[HBASE-9286] [0.94] ageOfLastShippedOp replication metric doesn't update if the slave regionserver is stalled

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.94.12
• Hadoop Flags: Reviewed
• Assignee: Alex Newman
• Reporter: Alex Newman
• Created: 21/Aug/13 04:42
• Updated: 09/May/14 06:34
• Resolved: 06/Sep/13 17:59

Description

In replicationmanager
HRegionInterface rrs = getRS();
rrs.replicateLogEntries(Arrays.copyOf(this.entriesArray, currentNbEntries));
....
this.metrics.setAgeOfLastShippedOp(
this.entriesArray[currentNbEntries-1].getKey().getWriteTime());
break;

which makes sense, but is wrong. The problem is that rrs.replicateLogEntries will block for a very long time if the slave server is suspended or unavailable but not down.

However this is easy to fix. We just need to call refreshAgeOfLastShippedOp();
on a regular basis, in a different thread. I've attached a patch which fixed this for cdh4. I can make one for trunk and the like as well if you need me to do but it's a small change.

Issue Links

• relates to: HBASE-11143 Improve replication metrics