Date
Meeting link: https://meet.google.com/bmc-uvwr-omd
Time: 9 am pacific time
Links: TimeAndDate.com
Attendees
- Andrey Ukhanov
- Christopher Ball
- Christos Malliaridis
- David Smiley
- Eric Pugh
- Hoss
- Jason Gerlowski
- Kevin Liang
- Matt Biscocho
- (Sorry for those I missed - please add yourself if you were able to make it!)
Discussion items
| Time | Item | Who | Notes |
|---|---|---|---|
| 10 | Update on Solr CLI work. | Eric | |
| 15m | Performance Testing Solr (between version) | Kevin Liang | |
| 15m | Admin UI | Christos Malliaridis |
Summary
Eric started us off discussing his overall progress on the SolrCLI modernization. Big takeaways are that he's nearing the home stretch of the effort, and that there are a few outstanding PRs that could use some review/attention, including PR #2580, PR #2577, and PR #2479 (which adds a new "stream" SolrCLI tool to expose streaming expressions with a bit more syntactic sugar than is currently provided by the 'bin/solr api' command). A few folks stuck around at the end of the call to review 2577 "live".
From there, Kevin Liang started a discussion around Solr Performance Testing. It seems some deployments where Kevin works have seen performance regressions (-10%) in certain types of queries after upgrading to Solr 9.x. There was broad agreement that the entire community would benefit greatly if there was ongoing performance testing that (1) could be run on discrete commits, and (2) that could be used to show longitudinal changes in performance. There was a good bit of discussion on the partial solutions that already exist including Solr's JMH "benchmark" module, the external solr-bench project, and Mike McCandless' lucene-bench site. And of course on the pros, cons, and limitations of each approach, and on the difficulty of picking representative data and query sets. Overall the group was very encouraging on any effort in this area, and several folks mentioned "progress over perfection".
Christos closed out the meeting by summarizing some of the recent mailing list discussions around the Admin UI revamp that he's interested in helping with. Big topics of conversation included the relationship between Solr and its Admin UI(s), the need to make progress in small slices and not aim for parity with the existing UI, and more technical concerns like specific architecture and framework tradeoffs. No concrete takeaways from this discussion, but a lot of folks got to meet Christos and ask questions about the initiative.