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

SOLR-11560: Specifying the replicationFactor parameter while restoring a collection would lead to extra tlog 

and pull replicas being created
This commit is contained in:
Varun Thacker 2017-10-31 13:32:19 -05:00
parent bc46de3b2a
commit 07652a5289
3 changed files with 27 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
solr/CHANGES.txt
View File

@ -95,6 +95,9 @@ Bug Fixes
* SOLR-9440: The ZkStateReader.removeCollectionStateWatcher method can cache a DocCollection reference and * SOLR-9440: The ZkStateReader.removeCollectionStateWatcher method can cache a DocCollection reference and
never update it causing stale state to be returned in ClusterState. (shalin) never update it causing stale state to be returned in ClusterState. (shalin)
* SOLR-11560: Specifying the replicationFactor parameter while restoring a collection would lead to extra tlog
and pull replicas being created (Peter Szantai-Kis, Varun Thacker)
Optimizations Optimizations
---------------------- ----------------------
* SOLR-11285: Refactor autoscaling framework to avoid direct references to Zookeeper and Solr * SOLR-11285: Refactor autoscaling framework to avoid direct references to Zookeeper and Solr

6
solr/core/src/java/org/apache/solr/cloud/RestoreCmd.java
View File

@ -350,8 +350,8 @@ public class RestoreCmd implements OverseerCollectionMessageHandler.Cmd {
} }
} }
private int getInt(ZkNodeProps message, String propertyName, Integer default1, int default2) { private int getInt(ZkNodeProps message, String propertyName, Integer count, int defaultValue) {
Integer value = message.getInt(REPLICATION_FACTOR, default1); Integer value = message.getInt(propertyName, count);
return value!=null?value:default2; return value!=null ? value:defaultValue;
} }
} }

24
solr/core/src/test/org/apache/solr/cloud/AbstractCloudBackupRestoreTestCase.java
View File

@ -57,6 +57,10 @@ public abstract class AbstractCloudBackupRestoreTestCase extends SolrCloudTestCa
protected static final int NUM_SHARDS = 2;//granted we sometimes shard split to get more protected static final int NUM_SHARDS = 2;//granted we sometimes shard split to get more
int replFactor;
int numTlogReplicas;
int numPullReplicas;
private static long docsSeed; // see indexDocs() private static long docsSeed; // see indexDocs()
@BeforeClass @BeforeClass
@ -84,9 +88,9 @@ public abstract class AbstractCloudBackupRestoreTestCase extends SolrCloudTestCa
public void test() throws Exception { public void test() throws Exception {
boolean isImplicit = random().nextBoolean(); boolean isImplicit = random().nextBoolean();
boolean doSplitShardOperation = !isImplicit && random().nextBoolean(); boolean doSplitShardOperation = !isImplicit && random().nextBoolean();
int replFactor = TestUtil.nextInt(random(), 1, 2); replFactor = TestUtil.nextInt(random(), 1, 2);
int numTlogReplicas = TestUtil.nextInt(random(), 0, 1); numTlogReplicas = TestUtil.nextInt(random(), 0, 1);
int numPullReplicas = TestUtil.nextInt(random(), 0, 1); numPullReplicas = TestUtil.nextInt(random(), 0, 1);
CollectionAdminRequest.Create create = isImplicit ? CollectionAdminRequest.Create create = isImplicit ?
// NOTE: use shard list with same # of shards as NUM_SHARDS; we assume this later // NOTE: use shard list with same # of shards as NUM_SHARDS; we assume this later
@ -241,6 +245,12 @@ public abstract class AbstractCloudBackupRestoreTestCase extends SolrCloudTestCa
CollectionAdminRequest.Restore restore = CollectionAdminRequest.restoreCollection(restoreCollectionName, backupName) CollectionAdminRequest.Restore restore = CollectionAdminRequest.restoreCollection(restoreCollectionName, backupName)
.setLocation(backupLocation).setRepositoryName(getBackupRepoName()); .setLocation(backupLocation).setRepositoryName(getBackupRepoName());
//explicitly specify the replicationFactor/pullReplicas/nrtReplicas/tlogReplicas .
//Value is still the same as the original. maybe test with different values that the original for better test coverage
if (random().nextBoolean()) {
restore.setReplicationFactor(replFactor);
}
if (backupCollection.getReplicas().size() > cluster.getJettySolrRunners().size()) { if (backupCollection.getReplicas().size() > cluster.getJettySolrRunners().size()) {
// may need to increase maxShardsPerNode (e.g. if it was shard split, then now we need more) // may need to increase maxShardsPerNode (e.g. if it was shard split, then now we need more)
restore.setMaxShardsPerNode((int)Math.ceil(backupCollection.getReplicas().size()/cluster.getJettySolrRunners().size())); restore.setMaxShardsPerNode((int)Math.ceil(backupCollection.getReplicas().size()/cluster.getJettySolrRunners().size()));
@ -306,6 +316,14 @@ public abstract class AbstractCloudBackupRestoreTestCase extends SolrCloudTestCa
v <= restoreCollection.getMaxShardsPerNode()); v <= restoreCollection.getMaxShardsPerNode());
}); });
assertEquals("Different count of nrtReplicas. Backup collection state=" + backupCollection + "\nRestore " +
"collection state=" + restoreCollection, replFactor, restoreCollection.getNumNrtReplicas().intValue());
assertEquals("Different count of pullReplicas. Backup collection state=" + backupCollection + "\nRestore" +
" collection state=" + restoreCollection, numPullReplicas, restoreCollection.getNumPullReplicas().intValue());
assertEquals("Different count of TlogReplica. Backup collection state=" + backupCollection + "\nRestore" +
" collection state=" + restoreCollection, numTlogReplicas, restoreCollection.getNumTlogReplicas().intValue());
// assert added core properties: // assert added core properties:
// DWS: did via manual inspection. // DWS: did via manual inspection.
// TODO Find the applicable core.properties on the file system but how? // TODO Find the applicable core.properties on the file system but how?