Apache9 commented on a change in pull request #1311: HBASE-23984 [Flakey Tests]
TestMasterAbortAndRSGotKilled fails in tea…
URL: https://github.com/apache/hbase/pull/1311#discussion_r395406458
##########
File path:
hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/AssignRegionHandler.java
##########
@@ -132,11 +131,11 @@ public void process() throws IOException {
// opening can not be interrupted by a close request any more.
region = HRegion.openHRegion(regionInfo, htd, rs.getWAL(regionInfo),
rs.getConfiguration(),
rs, null);
+ rs.postOpenDeployTasks(new PostOpenDeployContext(region, openProcId,
masterSystemTime));
Review comment:
No...
IIRC, the design here is that, postOpenDeployTasks is the PONR, if we arrive
here, then we can not revert back, the only way to address the exception is to
abort the region server.
The fact is that, if we haven't told master anything, it is fine for us to
close the region and tell master the failure, but once we have already called
master with the succeeded message, even if the rpc call fails, we do not know
whether the other side(the master) has received and processed the request
already, so the only way is to retry for ever, and if this can not be done, the
only way is to abort ourselves...
----------------------------------------------------------------
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
