Wikimedia Release Engineering Team/Checkin archive/20150310

From mediawiki.org

2015-03-10[edit]

Team Business[edit]

Team Quarterly Goals[edit]

https://phabricator.wikimedia.org/maniphest/query/O9isnUt5IGLP/#R Next quarter goals...

Made by groups. Our is ops + all platform Next year's budget: 

   * send doc to team (Greg)

Scrum of Scrums[edit]

https://phabricator.wikimedia.org/project/board/64/
Blocked on us: https://phabricator.wikimedia.org/maniphest/?statuses=open%2Cstalled&allProjects=PHID-PROJ-arpazvuktn2l647rb6us#R

Beta Cluster stability[edit]

https://phabricator.wikimedia.org/project/board/497/?order=priority
Quarterly Priority: Green nightly builds on staging: https://phabricator.wikimedia.org/T88701
  • Working on merging staging and production into unified puppet roles
Quarterly Priority: Stable uptime metrics of the Staging cluster: https://phabricator.wikimedia.org/T88705
  • mmodell proposed and prototyped a solution to create metrics from the varnish proxy logs/stats. Essentially the metric is error_rate = 5xx_responses / 2xx_responses (per time period)
  • Something went badly wrong with either beta or Jenkins or both overnight 9 March PST (aka last night?). I've looked in the logs on beta labs and didn't find any smoking guns.

Test history[edit]

Quarterly Priority: By team test history: https://phabricator.wikimedia.org/T88706
  • Outlined a few different options with varying scope and complexity
  • Option #2 is worth a spike (Mama Bear's option; just right); can fallback on option #1
    • Need elasticsearch node for storing build/cucumber results (extend json formatter for structured build info and cucumber scenario results)
    • Dashboard to display results

Trying to work around a Selenium bug with Chrome and how it interacts with WMF-style "overlays". Affects MobileFrontend, Echo, probably VisualEditor repos at least. Antoine: 

   That one is new to me. OpenStack send results to logstash / ElaticSearch and build artifacts to Swift. One can reach out #openstack-infra to figure out how it is done for them.

   IIRC they are using a test result protocol named 'subunit' and have plans to collect all their tests in a huge central DB to build report from.   Subunt infos: http://www.tech-foo.net/making-the-most-of-subunit.html

   Dan: slide 29 of http://docs.openstack.org/infra/publications/2014-gerrit_user_summit-overview/#%2829%29

   Demo: http://logstash.openstack.org/

Isolated CI instances[edit]

CI board https://phabricator.wikimedia.org/tag/continuous-integration/board/?order=priority

Kunal "Legoktm" / Timo "Krinkle" have been quite busy.

https://phabricator.wikimedia.org/tag/continuous-integration-isolation/
Quarterly Priority: Disposable VMs - https://phabricator.wikimedia.org/T47499

Next:

  • test / refine it
  • create one for Trusty

Packaging with dh-virtualenv (which embeds python modules in the .deb) is a mess. We should probably migrate the whole infra to Debian/Jessie including the production boxes.

MediaWiki Releases[edit]

Quarterly Priority: Release MediaWiki 1.25: https://phabricator.wikimedia.org/T88709

Other Work[edit]

Hiring[edit]

Vacations/Confs/etc[edit]

  • Chad at Elasticon with Nik 10-11 March, in SF
  • Dan in France the week before offsite (May 11-15)
  • Elena on vacation May26-June07
  • Week before hackathon: Team offsite in France - https://phabricator.wikimedia.org/T89036
  • May: Hackathon in Lyon, France
    • Lyon used to be the capital of the area a long time ago. Nice old city, lot of great food, reasonably sunny/hot.
  • Chad vacation after offsite
Retrieved from "https://www.mediawiki.org/w/index.php?title=Wikimedia_Release_Engineering_Team/Checkin_archive/20150310&oldid=4952640"