sharmaar12 commented on PR #7149: URL: https://github.com/apache/hbase/pull/7149#issuecomment-3245828917
> So if active cluster splits a region, and daughters get compacted before refresh_meta is called, we may have the region removed by catalog janitor running in active cluster. In this case, read replica cluster would get FNFE when trying to read the removed region. I guess that's documented? We will be documenting following: - The Switch Over from active to read should strictly follow: Make Active cluster Read-only > Flush on Active > Refresh_Meta on read replica > refresh_hfiles on read replica - There may be inconsistencies between active and read-replica cluster until the switch over completely happens As per the discussion over the call, we may need to see the feasibility of disabling the catalog janitor or making it aware of read-replica cluster is there, to avoid removal of any Hfiles. I will do bit of investigation and lets discuss this with wider audience in the next sync up. -- 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]
