siddhantsangwan commented on code in PR #4227:
URL: https://github.com/apache/ozone/pull/4227#discussion_r1102359872
##########
hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/container/replication/health/TestRatisReplicationCheckHandler.java:
##########
@@ -363,6 +363,28 @@ public void
testHandlerReturnsFalseWhenAllReplicasAreUnhealthy() {
Assert.assertFalse(result.isReplicatedOkAfterPending());
Assert.assertFalse(result.underReplicatedDueToDecommission());
+ Assert.assertFalse(healthCheck.handle(requestBuilder.build()));
+ Assert.assertEquals(0, repQueue.underReplicatedQueueSize());
+ Assert.assertEquals(0, repQueue.overReplicatedQueueSize());
+ Assert.assertEquals(0, report.getStat(
+ ReplicationManagerReport.HealthState.UNDER_REPLICATED));
Review Comment:
Handling all UNHEALTHY replicas works as it does in legacy. This is checked
and queued for over/under rep in `RatisUnhealthyReplicationCheckHandler` and
the required commands are created in `RatisOverReplicationHandler` and
`RatisUnderReplicationHandler`.
Point 5 is about situations where we have excess UNHEALTHY and mismatched
replicas. `RatisReplicationCheckHandler` won't queue for over rep as long as
there are some mismatched replicas. `RatisUnhealthyReplicationCheckHandler` is
however still queuing them - this is inconsistent as you pointed out in [the
comment
above](https://github.com/apache/ozone/pull/4227#discussion_r1099303927). I can
fix this by either making both checks more conservative (not queuing if there
are mismatched replicas) or by just checking if we have replication factor
number of matching replicas and then queuing.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
