MediaWiki Engineering/Runbook/Daily duties

Wikimedia infrastructure

Data centers

Networking

Global traffic routing

MediaWiki SRE

MediaWiki Engineering

• Parser cache
• MediaWiki JobQueue
• Performance review
• PHP upgrade process for MediaWiki
• performance.wikimedia.org
• Web Perf Hero award
• Guides:
  • Frontend best practices
  • Backend best practices
  • more…
• Runbooks:
  • Daily duties

Multimedia

Data Engineering

Search

SRE Data Persistence

SRE Infra Foundations

SRE Observability

Wikidata Platform

Wikimedia Performance

Event Platform

Release Engineering

Fundraising

[]

These are daily duties for weekly rotation among team members, as well as individual responsibilities that don't rotate.

Individual

For Phabricator tasks:

• Generally continue work on tasks in progress (i.e. current sprint) before starting work on other tasks.
• Review new comments on tasks you're subscribed to, especially comments that mention you or on tasks assigned to you. One approach is to use email notifications (how to: Reduce notifications).

For Gerrit code review:

• Your patches: Review failing tests or comments addressed to you on your own patches. If a patch is not ready for review, mark it as WIP or abandoning it, which helps maintainers by excluding it from certain most code review listings.
• Your reviews: Maintain awareness of patches that have specifically asked for your review. And, if you start reviewing a patch, review follow-up comments addressed to you.
• Incoming: As maintainer of a component, maintain awareness of new patches. We generally focus only on patches that are open, have passing tests, and are not marked WIP. Other patches are not considered as publicly seeking review (if they need help during this stage, they may ask on Phab, IRC, or by asking an individual for review). If a new patch seems small enough and aligned with current direction, consider reviewing it yourself, or assigning it to a team mate. Otherwise, consider acknowledging the patch by leaving a comment and setting expectations.

One approach to code review is through a team-specific Gerrit dashboard (how to: Add Gerrit navigation link).

The "Your turn" section lists changes that are open, not-WIP, and have you in the "Attention Set". Gerrit automatically adds you to the attention set when a comment is addressed to you (on your own patch, or as follow-up on one of your reviews), and when new patches ask your review.

The "Incoming" sections are used for maintainers to spread the load of new reviews (avoid using mw:Git/Reviewers if you follow this approach, since it might cookielick patches and overload your personal assignments).

MediaWiki Platform Team

You can check who is on duty in the rotation calendar.

Individual

• Phabricator workboard: MediaWiki Platform Team / Assigned to me
• Gerrit dashboard: MediaWiki Platform Team

Rotating

Daily:

• Maintain public presence on IRC in #mediawiki-core, and on mailing lists (Wikitech-l). As our look-out, look for questions and announcements that relate to our team's work. Feel free to answer questions if you feel able to. Otherwise, consider bringing the question to the attention of others in the team.
• Look into Grafana alerts sent to mediawiki-platform-team@ mailing list.
• Review the team inbox on Phabricator and work on the following (as for help or delegate as-needed). You can leave any other tasks in the inbox for the team triage on Monday. To follow Phabricator activity around our team you can optionally use the #mediawiki-core-bots channel on IRC.
  • Task for stuck global renames.
  • Task with "Unbreak Now" status.
  • Task marked as train blocker (child of "deployment blockers").
  • Task for production alert, created by SRE bots such as phaultfinder.
• Review Wikipedia:Village pump (technical) for issues related to our team's work and respond if you feel comfortable, or mention the discussion in our team channel if appropriate.

Once a week:

• Triage new PHP errors from the Logstash: MediaWiki Platform dashboard (how-to guide)
• Triage new JavaScript errors from the Logstash: MediaWiki Platform Client dashboard (check the old Reading team's guide to triaging client-side errors)
• Review ResourceLoader health dashboard:
  • Check Grafana: ResourceLoader Alerts dashboard, e.g., in case alert emails didn't trigger and for awareness of our outliers more generally.
  • Check the last 7 days of ResourceLoader service stats, and the last 14 days to compare current charts against last week.
  • Consider having a glance at a longer time frame, such as the last 30 days, as client-side regressions may be hard to notice until more time has passed.
  • Check MediaWiki-static service stats.
• Review the Authentication health dashboard:
  • Check the last 7 days of Authentication metrics, and the last 14 days to compare current charts against last week. Some of these can be irregular due to spambots, but huge changes should be noted.
  • Consider having a glance at a longer time frame, such as the last 30 days, as regressions may be hard to notice until more time has passed due to bot traffic noise.
• Review session store dashboard:
  • Check the last 7 days of session store metrics, and the last 14 days to compare current charts against last week. Some of these can be irregular due to spambots, but huge changes should be noted.
  • Consider having a glance at a longer time frame, such as the last 30 days, as regressions may be hard to notice until more time has passed due to bot traffic noise.

Content Transform Team

See Content Transform Team/Chores on mediawiki.org.