Page MenuHomePhabricator

Mentor dashboard: M1 mentee overview: Update more frequently than daily
Closed, ResolvedPublic

Description

Thanks to the significant performance improvements made to updateMenteeData (T290609#7429319), we can now run the updates more frequently than daily.

I think updating every three hours is now possible from the technical PoV (potentially even hourly, but I want to keep some breathing room in case it slows down for some reason).

What would be the ideal update frequency from the product PoV? @Trizek-WMF @MMiller_WMF @RHo Maybe you have some opinions on this topic?

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald Transcript

Hi @Urbanecm_WMF - 3 hours sounds like a prudent improvement to daily based on your assessment. However I wonder if in future, we may want to add a indication of how 'fresh' the data is, such as an Updated Xmin ago label. And furthermore, if it is technically possible (and desirable) for someone to manually refresh/re-run the query?

Hi @Urbanecm_WMF - 3 hours sounds like a prudent improvement to daily based on your assessment. However I wonder if in future, we may want to add a indication of how 'fresh' the data is, such as an Updated Xmin ago label. And furthermore, if it is technically possible (and desirable) for someone to manually refresh/re-run the query?

Both things should be possible (and fairly easy). Because of the performance improvements, updating data for a given mentor should be reasonably fast (a couple of minutes at most). It will still run in the background, but we can let mentors to run the background job on-demand, no problem.

Hi @Urbanecm_WMF - 3 hours sounds like a prudent improvement to daily based on your assessment. However I wonder if in future, we may want to add a indication of how 'fresh' the data is, such as an Updated Xmin ago label. And furthermore, if it is technically possible (and desirable) for someone to manually refresh/re-run the query?

Both things should be possible (and fairly easy). Because of the performance improvements, updating data for a given mentor should be reasonably fast (a couple of minutes at most). It will still run in the background, but we can let mentors to run the background job on-demand, no problem.

That's great, thanks! I created T293454 for this as a improvement.

Experienced users (hence mentors) are used to instant refresh. This is how Watchlist or RecentChanges work. Our table of activity is a recap, and filters in RCs are available for instant monitoring.

Maybe we should:

  • explain how the table is updates
  • point our the existing filters

Change 731192 had a related patch set uploaded (by Urbanecm; author: Urbanecm):

[operations/puppet@production] growthexperiments/updatementeedata: Run updates every three hours

https://gerrit.wikimedia.org/r/731192

Change 731192 merged by Dzahn:

[operations/puppet@production] growthexperiments/updatementeedata: Run updates every three hours

https://gerrit.wikimedia.org/r/731192

[mwmaint1002:~] $ sudo systemctl list-timers | grep MenteeData
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:34 UTC  22min ago          mediawiki_job_growthexperiments-updateMenteeData-s1.timer              mediawiki_job_growthexperiments-updateMenteeData-s1.service
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:34 UTC  22min ago          mediawiki_job_growthexperiments-updateMenteeData-s2.timer              mediawiki_job_growthexperiments-updateMenteeData-s2.service
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:35 UTC  22min ago          mediawiki_job_growthexperiments-updateMenteeData-s3.timer              mediawiki_job_growthexperiments-updateMenteeData-s3.service
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:35 UTC  22min ago          mediawiki_job_growthexperiments-updateMenteeData-s4.timer              mediawiki_job_growthexperiments-updateMenteeData-s4.service
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:35 UTC  21min ago          mediawiki_job_growthexperiments-updateMenteeData-s5.timer              mediawiki_job_growthexperiments-updateMenteeData-s5.service
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:36 UTC  21min ago          mediawiki_job_growthexperiments-updateMenteeData-s6.timer              mediawiki_job_growthexperiments-updateMenteeData-s6.service
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:36 UTC  21min ago          mediawiki_job_growthexperiments-updateMenteeData-s7.timer              mediawiki_job_growthexperiments-updateMenteeData-s7.service
Wed 2021-10-27 21:15:00 UTC  1h 31min left       Wed 2021-10-27 19:21:37 UTC  21min ago          mediawiki_job_growthexperiments-updateMenteeData-s8.timer              mediawiki_job_growthexperiments-updateMenteeData-s8.service

^ updated. one timer per DB cluster though that all start at the same time.

Change 735365 had a related patch set uploaded (by Urbanecm; author: Urbanecm):

[mediawiki/extensions/GrowthExperiments@master] Mentee overview: Updates now happen every 3 hours

https://gerrit.wikimedia.org/r/735365

Thanks @Dzahn, confirmed by looking at the logs. Uploaded a patch to inform users about more frequent updates, which is the last thing to do here.

Change 735365 merged by jenkins-bot:

[mediawiki/extensions/GrowthExperiments@master] Mentee overview: Updates now happen every 3 hours

https://gerrit.wikimedia.org/r/735365