[
https://issues.apache.org/jira/browse/SOLR-9555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15969334#comment-15969334
]
Mark Miller commented on SOLR-9555:
-----------------------------------
more precisely, if the issue is as stated, it would seem the problem is that we
stop the replicas from being able to receive requests, then we index 2 docs,
then we kill the leader and close it's proxy. That is where there seems to be
room for a replica to recover from the leader. If we do
getProxyForReplica(leader).close() when we close the replica proxies, that
should prevent any replica from being able to recover from it even before we
kill the leader jetty. Doesn't seem like it would have any negative
consequences either on first look. It's not like the test expects a replica to
still be able to talk to the leader during this point, right?
{code}
// ok, now introduce a network partition between the leader and both
replicas
log.info("Closing proxies for the non-leader replicas...");
for (SocketProxy proxy : nonLeaderProxies)
proxy.close();
// indexing during a partition
log.info("Sending a doc during the network partition...");
sendDoc(2);
// Wait a little
Thread.sleep(2000);
// Kill the leader
log.info("Killing leader for shard1 of " + collectionName + " on node " +
leader.getNodeName() + "");
JettySolrRunner leaderJetty = getJettyOnPort(getReplicaPort(leader));
getProxyForReplica(leader).close();
leaderJetty.stop();
{code}
> Leader incorrectly publishes state for replica when it puts replica into LIR.
> -----------------------------------------------------------------------------
>
> Key: SOLR-9555
> URL: https://issues.apache.org/jira/browse/SOLR-9555
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Alan Woodward
> Attachments: SOLR-9555-WIP.patch
>
>
> See
> https://jenkins.thetaphi.de/job/Lucene-Solr-master-Linux/17888/consoleFull
> for an example
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
