FWIW I do think it's weird for a replica to tell the leader to do something (anything) during recovery. It'd be great to eliminate this somehow. If all tests pass without it; I'm in favor of making it opt-in internal feature-flag for eventual removal.
On Fri, Oct 17, 2025 at 10:37 AM Kevin Liang (BLOOMBERG/ 919 3RD A) < [email protected]> wrote: > Thanks for that source/tip. Will dig further there. The "it seems like > forcing a commit on the leader might never be necessary" comment kind of > reinforces where my thinking is going > > From: [email protected] At: 10/16/25 17:17:37 UTC-4:00To: > [email protected] > Subject: Re: Why recovery requires commit? > > Not a direct answer to your question but I find it helps tremendously to > look at source control history. You can get the history for a method (cool > feature of Intellij), scan through, and voila -- > https://issues.apache.org/jira/browse/SOLR-16689 with great info > > IMO it doesn't hurt to ask questions on old JIRAs/PRs > > On Thu, Oct 16, 2025 at 4:38 PM Kevin Liang (BLOOMBERG/ 919 3RD A) < > [email protected]> wrote: > > > I've been looking at the code in RecoveryStrategy for a while and I can't > > wrap my head around why a TLOG node replicating from leader would need to > > send a commit back to the leader and wait for a new searcher on the > leader? > > > > https://github.com/apache/solr/blob/06a3b5e77e94771ee35c407cc90be0ce46d7a748/sol > r/core/src/java/org/apache/solr/cloud/RecoveryStrategy.java#L235 > <https://github.com/apache/solr/blob/06a3b5e77e94771ee35c407cc90be0ce46d7a748/solr/core/src/java/org/apache/solr/cloud/RecoveryStrategy.java#L235> > > > > Can someone help explain this? Thanks > > >
