mcvsubbu commented on issue #11314: URL: https://github.com/apache/pinot/issues/11314#issuecomment-1675585363
Correct, I realized after posting that you may have had a problem with one of the replicas. So, the thought was as follows: In general, if there are consumption issues from a stream, it is likely that all replicas face issues communicating with the stream. So the likely case is that all replicas go down eventually, perhaps within short time of each other. Further, if only a subset of replicas go down, there is still at least one replica alive to complete the segment. Once the segment is complete, the transition from OFFLINE to ONLINE would arrive, and everything would return back to normal.The thought here is that the other replicas are enough to serve queries. After all , only one segment will be routed to some other replica. The server in question can continue to serve other segments of the table. Lastly, let us say we try to modify the state back to CONSUMING in idealstate. We will need to worry about race conditions where the segment may be in completing stage. Also, the host may have to feverishly catch up to the data only to find that some other replica is at the point of completing the segment. For these reasons, we decided to fix the situation only if all replicas are OFFLINE. -- 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: commits-unsubscr...@pinot.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org For additional commands, e-mail: commits-h...@pinot.apache.org
