[HBASE-8910] HMaster.abortNow shouldn't try to become a master again if it was stopped

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.98.0, 0.95.2, 0.94.10
• Hadoop Flags: Reviewed
• Assignee: Jean-Daniel Cryans
• Reporter: Jean-Daniel Cryans
• Created: 10/Jul/13 00:37
• Updated: 24/Jul/13 19:08
• Resolved: 10/Jul/13 17:35

Description

Here's a case where TestReplicationDisableInactivePeer fails while re-starting the second master:

http://54.241.6.143/job/HBase-0.95-Hadoop-2/574/org.apache.hbase$hbase-server/testReport/junit/org.apache.hadoop.hbase.replication/TestReplicationDisableInactivePeer/testDisableInactivePeer/

The reason is that when we first shutdown the master, it comes back to life thinking it just lost its session:

2013-07-07 04:27:03,989 FATAL [pool-1-thread-1-EventThread] master.HMaster(2062): Master server abort: loaded coprocessors are: [org.apache.hadoop.hbase.coprocessor.MultiRowMutationEndpoint]
2013-07-07 04:27:03,989 INFO  [pool-1-thread-1-EventThread] master.HMaster(2155): Primary Master trying to recover from ZooKeeper session expiry.

And after that it tries to assign .META. fails since the RS are down.

One way I think we can prevent this is by skipping recovering the session if we are stopping.