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

eliminate CPU hogging spin-loops in OverseerTest 

this test already uses waitForState (frequently via verifyReplicaStatus) so there is no reason to include CPU/network/ZK intensive infinite loop checks looking for udpated cluster state
This commit is contained in:
Chris Hostetter 2019-06-17 15:42:07 -07:00
parent 7eb8703df6
commit 3030ea9d94
1 changed files with 7 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
solr/core/src/test/org/apache/solr/cloud/OverseerTest.java
View File

@ -670,42 +670,33 @@ public class OverseerTest extends SolrTestCaseJ4 {
waitForCollections(reader, COLLECTION); waitForCollections(reader, COLLECTION);
verifyReplicaStatus(reader, COLLECTION, "shard1", "core_node1", Replica.State.RECOVERING); verifyReplicaStatus(reader, COLLECTION, "shard1", "core_node1", Replica.State.RECOVERING);
int version = getClusterStateVersion(zkClient);
mockController.publishState(COLLECTION, core, core_node, "shard1", Replica.State.ACTIVE, mockController.publishState(COLLECTION, core, core_node, "shard1", Replica.State.ACTIVE,
numShards, true, overseers.get(0)); numShards, true, overseers.get(0));
while (version == getClusterStateVersion(zkClient));
verifyReplicaStatus(reader, COLLECTION, "shard1", "core_node1", Replica.State.ACTIVE); verifyReplicaStatus(reader, COLLECTION, "shard1", "core_node1", Replica.State.ACTIVE);
version = getClusterStateVersion(zkClient);
mockController.publishState(COLLECTION, core, core_node, "shard1", mockController.publishState(COLLECTION, core, core_node, "shard1",
Replica.State.RECOVERING, numShards, true, overseers.get(0)); Replica.State.RECOVERING, numShards, true, overseers.get(0));
overseerClient.close(); overseerClient.close();
version = getClusterStateVersion(zkClient);
overseerClient = electNewOverseer(server.getZkAddress()); overseerClient = electNewOverseer(server.getZkAddress());
while (version == getClusterStateVersion(zkClient));
verifyReplicaStatus(reader, COLLECTION, "shard1", "core_node1", Replica.State.RECOVERING); verifyReplicaStatus(reader, COLLECTION, "shard1", "core_node1", Replica.State.RECOVERING);
assertEquals("Live nodes count does not match", 1, reader assertEquals("Live nodes count does not match", 1, reader
.getClusterState().getLiveNodes().size()); .getClusterState().getLiveNodes().size());
assertEquals(shard+" replica count does not match", 1, reader.getClusterState() assertEquals(shard+" replica count does not match", 1, reader.getClusterState()
.getCollection(COLLECTION).getSlice(shard).getReplicasMap().size()); .getCollection(COLLECTION).getSlice(shard).getReplicasMap().size());
version = getClusterStateVersion(zkClient);
mockController.publishState(COLLECTION, core, core_node, "shard1", null, numShards, true, overseers.get(1)); mockController.publishState(COLLECTION, core, core_node, "shard1", null, numShards, true, overseers.get(1));
while (version == getClusterStateVersion(zkClient));
assertTrue(COLLECTION +" should remain after removal of the last core", // as of SOLR-5209 core removal does not cascade to remove the slice and collection
reader.getClusterState().hasCollection(COLLECTION));
reader.waitForState(COLLECTION, 5000, reader.waitForState(COLLECTION, 5000,
TimeUnit.MILLISECONDS, (liveNodes, collectionState) -> collectionState != null && collectionState.getReplica(core_node) == null); TimeUnit.MILLISECONDS, (liveNodes, collectionState) -> collectionState != null && collectionState.getReplica(core_node) == null);
reader.forceUpdateCollection(COLLECTION);
// as of SOLR-5209 core removal does not cascade to remove the slice and collection
assertTrue(COLLECTION +" should remain after removal of the last core",
reader.getClusterState().hasCollection(COLLECTION));
assertTrue(core_node+" should be gone after publishing the null state", assertTrue(core_node+" should be gone after publishing the null state",
null == reader.getClusterState().getCollection(COLLECTION).getReplica(core_node)); null == reader.getClusterState().getCollection(COLLECTION).getReplica(core_node));
} finally { } finally {
@ -1055,18 +1046,11 @@ public class OverseerTest extends SolrTestCaseJ4 {
mockController.close(); mockController.close();
int version = getClusterStateVersion(zkClient);
mockController = new MockZKController(server.getZkAddress(), "node1", overseers); mockController = new MockZKController(server.getZkAddress(), "node1", overseers);
mockController.publishState(COLLECTION, "core1", "core_node1","shard1", Replica.State.RECOVERING, 1, true, overseers.get(0)); mockController.publishState(COLLECTION, "core1", "core_node1","shard1", Replica.State.RECOVERING, 1, true, overseers.get(0));
try { reader.forceUpdateCollection(COLLECTION);
reader.waitForState(COLLECTION, 5, TimeUnit.SECONDS, (liveNodes, collectionState) -> version == zkController
.getZkStateReader().getClusterState().getZkClusterStateVersion());
} catch (TimeoutException e) {
// okay
}
ClusterState state = reader.getClusterState(); ClusterState state = reader.getClusterState();
int numFound = 0; int numFound = 0;
@ -1397,12 +1381,6 @@ public class OverseerTest extends SolrTestCaseJ4 {
} }
} }
private int getClusterStateVersion(SolrZkClient controllerClient)
throws KeeperException, InterruptedException {
return controllerClient.exists(ZkStateReader.CLUSTER_STATE, null, false).getVersion();
}
private SolrZkClient electNewOverseer(String address) private SolrZkClient electNewOverseer(String address)
throws InterruptedException, TimeoutException, IOException, throws InterruptedException, TimeoutException, IOException,
KeeperException, ParserConfigurationException, SAXException, NoSuchFieldException, SecurityException { KeeperException, ParserConfigurationException, SAXException, NoSuchFieldException, SecurityException {