sodonnel commented on code in PR #4061:
URL: https://github.com/apache/ozone/pull/4061#discussion_r1045990695
##########
hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/SCMCommonPlacementPolicy.java:
##########
@@ -480,8 +483,21 @@ public Set<ContainerReplica>
replicasToCopyToFixMisreplication(
requiredNumberOfPlacementGroups -= 1;
if (numberOfReplicasToBeCopied > 0) {
List<ContainerReplica> replicasToBeCopied = replicaList.stream()
+ .filter(replicas::get)
.limit(numberOfReplicasToBeCopied)
.collect(Collectors.toList());
+ if (numberOfReplicasToBeCopied > replicasToBeCopied.size()) {
+ Node rack = this.getPlacementGroup(
+ replicaList.stream().findAny()
+ .map(ContainerReplica::getDatanodeDetails)
+ .orElse(null));
+ throw new IOException(String.format(
Review Comment:
I am not sure if we should totally fail here. We might be able to improve
things somewhat, eg if we ideally had 3 replicas needing a copy, 2 on the first
rack and 1 on the second.
If, when we process the first rack, we can only find 1 or 0 to copy instead
of 2, and then if we were to process the second rack we might have been able to
find 1 to copy there. So we could have improved the situation by 1 or 2 out of
3.
It would probably be better to log a warning here and continue rather than
throwing an exeception.
--
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]
