CIDuty/SVMeetings/Mar13-Mar17

Meeting Details

• Room Link: https://v.mozilla.com/flex.html?roomdirect.html&key=RSAkZHawz3ED
• Kmoir’s room link: https://v.mozilla.com/flex.html?roomdirect.html&key=yUCdDe4Ly4bm

• Meetings every Wednesday
• Meeting notes archive:
  • https://wiki.mozilla.org/ReleaseEngineering/Buildduty/SVMeetings

• Upcoming vacation/PTO:
  • Alin:
  • Andrei:
  • Sebastian:
  • Florin:
  • Coop:
  • Kmoir:
• Holidays:
  • Canada:
  • Romania:

Possible bugs

• Write a tool to update bugzilla bugs for outstanding machine loans and ask if the loaner is still needed on say, a weekly basis: https://bugzilla.mozilla.org/show_bug.cgi?id=1171165
• Mitigating DoS Potential on BuildBot Masters web service on port 8201: https://bugzilla.mozilla.org/show_bug.cgi?id=1324802

From kmoir:

• Other possible bugs:
  • Is there further automation that you can implement to make regular tasks easier i.e. the machine loan email as a template
  • [alin]: as discussed in Hawaii, we’ll start investigating what improvements can be done here
• Other goals
  • Puppet master upgrades: we're on a really old version, and should have a better story for upgrades going forward. We can move a single platform to start(kmoir needs to open bug here and ask arr on current state)
  • Talked to Rok this week re moving releng web properties to releng services. He said that his documentation is ready to use and he is willing to pair program with people as needed. We could move one of this projects (slave health, buildapi, slavealloc) as a start, depending on what you think is most useful. https://docs.mozilla-releng.net/.
• Priorities update. Lawrence had a meeting with his mgmt team in Toronto recently and established the following ordered priorities
  • 1. Data centre move - two different data centres (will ask Amy for plan)
  • 2. OSX on TC (we need to do this so we don’t have to move mac builders into new DC)
  • 3. Quantum
  • 4. Security, Disaster recovery (part of having two data centres now). Securing the autoland pipeline
  • 5. Phase out aurora (Q1 planning)
  • 6. Reduce orange factor
  • 7. Blocking bad commits
  • 8. TC for Linux (almost done) talos needs to be done
  • 9. TC Windows
• How is your workload these days? Do you feel like there is too much to do or too little? What percentage of time are you spending on operational requests vs writing patches for automation etc?
• How are things after the s3 outage yesterday?

SV: 2017/03/20 - 2017/03/24

• New buidduty team member: Sebastian Păcurar (:spacurar) \o/
  • It would be nice if Sebi could get access to Nagios dashboard too (not urgent)
    • Kmoir - will ask arr about this
    • [alin] there’s a bug on file for this, still not working atm

• Bug 1332337 - migrate machine health dashboard to releng services
  • working on loading the informations for machine page (example)
  • Need to speak with Rok after his return to clarify some possible issue

• Bug 1233170 - Slave health page: details page opens several/multiple http auth requests
  • Callek spoke with :jabba and added Aryx to vpn_sheriffs group
  • Now he should be able to login to that page

• Bug 1338871 - Enable Talos tests for linux64-stylo builds
  • Still investigate why the linux64-stylo talos jobs aren’t appearing on the master
  • The last patch was not pushed so for talos-chrome we have linux64-stylo/.*: [] instead of linux64-stylo/.*: ['mozilla-central']

• Bug 1338239 - Remove tc() prefix from platforms that no longer exist in buildbot
  • currently investigating this
  • will need to find a way to differentiate platforms when setting up group symbols

• Releng tech-talks for the Buildduty team
  • we really appreciate the initiative and feel that’ll help us a lot
  • Link: https://docs.google.com/document/d/1EMneZ8F9YpxljpajvlE6mYA_xHKSBrCXbQ8AXsy_rmA/

• Bug 1309812 - Windows 7 VM reftest jobs are not running by default if 'Windows 7' is not explicitly specified
  • Asked catlee if we should have reftest jobs running by default on try, without specifying "Windows 7" in the try syntax.
    • Catlee answered that we should give it a try but we should keep a close eye on capacity and problems getting new instances from Amazon.

• Bug 1342825 - Don't run mochitest browser screenshots on PGO builds
  • I already provided a patch to remove "linux64-nightly opt" variant in TaskCluster (not landed yet)
  • I don't see an easy way to remove the pgo variant of mochitest-browser-screenshots.