use of org.elasticsearch.cluster.metadata.Metadata in project elasticsearch by elastic.
the class FailedShardsRoutingTests method testFailAllReplicasInitializingOnPrimaryFail.
public void testFailAllReplicasInitializingOnPrimaryFail() {
AllocationService allocation = createAllocationService(Settings.builder().build());
MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(2)).build();
RoutingTable routingTable = RoutingTable.builder().addAsNew(metaData.index("test")).build();
ClusterState clusterState = ClusterState.builder(CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(routingTable).build();
ShardId shardId = new ShardId(metaData.index("test").getIndex(), 0);
// add 4 nodes
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2")).add(newNode("node3")).add(newNode("node4"))).build();
clusterState = ClusterState.builder(clusterState).routingTable(allocation.reroute(clusterState, "reroute").routingTable()).build();
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().shardsWithState(UNASSIGNED).size(), equalTo(2));
// start primary shards
clusterState = allocation.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(2));
// start one replica so it can take over.
clusterState = allocation.applyStartedShards(clusterState, Collections.singletonList(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).get(0)));
assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(2));
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
ShardRouting startedReplica = clusterState.getRoutingNodes().activeReplica(shardId);
// fail the primary shard, check replicas get removed as well...
ShardRouting primaryShardToFail = clusterState.routingTable().index("test").shard(0).primaryShard();
ClusterState newState = allocation.applyFailedShard(clusterState, primaryShardToFail);
assertThat(newState, not(equalTo(clusterState)));
clusterState = newState;
// the primary gets allocated on another node, replicas are initializing
assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(2));
ShardRouting newPrimaryShard = clusterState.routingTable().index("test").shard(0).primaryShard();
assertThat(newPrimaryShard, not(equalTo(primaryShardToFail)));
assertThat(newPrimaryShard.allocationId(), equalTo(startedReplica.allocationId()));
}
use of org.elasticsearch.cluster.metadata.Metadata in project elasticsearch by elastic.
the class FailedShardsRoutingTests method testSingleShardMultipleAllocationFailures.
public void testSingleShardMultipleAllocationFailures() {
AllocationService strategy = createAllocationService(Settings.builder().put("cluster.routing.allocation.node_concurrent_recoveries", 10).put(ClusterRebalanceAllocationDecider.CLUSTER_ROUTING_ALLOCATION_ALLOW_REBALANCE_SETTING.getKey(), "always").build());
logger.info("Building initial routing table");
int numberOfReplicas = scaledRandomIntBetween(2, 10);
MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(numberOfReplicas)).build();
RoutingTable initialRoutingTable = RoutingTable.builder().addAsNew(metaData.index("test")).build();
ClusterState clusterState = ClusterState.builder(CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(initialRoutingTable).build();
logger.info("Adding {} nodes and performing rerouting", numberOfReplicas + 1);
DiscoveryNodes.Builder nodeBuilder = DiscoveryNodes.builder();
for (int i = 0; i < numberOfReplicas + 1; i++) {
nodeBuilder.add(newNode("node" + Integer.toString(i)));
}
clusterState = ClusterState.builder(clusterState).nodes(nodeBuilder).build();
while (!clusterState.routingTable().shardsWithState(UNASSIGNED).isEmpty()) {
// start all initializing
clusterState = strategy.applyStartedShards(clusterState, clusterState.routingTable().shardsWithState(INITIALIZING));
// and assign more unassigned
clusterState = strategy.reroute(clusterState, "reroute");
}
int shardsToFail = randomIntBetween(1, numberOfReplicas);
ArrayList<FailedShard> failedShards = new ArrayList<>();
RoutingNodes routingNodes = clusterState.getRoutingNodes();
Set<String> failedNodes = new HashSet<>();
Set<ShardRouting> shardRoutingsToFail = new HashSet<>();
for (int i = 0; i < shardsToFail; i++) {
String failedNode = "node" + Integer.toString(randomInt(numberOfReplicas));
logger.info("failing shard on node [{}]", failedNode);
ShardRouting shardToFail = routingNodes.node(failedNode).iterator().next();
if (shardRoutingsToFail.contains(shardToFail) == false) {
failedShards.add(new FailedShard(shardToFail, null, null));
failedNodes.add(failedNode);
shardRoutingsToFail.add(shardToFail);
}
}
clusterState = strategy.applyFailedShards(clusterState, failedShards);
routingNodes = clusterState.getRoutingNodes();
for (FailedShard failedShard : failedShards) {
if (routingNodes.getByAllocationId(failedShard.getRoutingEntry().shardId(), failedShard.getRoutingEntry().allocationId().getId()) != null) {
fail("shard " + failedShard + " was not failed");
}
}
for (String failedNode : failedNodes) {
if (!routingNodes.node(failedNode).isEmpty()) {
fail("shard was re-assigned to failed node " + failedNode);
}
}
}
use of org.elasticsearch.cluster.metadata.Metadata in project elasticsearch by elastic.
the class FailedShardsRoutingTests method testFailedShardPrimaryRelocatingToAndFrom.
public void testFailedShardPrimaryRelocatingToAndFrom() {
AllocationService allocation = createAllocationService(Settings.builder().put("cluster.routing.allocation.node_concurrent_recoveries", 10).put(ClusterRebalanceAllocationDecider.CLUSTER_ROUTING_ALLOCATION_ALLOW_REBALANCE_SETTING.getKey(), "always").build());
logger.info("--> building initial routing table");
MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(1)).build();
RoutingTable routingTable = RoutingTable.builder().addAsNew(metaData.index("test")).build();
ClusterState clusterState = ClusterState.builder(CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(routingTable).build();
logger.info("--> adding 2 nodes on same rack and do rerouting");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
clusterState = allocation.reroute(clusterState, "reroute");
// starting primaries
clusterState = allocation.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
// starting replicas
clusterState = allocation.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
logger.info("--> verifying all is allocated");
assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().node("node1").iterator().next().state(), equalTo(STARTED));
assertThat(clusterState.getRoutingNodes().node("node2").size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().node("node2").iterator().next().state(), equalTo(STARTED));
logger.info("--> adding additional node");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node3"))).build();
clusterState = allocation.reroute(clusterState, "reroute");
assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().node("node1").iterator().next().state(), equalTo(STARTED));
assertThat(clusterState.getRoutingNodes().node("node2").size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().node("node2").iterator().next().state(), equalTo(STARTED));
assertThat(clusterState.getRoutingNodes().node("node3").size(), equalTo(0));
String origPrimaryNodeId = clusterState.routingTable().index("test").shard(0).primaryShard().currentNodeId();
String origReplicaNodeId = clusterState.routingTable().index("test").shard(0).replicaShards().get(0).currentNodeId();
logger.info("--> moving primary shard to node3");
AllocationService.CommandsResult commandsResult = allocation.reroute(clusterState, new AllocationCommands(new MoveAllocationCommand("test", 0, clusterState.routingTable().index("test").shard(0).primaryShard().currentNodeId(), "node3")), false, false);
assertThat(commandsResult.getClusterState(), not(equalTo(clusterState)));
clusterState = commandsResult.getClusterState();
assertThat(clusterState.getRoutingNodes().node(origPrimaryNodeId).iterator().next().state(), equalTo(RELOCATING));
assertThat(clusterState.getRoutingNodes().node("node3").iterator().next().state(), equalTo(INITIALIZING));
logger.info("--> fail primary shard recovering instance on node3 being initialized");
clusterState = allocation.applyFailedShard(clusterState, clusterState.getRoutingNodes().node("node3").iterator().next());
assertThat(clusterState.getRoutingNodes().node(origPrimaryNodeId).iterator().next().state(), equalTo(STARTED));
assertThat(clusterState.getRoutingNodes().node("node3").size(), equalTo(0));
logger.info("--> moving primary shard to node3");
commandsResult = allocation.reroute(clusterState, new AllocationCommands(new MoveAllocationCommand("test", 0, clusterState.routingTable().index("test").shard(0).primaryShard().currentNodeId(), "node3")), false, false);
assertThat(commandsResult.getClusterState(), not(equalTo(clusterState)));
clusterState = commandsResult.getClusterState();
assertThat(clusterState.getRoutingNodes().node(origPrimaryNodeId).iterator().next().state(), equalTo(RELOCATING));
assertThat(clusterState.getRoutingNodes().node("node3").iterator().next().state(), equalTo(INITIALIZING));
logger.info("--> fail primary shard recovering instance on node1 being relocated");
clusterState = allocation.applyFailedShard(clusterState, clusterState.getRoutingNodes().node(origPrimaryNodeId).iterator().next());
// check promotion of replica to primary
assertThat(clusterState.getRoutingNodes().node(origReplicaNodeId).iterator().next().state(), equalTo(STARTED));
assertThat(clusterState.routingTable().index("test").shard(0).primaryShard().currentNodeId(), equalTo(origReplicaNodeId));
assertThat(clusterState.routingTable().index("test").shard(0).replicaShards().get(0).currentNodeId(), anyOf(equalTo(origPrimaryNodeId), equalTo("node3")));
}
use of org.elasticsearch.cluster.metadata.Metadata in project elasticsearch by elastic.
the class FailedShardsRoutingTests method testFailAllReplicasInitializingOnPrimaryFailWhileHavingAReplicaToElect.
public void testFailAllReplicasInitializingOnPrimaryFailWhileHavingAReplicaToElect() {
AllocationService allocation = createAllocationService(Settings.builder().build());
MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(2)).build();
RoutingTable routingTable = RoutingTable.builder().addAsNew(metaData.index("test")).build();
ClusterState clusterState = ClusterState.builder(CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(routingTable).build();
// add 4 nodes
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2")).add(newNode("node3")).add(newNode("node4"))).build();
clusterState = allocation.reroute(clusterState, "reroute");
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().shardsWithState(UNASSIGNED).size(), equalTo(2));
// start primary shards
clusterState = allocation.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(2));
// start another replica shard, while keep one initializing
clusterState = allocation.applyStartedShards(clusterState, Collections.singletonList(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).get(0)));
assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(2));
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
// fail the primary shard, check one replica gets elected to primary, others become INITIALIZING (from it)
ShardRouting primaryShardToFail = clusterState.routingTable().index("test").shard(0).primaryShard();
ClusterState newState = allocation.applyFailedShard(clusterState, primaryShardToFail);
assertThat(newState, not(equalTo(clusterState)));
clusterState = newState;
assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(1));
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(2));
ShardRouting newPrimaryShard = clusterState.routingTable().index("test").shard(0).primaryShard();
assertThat(newPrimaryShard, not(equalTo(primaryShardToFail)));
}
use of org.elasticsearch.cluster.metadata.Metadata in project elasticsearch by elastic.
the class FilterAllocationDeciderTests method createInitialClusterState.
private ClusterState createInitialClusterState(AllocationService service, Settings settings) {
RecoverySource.Type recoveryType = randomFrom(RecoverySource.Type.EMPTY_STORE, RecoverySource.Type.LOCAL_SHARDS, RecoverySource.Type.SNAPSHOT);
MetaData.Builder metaData = MetaData.builder();
final Settings.Builder indexSettings = settings(Version.CURRENT).put(settings);
final IndexMetaData sourceIndex;
if (recoveryType == RecoverySource.Type.LOCAL_SHARDS) {
//put a fake closed source index
sourceIndex = IndexMetaData.builder("sourceIndex").settings(settings(Version.CURRENT)).numberOfShards(2).numberOfReplicas(0).putInSyncAllocationIds(0, Collections.singleton("aid0")).putInSyncAllocationIds(1, Collections.singleton("aid1")).build();
metaData.put(sourceIndex, false);
indexSettings.put(INDEX_SHRINK_SOURCE_UUID.getKey(), sourceIndex.getIndexUUID());
indexSettings.put(INDEX_SHRINK_SOURCE_NAME.getKey(), sourceIndex.getIndex().getName());
} else {
sourceIndex = null;
}
final IndexMetaData.Builder indexMetaDataBuilder = IndexMetaData.builder("idx").settings(indexSettings).numberOfShards(1).numberOfReplicas(1);
if (recoveryType == RecoverySource.Type.SNAPSHOT) {
indexMetaDataBuilder.putInSyncAllocationIds(0, Collections.singleton("_snapshot_restore"));
}
final IndexMetaData indexMetaData = indexMetaDataBuilder.build();
metaData.put(indexMetaData, false);
RoutingTable.Builder routingTableBuilder = RoutingTable.builder();
switch(recoveryType) {
case EMPTY_STORE:
routingTableBuilder.addAsNew(indexMetaData);
break;
case SNAPSHOT:
routingTableBuilder.addAsRestore(indexMetaData, new RecoverySource.SnapshotRecoverySource(new Snapshot("repository", new SnapshotId("snapshot_name", "snapshot_uuid")), Version.CURRENT, indexMetaData.getIndex().getName()));
break;
case LOCAL_SHARDS:
routingTableBuilder.addAsFromCloseToOpen(sourceIndex);
routingTableBuilder.addAsNew(indexMetaData);
break;
default:
throw new UnsupportedOperationException(recoveryType + " is not supported");
}
RoutingTable routingTable = routingTableBuilder.build();
ClusterState clusterState = ClusterState.builder(org.elasticsearch.cluster.ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(routingTable).build();
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
return service.reroute(clusterState, "reroute", false);
}
Aggregations