameworks are synchronized with NICOS jobs and run larger production jobs with the nightly releases. NICOS web pages dynamically provide information about the progress and results of the builds. For faster feedback the e-mail notifications about nightly build problems are automatically distributed to responsible developers.

Nightly FrameworkNightly Framework

References[1] A. Undrus, CHEP 03, La Jolla, USA, 2003, eConf C0303241, TUJT006 [hep-ex/0305087]; http://www.usatlas.bnl.gov/computing/software/nicos/index.html

[2] A. Undrus, CHEP 04, Interlaken, 2004, Conference proceedings, p. 521

[3] http://www.cmtsite.org

[4] Tag Collector reference

[5] AMI reference

[6] PackDist reference

[7] RTT reference

[8] FCT reference

[9] TCT reference

[10] KV rererence

References[1] A. Undrus, CHEP 03, La Jolla, USA, 2003, eConf C0303241, TUJT006 [hep-ex/0305087]; http://www.usatlas.bnl.gov/computing/software/nicos/index.html

[2] A. Undrus, CHEP 04, Interlaken, 2004, Conference proceedings, p. 521

[3] http://www.cmtsite.org

[4] Tag Collector reference

[5] AMI reference

[6] PackDist reference

[7] RTT reference

[8] FCT reference

[9] TCT reference

[10] KV rererence

NICOS FRAMEWORK

NIGHTLY CONFIGURATION

DISTRIBUTION

CODEMANAGEMENT

TESTING

nightlyrelease

CMT build

ATN testing

AMI DB

Tag Collector

CVSRepository

RTT

FCT

TCT

KITVALIDATION

KITSCERN AFSATLAS

TierCenters

NICOSWEB PAGES

NICOS is connected with ATLAS collaborative tools:CMT [3] code management and build toolTag Collector [4] web based tool for managing the tags of packages in releaseATLAS CVS code repositoryATLAS metadata DB (AMI) [5] for storage of NICOS configurations for different nightlies branchesNightly releases are installed for worldwide access at CERN AFS and also can be downloaded with ATLAS distribution kits tools [6]Integrated ATN [2] and external RTT [7], FCT [8], TCT [9], Kit Validation [10] testing frameworks

NICOS Web PagesNICOS Web Pages

All nightlies summary with message of day, distribution kit and tests status

Nightlies computer farm statuswith dynamic information on nodes

Log files with error diagnostics with links to problem messages

Nightly branch summary with results for inside projects and platforms

Summary for project/platformwith latest releases comparisons

Nightly release summary with diagnostics for individual packages

Software Release StructureSoftware Release Structure

AtlasOffline

ATLAS PROJECTS

ATLAS PATCH PROJECTS

PACKAGESa) Containers: directories for related packagesb) Leaf: source code and/or scriptsc) Glue: interfaces to externals ……….Number of leaf + container packages:

EXTERNALS

AtlasAnalysis

AtlasTrigger

AtlasReco

AtlasSimulation

AtlasEvent

AtlasConditions

AtlasCore

GAUDI Framework

DetCommon

Data QualityMonitoring

Data AcquisitionSoftware

Interfacesto externals(“LCGCMT”)

Ext

erna

ls (

RO

OT

, G

eant

, …

)

26+20

234+84

187+30

452+117

190+35

293+90

206+55

178+20

15+5

Total: 1781 leaf,456 container

packages as of 02/15/09

Projects are groups of packages with similar dependencies managed by CMT [3]. They are built as units and can evolve at different rates. Patch projects sit at the top of project hierarchy. They contain override versions of packages from downstream projects.

Software ValidationSoftware Validation

Release Coordinatorensure design goals

Chief Architect

Validation Data Managerprovides data samples

Production Manageroversees jobs definitions

Project Coordinators: evaluate new software submissions

Package Managers: request upgrades through Tag Collector

New software versions are verified in validation or migration nightlies. Upon approval from coordinators they are moved to the primary branch. Stable software releases are created from successful primaries. All patches to stable releases are sent to primary branch. Package versions associated with releases and project dependencies are handled by the Tag Collector.

Primary development nightly branch

Validation nightly branch

Migration nightlies:Verification of disruptive changes

LCG nightlies:Verification of externals upgrades

Stablerelease

Stablerelease

Stablerelease

Patchrelease

Patchrelease

ATLAS Nightlies in Numbers

Number of branches: 25 Total number of platforms in all branches: 40 All nightlies are rebuilt and tested daily Old nightly releases are kept for 2 - 7 days Nightlies computing farm:• 40 nodes, 4x3.0 or 8x2.33 GHz• 8 - 16 GB RAM Time to rebuild nighty release: 10 hours Number of ATN tests: 300 Additional time to complete ATN tests: 5 hours

ATLAS Nightlies in Numbers

Number of branches: 25 Total number of platforms in all branches: 40 All nightlies are rebuilt and tested daily Old nightly releases are kept for 2 - 7 days Nightlies computing farm:• 40 nodes, 4x3.0 or 8x2.33 GHz• 8 - 16 GB RAM Time to rebuild nighty release: 10 hours Number of ATN tests: 300 Additional time to complete ATN tests: 5 hours

NICOS DesignNICOS nightly jobs consist of tightly synchronized streams. The results for different platforms are combined in a single release.

NICOS DesignNICOS nightly jobs consist of tightly synchronized streams. The results for different platforms are combined in a single release.

Tag Collector access Code checkout proj.1 Code checkout proj.2 Code checkout proj.3

Build project 1 Build project 2 Build project 3

Install 2Install 1

Tests project 1 Tests project 2

Build project 1 Build project 2 Build project 3

Install 1 Install 2wait