This week in Mozilla RelEng – February 28th, 2014
Highlights:
- Catlee worked on a lot of tooling for Jacuzzis, and we enabled them for more builders, which is greatly speeding up turnaround time for them.
- He also changed the algorithm used when spinning up AWS machines which has significantly reduced the number of unused hours our machines run for.
- Armen submitted a talk proposal for next year's RelEng Workshop.
- John Hopkins got a large portion of machines sending data to hosted Graphite service which will help us analyze utilization more easily.
- John Zeller finished work on a patch that allows developers and sheriffs to trigger builds for any revision. This is a feature that's been wanted for quite awhile, and helps pave the way for things like automatic bisection.
- Pete and Simone learned the basics of Buildduty from Coop, which should help us be more responsive to infrastructure issues during the European working day.
- Balrog: Backend
- API for manipulating balrog rules
- balrog should return empty snippets instead of 500s
- Balrog shouldn't serve updates to older builds
- Buildduty
- buildbot-master70 is having trouble with puppet
- Integration Trees closed, high number of pending linux compile jobs about ±2 hour backlog
- What happened at 18:00 PT that disconnected Windows builds across all trees?
- upload mozlog-1.5 to http://pypi.pub.build.mozilla.org/pub
- Deploy python 2.7.3 to all build machines
- upload mozinfo-0.7 to http://pypi.pub.build.mozilla.org/pub
- General Automation
- B2G sources.xml (or
) should state which github repos map to the repos/revisions listed - Rooting analysis mozconfig should be in the tree
- [Meta] Tree closing/reliability issues with builds-pending.js, builds-running.js, builds-4hr.js.gz or buildapi in general
- aws_watch_pending.py should use jacuzzis
- Set --timeout 10000 on gaia-ui-test command line
- Don't reset mock environments if we don't have to
- Schedule multimedia b2g mochitests for emulator-jb on Cedar
- Do our own hg purge
- aws sanity check shouldn't report instances that are actively doing work as long running
- move off of dev-master01
- Update b2g_bumper.py with a git dict
- Remove 'update_files' logic from B2G unittest mozharness scripts
- rewrite watch pending to cope better with spot requests that aren't being fulfilled
- Use spot instances for regular builds
- write mozconfigs for non-unified builds
- Mozharness swallows return_code of a talos run if it is 1
- [tracking bug] migrate hosts off of KVM
- Create IAM roles for EC2 instances
- Configure holly like Aurora properly
- Remove linux-opt arm-sim builder
- Loan Requests
- Please loan dminor Android 2.3 Emulator test instance
- Please loan dminor instance to build Android 2.3 Emulator
- Linux 64-bit test slave for bug 926264
- loan linux slave to jmaher
- Other
- buildbot master manifests should ensure firewall is disabled
- Create (private?) repo holding TEF customization data
- The stage.m.o:/mnt/cm-ixstore01 mount is filling up
- Need to increase file size limit for mar files to enable linux+mac mar signing
- selfserve agent drops requests
- Platform Support
- Re-allocate tegras that were on decommissioned foopy118
- [tracker] move jobs away from obsoleted machines
- remove joduinn from puppet controlled access
- Setup in-house buildbot masters for remaining in-house testers
- release-automation
- replace TuxedoEntrySubmitterFactory with a mozharness script
- Support modifying update-settings.ini when doing update verify
- Create SSL products in bouncer as part of release automation
- Request notification of events affecting release timelines
- Repos and Hooks
- Request for a /gaia-l10n/lt repository
- Please create git.mozilla.org mirror for https://github.com/mozilla-b2g/gaia-node-modules
- Tools
- Balrog: Backend
- Balrog: Frontend
- Buildduty
- General Automation
- Override ASan malloc_context_size option
- ASan is the only Linux platform running the separate jittest suite on try
- Self-serve should be able to request arbitrary builds on a push (not just retriggers or complete sets of dep/PGO/Nightly builds)
- Provision enough in-house master capacity
- Tracking bug for 17-mar-2014 migration work
- Change birch config and add jacuzzis
- Provide B2G Emulator builds for Darwin x86
- Do debug B2G desktop builds
- decommission (releng side) aws buildbot masters migrated to scl3
- Add support for add-if-not instruction added by bug 759469 to the mar generation scripts
- Tweak buildername for Android l10n nightlies to make them more consistent with other multi-part jobs
- use tbirdbld account to submit thunderbird data to balrog
- [Meta] Some "Android 4.0 debug" tests fail
- Remove --no-update from B2G media test invocation
- Normalize builder names
- [Meta] Fix + unhide broken testsuites or else turn them off to save capacity
- b2g_build.py steps aren't independent
- Please schedule linux jit-tests in two chunks
- add --dump-config and --dump-config-hierarchy to mozharness
- port MockReset to mozharness
- allow for multiple basedirs to be passed to PurgeMixin.purge_builds()
- allow mozharness's tbox_print_summary() to be used outside of unittests
- Schedule gaia-integration tests on OSX on cedar
- Create SpiderMonkey builds for Windows on TBPL again
- desktop_unittest.py should allow using existing files
- Do nightly builds with profiling disabled
- Install ant on builders
- Add Linux32 debug SpiderMonkey ARM simulator build
- Run mozbase unit tests from test package
- Allow the possibility of spidermonkey builds on more platforms
- Please add non-unified builds to mozilla-central
- Device Builds Tests/Builds for mozilla-inbound
- submit release builds to balrog automatically
- Intermittent Linux spot builder "command timed out: 2700 seconds without output, attempting to kill" while trying to install mock
- Loan Requests
- loan windows 7 slave to jmaher
- Loan a B2G emulator machine to baku
- Loan t-w864-ix-003 to jrmuizel
- Loan t-w732-ix-003, t-w732-ix-004 to Q
- Loan felipe an AWS unit test machine
- Other
- give larsberg access to servo machines
- [tracking] infrastructure improvements
- Decide on document workflow for releng docs
- s/m1.large/m3.medium/
- stage NFS volume about to run out of space
- Make AWS node type available to graphite & build metadata
- Platform Support
- Deploy git to windows builders
- Update wget to 1.15 on foopies
- Migrate ESR, b2g26_v1_2, b2g28_v1_3 branches to win64-rev2
- Update builders to Android SDK platform 17
- signing win64 builds is busted
- Update VC++2010 runtime to MS11-025 updated version
- Add msttcorefonts package to Servo buildbots
- Produce crash report symbols for darwin-on-linux builds
- foopy watch_devices does not clear lockfile when it fails to get data from slavealloc
- Configure VS2012 builds in staging
- Deploy new version of MozillaBuild from bug 927213 on windows build slaves
- [tracker] Determine how to support and roll out darwin-on-linux in production
- deploy mac on linux build tools to tooltool
- puppet should setup directory structure on foopies
- Update all foopies to Python 2.7.3
- Deploy patched version of Mesa 8.0.4
- Upgrade Windows 8 Nvidia drivers to 334.89
- Setup Diamond to send metrics to hosted graphite
- Create a Windows-native alternate to msys rm.exe to avoid common problems deleting files
- release-automation
- Releases
- Releases: Custom Builds
- Fix outdated and duplicated landing pages for MSN/Bing builds
- Update Rambler pack with updated extension 4.9.3
- Repos and Hooks
- Tools
- more vcs-sync email tweaks
- build-tools: Can't push /opt/vcs2vcs/vcs_sync/build/conversion/build-tools to build-tools-github!
- Slaveapi should support mobile devices for reboots
- Add an option to trychooser to select Talos profiling options
- Create a new landing page for releng
- slaveapi still files IT bugs for some slaves that aren't actually down
- legacy vcs-sync process doesn't auto retry when it should
- Add buildbot database sources into Tableau