[
https://issues.apache.org/jira/browse/SOLR-18327?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18101281#comment-18101281
]
ASF subversion and git services commented on SOLR-18327:
--------------------------------------------------------
Commit 5e58e05f0f34fe2d1c878fd78a3c556297359179 in solr's branch
refs/heads/main from Jan Høydahl
[ https://gitbox.apache.org/repos/asf?p=solr.git;h=5e58e05f0f3 ]
SOLR-18327: Balance replicas API does not balance duplicate replicas on same
node (#2503)
> Balance replicas API does not balance duplicate replicas on same node
> ---------------------------------------------------------------------
>
> Key: SOLR-18327
> URL: https://issues.apache.org/jira/browse/SOLR-18327
> Project: Solr
> Issue Type: Bug
> Reporter: Jan Høydahl
> Assignee: Jan Høydahl
> Priority: Critical
> Labels: pull-request-available
> Time Spent: 50m
> Remaining Estimate: 0h
>
> The BALANCE_REPLICAS API (OrderedNodePlacementPlugin /
> AffinityPlacementFactory) does not correct the situation where a node hosts
> more than one replica of the same shard. Given a cluster where such
> duplicates exist (e.g. created via ADDREPLICA with an explicit node, or
> restore), a balance request computes an empty plan even though the cluster
> violates the "one replica per shard per node" rule.
> *Reproduction:*
> 4 nodes, two collections, each with one shard of 4 NRT replicas placed
> pairwise on two nodes (collection "a": 2 replicas on node 0, 2 on node 2;
> collection "b": 2 on node 1, 2 on node 3). computeBalancing() returns zero
> movements. See test testBalancingAvoidMultiReplicaOnNode in PR #2503 — fails
> deterministically with "expected 4 computed 0" (verified with
> -Dtests.iters=10).
> Fix in https://github.com/apache/solr/pull/2503
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
