huaxiangsun commented on issue #1514: HBASE-24183 [flakey test] replication.TestAddToSerialReplicationPeer URL: https://github.com/apache/hbase/pull/1514#issuecomment-613757958 There are two flakies. testAddToSerialPeer failure is that it just needs to make sure the source RS's inmemory map contains only the new wal file. (not the RS which region moves to) After that, there is still one failure which is common to testAddToSerialPeer and testChangeToSerial. If the old wal file before rollover is still in the inmemory map of replicateSourceManager, during peer disable/enable/config update, it could be still be replicated over from the begin to peer cluster. If that happens, the old wal entries and the new wal entries will be written to the same wal file (which results in out of order seq numbers). waitUntilReplicatedToTheCurrentWALFile() does not really guarantee that the inmemory map is forwarded to the new wal file, there is a small window that there is only one wal file in map which is the old wal file. Added a new check to make sure that inmemory map only has the new wal file. This could happen in the production cluster as well, however I do not think this is the purpose of these two test cases. @Apache9, please provide your input, thanks.
---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
