jtuglu1 commented on PR #19644: URL: https://github.com/apache/druid/pull/19644#issuecomment-4857864216
> > My proposed solution is allow follower Coordinator replicas to also fill their cache and rely on a delta sync upon becoming the leader. The delta sync is a much cheaper operation and allows the Coordinator to come up faster with a healthy cache. While this faster recovery/higher availability is at the expense of higher DB load, this is particularly useful for situations like rolling deployments where you may only have > 1 coordinators running for a short period (then back to 1). > > @jtuglu1 , this is already the case. Follower coordinator/overlords already perform delta syncs in the background. Only the first sync (that happens upon startup) is a full sync for both Leader and Follower. This can be verified from logs as well as cache metrics. The only requirement is that upon becoming Leader, the service needs to finish another delta sync (or full sync only if no sync has happened yet) before it starts to use the cache. > > Please let me know if you have noticed otherwise. 👍 I think I'll need to look into why a delta sync is taking so long then. Likely -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
