Wikimedia Search Team/Meeting minutes/2015-05-18

2015-05-18 edit

Nik

Did rolling restarts of elastic
Worked on query parser (has 50+ tests mimicking our integration tests)
- Maybe 4 person-days away from being able _start_ replace the hated query_string
Trying to figure out when upstream releases Elasticsearch 1.6 - we wants it so badly!
Fixing some issues Elasticsearch has filed against features I wrote in Elasticsearch
Can now remove a bunch of dynamic groovy which will make ops happy.
- We’ll need one more rolling restart after we merge and deploy some Cirrus changes to make turning it off really really real.

Stas

Released 0.0.1 to maven central
Need to set up semi-public endpoint for hackathon and others
Should we use blazegraph demo or create a front-end
- Standard front end has a lot of features we intentionally don’t support (e.g. readonly)
Needs help. Can’t be alone. Especially around all this Java.

James

Has been working with S Page on API docs (focus on simplicity)
- Want to create a portal, maybe including snippets, demos, etc.
- Created quickie front-end for current WDQ front end
  - Could be a good starting point for WDQS front end
Meetings, reviews. Later in the week, continue on docs. Will create phab tasks

Max

Yuri

Gave presentation of how to use new graph feature in wiki
- Hungarian already uses over 200 graphs
Working w/Max on mapping features
Creating map presentation for hackathon (will also give graph presentation)
Plan to create vector graphics demo
Hope to create “wow” demo, maybe something with overlays

Erik

Wildcards now enabled in cirrus for other wikis, without performance problems
Improving multitasking in hhvm. Forking scripts causes some issues.
Added tests to some of Nik’s code
Will be making search index updates pause-able
- Nik thinks they might be related to multi datacenter work
Will be somewhat unavailable intermittent over the next couple weeks

Oliver

Still working on data visualisation dashboard
- Still from static snapshot, until things settle out
“Find out how API users are using search”
Working on non-search project
Wrote a C++ tool to geolocate readers/editors - https://github.com/Ironholds/rgeolocate
Will continue on API and non-search work, as well as hiring

Dan

Wes (Didn’t attend, but sent these notes)

primarily planning for hackathon, wikidata conversations there and planning offsite
Met last week with Google discussing Wikidata Query Service
- will follow up with Dan regarding that, and planning.

Should James learn more about metrics/measurement?
- Hiring is going slower than hoped (opening req took a while too)
- Nik thinks we should build internal expertise rather than hiring
- James is already sort of on that path; happy to do it; overburdened?
- Oliver may have some sources of potential talent (less engineer-y)
Are we hiring a front-end developer?
- We hope to open a req “soon”
Who can look at dynamic script removal?
- James volunteered
Who can help Stas with Java stuff
- James is interviewing a candidate w/Java experience
Proposal to keep feature-specific projects around
- Yuri had some concerns, but they were resolved in offline discussions w/Kevin
Pushback on our team’s wiki page name (“Search and Discovery”)
- Should it have “Team” at the end? “Department”
- Is it reasonable that “Search” redirects to our team page?
- “Editing” is in a similar position
Initial WDQS release (0.0.1) Status https://phabricator.wikimedia.org/T96579