Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

SOLR-5904: ElectionContext can cancel an election when it should not if there was an exception while trying to register as the leader. 

git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/trunk@1587702 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Mark Robert Miller 2014-04-15 19:28:45 +00:00
parent 3dc2074d1b
commit c7c9fc35a1
2 changed files with 11 additions and 40 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
solr/CHANGES.txt
View File

@ -94,6 +94,13 @@ Upgrading from Solr 4.8
Detailed Change List
----------------------
Bug Fixes
----------------------
* SOLR-5904: ElectionContext can cancel an election when it should not if there
was an exception while trying to register as the leader.
(Mark Miller, Alan Woodward)
Other Changes
---------------------

44
solr/core/src/java/org/apache/solr/cloud/ElectionContext.java
View File

@ -286,63 +286,27 @@ final class ShardLeaderElectionContext extends ShardLeaderElectionContextBase {
+ ZkCoreNodeProps.getCoreUrl(leaderProps) + " " + shardId);
core.getCoreDescriptor().getCloudDescriptor().setLeader(true);
}
boolean success = false;
try {
super.runLeaderProcess(weAreReplacement, 0);
success = true;
} catch (Exception e) {
SolrException.log(log, "There was a problem trying to register as the leader", e);
try (SolrCore core = cc.getCore(coreName)) {
if (core == null) {
throw new SolrException(ErrorCode.SERVER_ERROR,
"Fatal Error, SolrCore not found:" + coreName + " in "
+ cc.getCoreNames());
log.debug("SolrCore not found:" + coreName + " in " + cc.getCoreNames());
return;
}
core.getCoreDescriptor().getCloudDescriptor().setLeader(false);
// we could not publish ourselves as leader - rejoin election
// we could not publish ourselves as leader - try and rejoin election
rejoinLeaderElection(leaderSeqPath, core);
} finally {
if (!success)
cancelElection();
}
}
}
private boolean areAnyOtherReplicasActive(ZkController zkController,
ZkNodeProps leaderProps, String collection, String shardId) {
ClusterState clusterState = zkController.getZkStateReader()
.getClusterState();
Map<String,Slice> slices = clusterState.getSlicesMap(collection);
Slice slice = slices.get(shardId);
if (!slice.getState().equals(Slice.ACTIVE)) {
//Return false if the Slice is not active yet.
return false;
}
Map<String,Replica> replicasMap = slice.getReplicasMap();
for (Map.Entry<String,Replica> shard : replicasMap.entrySet()) {
String state = shard.getValue().getStr(ZkStateReader.STATE_PROP);
// System.out.println("state:"
// + state
// + shard.getValue().get(ZkStateReader.NODE_NAME_PROP)
// + " live: "
// + clusterState.liveNodesContain(shard.getValue().get(
// ZkStateReader.NODE_NAME_PROP)));
if (state.equals(ZkStateReader.ACTIVE)
&& clusterState.liveNodesContain(shard.getValue().getStr(
ZkStateReader.NODE_NAME_PROP))
&& !new ZkCoreNodeProps(shard.getValue()).getCoreUrl().equals(
new ZkCoreNodeProps(leaderProps).getCoreUrl())) {
return true;
}
}
return false;
}
private void waitForReplicasToComeUp(boolean weAreReplacement, int timeoutms) throws InterruptedException {
long timeoutAt = System.nanoTime() + TimeUnit.NANOSECONDS.convert(timeoutms, TimeUnit.MILLISECONDS);