Date
Attendees
- Jason Gerlowski
- Jan Hoydahl
- Hoss
- Christopher Ball
- Eric Pugh
- Mikhail Khludnev
- Kevin Risden
- Ilan Ginzburg
- (I added a few of the names above, but have missed many folks - please add others who were in attendance)
Google Meetup link:
https://meet.google.com/cbf-hsns-mxh
Discussion items
| Time | Topic | People | Notes |
|---|---|---|---|
| Make Solr Cloud opt out instead of opt in? | Jason G | ||
| Starting an "Official" Solr Blog | Jason G | ||
| Discuss Plan for Moving to Newer Versions of Java and Lucene 10 | Eric Pugh | ||
| Demo of user managed clusters, with a twist | Eric Pugh | ||
| Applying fix for block join tests. Please chime in if you aware of block join query/logic test, which is out of scope yet | Mikhail Khludnev |
Notes and Action items
We had ~20 folks show up for our meeting on the 23rd. There were no explicit action items, but there was lively discussion on the topics above. See below for a brief summary:
Jason started discussion by raising a question that's come up a few times in the past: "SolrCloud has been around for somewhere near 10 years, it's the preeminent way our users deploy Solr, and many features are available only when running SolrCloud - is it worth considering making it the default mode of operation, and what should that look like?" No clear outcome was reached, but a lot of good points were raised, including: questions about the production-readiness of embedded ZK for single-node deployments, how to clearly communicate to users when they're running embedded ZK, and the relationship between switching the default mode and the work outlined in SIP-14. Discussion also at times strayed towards the standalone vs SolrCloud feature gap. Ilan highlighted a few features available in standalone but not SolrCloud, most notably "transient cores".
Tangentially related to the SolrCloud/standalone discussion, Eric walked the group through a small test which show how "repeater-style" replication surprisingly actually works in SolrCloud mode. This was one feature that was thought to be a gap between the two deployment modes, but surprisingly works in SolrCloud - albeit in a kindof hacky manner. Various Jurassic Park-style comments ensued.
Jason gave a short (~5m) demo of a PR he's opened for SOLR-16833, which considers adding a very basic blog structure to the Solr website.
Eric initiated some discussion around whether Solr 10 should bump the minimum required Java version. It looks like Lucene will do something similar in an upcoming release, so it likely makes sense to do the same. Hoss pointed out that before doing this it might make sense to see if we can modify our build to support a different (wider?) version range for SolrJ than for Solr's server-side code, due to how high-impact a Java version change would be on SolrJ users.
Lastly, Mikhail used a few minutes to raise visibility on a fix he found for a block-join related test bug.