Wikimedia Discovery/Meetings/Search retrospective 2016-06-01
Covering whatever has happened related to the team since the last retro (2016-04-18)
- Team offsite
- Tomasz leaving; Katie joining
- Prep for Dan's vacation
- TextCat A/B test
- ElasticSearch upgrade
- Discernatron
- Search off-site
- Data mapping (did this happen between the last retro and this one?) -> it seems to be in the "What went well", so I guess yes... but I was the one that put it in there too :-p
Review action items from before
edit- Erik: Post action plans with rough timelines for finishing Q4 goals
- Team has shared understanding; nothing was posted. We won't carry this action item forward.
- Erik: Advance the conversation about improving browser tests
- Had various conversations, so
Done. It had been running well until yesterday (machine config issue unrelated to earlier problems).
- Had various conversations, so
- Chris: See if there are any past tests that weren't announced that should have been
- Didn't find any. Calling this one Done.
- Didn't find any. Calling this one
- Chris: Update Discovery page on Mw.org to reflect q4 goals - Done
- Chris: Request feedback on search team draft page - Done
What went well?
edit- Tomasz has been gone two days and nothing exploded (yet) (well, it was a holiday weekend ;) )
- The search offsite was productive and fun, and it was great to see everyone who could come +1
- Elasticsearch upgrade is so far relatively painless
- Guillaume is getting better at doing full elasticsearch cluster restarts
- Implementing feature requests for Reading is going well *
- Data mapping went well and was straightforward for us to do+1*
What could we improve?
edit- Monitor github activity on our plugins: e.g. https://github.com/wikimedia/search-highlighter/issues/18 **
- We can't properly test major production changes to search infrastructure because we can't duplicate LVS ***
- Didn't notice breakage of elasticsearch logging in beta, noticed after initial production rollout
- Not sure that I (Guillaume) give enough visibility on what I'm doing, since I'm split between different teams. Not always easy to jump from one project to the other.
- Didn't notice that live updating for completion suggester was removed from Elasticsearch 2
- Haven't done much analysis beyond a brief high level on the swap2and3 test *
- Some issues with the data collected for the textcat test. We were able to work through it, but could use better processes to validate before shipping+1
What else should be noted?
edit- Many issues with WDQS, we did a postmortem (Stas & Guillaume - Thanks for taking the time!). We have a few actions items on that
- Discernatron functionality is good; "usability" is less good (soooo many results)—overall holds much promise ***
- As we look toward Q1 how and where can the community liaison help? - selfish prod to think of me. :) ****
Discussion:
edit- As we look toward Q1 how and where can the community liaison help? - selfish prod to think of me. :) ****
- Draft goals: https://www.mediawiki.org/wiki/Wikimedia_Engineering/2016-17_Q1_Goals#Discovery
- DG: For some internal tech issues like changing elastic settings, an announcement should be enough. For something like interwiki search, more communication makes sense. We can sync up offline about that.
- EB: For the new search, it would be backward compatible so users wouldn't see a change at rollout, but it would open up new features in the future.
- GL: Do you think you get enough information from tech folks like me?
- CK: I get a lot of info from the weekly status reports. I'll send out a link to a video of my presentation of "what we do as CL's" (and don't)
- (I'll send this out to the internal discovery list but if you're here now): https://www.mediawiki.org/wiki/EMWCon_Spring_2016/Technical_Collaboration_with_the_Wikimedia_Foundation
- CK: I get a lot of info from the weekly status reports. I'll send out a link to a video of my presentation of "what we do as CL's" (and don't)
- Draft goals: https://www.mediawiki.org/wiki/Wikimedia_Engineering/2016-17_Q1_Goals#Discovery
- Discernatron functionality is good; "usability" is less good (soooo many results)—overall holds much promise ***
- DG: Combination of being busy and lack of design support?
- EB: I have been busy.
- DG: Maybe host a lunch in SF where people can eat if they help with Discernatron?
- We can't properly test major production changes to search infrastructure because we can't duplicate LVS ***
- EB: We have discussed outside this meeting; things are moving; just an annoyance
- GL: Not a major problem. It's not too bad testing in production.
Action Items
edit- Chris: post the link to his "what technical collaboration team does" presentation
- Chris: Chris and Erik should talk about implications of interwiki search indices
- Trey & Deb: Chris needs to be aware of the ? at the end of queries
- Erik: Figure out a plan to reliably monitor github (David and Guillaume have started to watch it)
Retro of retro:
edit- needs to be longer than 1/2 an hour *****