use of org.elasticsearch.cluster.metadata.MetaData in project elasticsearch by elastic.
the class DiskThresholdDeciderTests method testShardRelocationsTakenIntoAccount.
public void testShardRelocationsTakenIntoAccount() {
Settings diskSettings = Settings.builder().put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_THRESHOLD_ENABLED_SETTING.getKey(), true).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_INCLUDE_RELOCATIONS_SETTING.getKey(), true).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_LOW_DISK_WATERMARK_SETTING.getKey(), 0.7).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), 0.8).build();
ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
// 60% used
usagesBuilder.put("node1", new DiskUsage("node1", "n1", "/dev/null", 100, 40));
// 60% used
usagesBuilder.put("node2", new DiskUsage("node2", "n2", "/dev/null", 100, 40));
// 60% used
usagesBuilder.put("node3", new DiskUsage("node3", "n3", "/dev/null", 100, 40));
ImmutableOpenMap<String, DiskUsage> usages = usagesBuilder.build();
ImmutableOpenMap.Builder<String, Long> shardSizesBuilder = ImmutableOpenMap.builder();
// 14 bytes
shardSizesBuilder.put("[test][0][p]", 14L);
shardSizesBuilder.put("[test][0][r]", 14L);
// 1 bytes
shardSizesBuilder.put("[test2][0][p]", 1L);
shardSizesBuilder.put("[test2][0][r]", 1L);
ImmutableOpenMap<String, Long> shardSizes = shardSizesBuilder.build();
final ClusterInfo clusterInfo = new DevNullClusterInfo(usages, usages, shardSizes);
DiskThresholdDecider decider = makeDecider(diskSettings);
AllocationDeciders deciders = new AllocationDeciders(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, new ClusterSettings(Settings.EMPTY, ClusterSettings.BUILT_IN_CLUSTER_SETTINGS)), decider)));
ClusterInfoService cis = new ClusterInfoService() {
@Override
public ClusterInfo getClusterInfo() {
logger.info("--> calling fake getClusterInfo");
return clusterInfo;
}
@Override
public void addListener(Listener listener) {
// noop
}
};
AllocationService strategy = new AllocationService(Settings.builder().put("cluster.routing.allocation.node_concurrent_recoveries", 10).put(ClusterRebalanceAllocationDecider.CLUSTER_ROUTING_ALLOCATION_ALLOW_REBALANCE_SETTING.getKey(), "always").put("cluster.routing.allocation.cluster_concurrent_rebalance", -1).build(), deciders, new TestGatewayAllocator(), new BalancedShardsAllocator(Settings.EMPTY), cis);
MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(1)).put(IndexMetaData.builder("test2").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(1)).build();
RoutingTable initialRoutingTable = RoutingTable.builder().addAsNew(metaData.index("test")).addAsNew(metaData.index("test2")).build();
ClusterState clusterState = ClusterState.builder(ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(initialRoutingTable).build();
logger.info("--> adding two nodes");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
clusterState = strategy.reroute(clusterState, "reroute");
logShardStates(clusterState);
// shards should be initializing
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(4));
logger.info("--> start the shards");
clusterState = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
logShardStates(clusterState);
// Assert that we're able to start the primary and replicas
assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).size(), equalTo(4));
logger.info("--> adding node3");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node3"))).build();
AllocationCommand relocate1 = new MoveAllocationCommand("test", 0, "node2", "node3");
AllocationCommands cmds = new AllocationCommands(relocate1);
clusterState = strategy.reroute(clusterState, cmds, false, false).getClusterState();
logShardStates(clusterState);
AllocationCommand relocate2 = new MoveAllocationCommand("test2", 0, "node2", "node3");
cmds = new AllocationCommands(relocate2);
try {
// The shard for the "test" index is already being relocated to
// node3, which will put it over the low watermark when it
// completes, with shard relocations taken into account this should
// throw an exception about not being able to complete
strategy.reroute(clusterState, cmds, false, false);
fail("should not have been able to reroute the shard");
} catch (IllegalArgumentException e) {
assertThat("can't be allocated because there isn't enough room: " + e.getMessage(), e.getMessage(), containsString("the node is above the low watermark cluster setting " + "[cluster.routing.allocation.disk.watermark.low=0.7], using more disk space than the maximum " + "allowed [70.0%], actual free: [26.0%]"));
}
}
use of org.elasticsearch.cluster.metadata.MetaData in project elasticsearch by elastic.
the class FailedShardsRoutingTests method testRebalanceFailure.
public void testRebalanceFailure() {
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");
MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(2).numberOfReplicas(1)).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 two nodes and performing rerouting");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
clusterState = strategy.reroute(clusterState, "reroute");
logger.info("Start the shards (primaries)");
RoutingNodes routingNodes = clusterState.getRoutingNodes();
ClusterState newState = strategy.applyStartedShards(clusterState, routingNodes.shardsWithState(INITIALIZING));
assertThat(newState, not(equalTo(clusterState)));
clusterState = newState;
assertThat(clusterState.routingTable().index("test").shards().size(), equalTo(2));
for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
assertThat(clusterState.routingTable().index("test").shard(i).size(), equalTo(2));
assertThat(clusterState.routingTable().index("test").shard(i).shards().size(), equalTo(2));
assertThat(clusterState.routingTable().index("test").shard(i).primaryShard().state(), equalTo(STARTED));
assertThat(clusterState.routingTable().index("test").shard(i).primaryShard().currentNodeId(), anyOf(equalTo("node1"), equalTo("node2")));
assertThat(clusterState.routingTable().index("test").shard(i).replicaShards().size(), equalTo(1));
assertThat(clusterState.routingTable().index("test").shard(i).replicaShards().get(0).state(), equalTo(INITIALIZING));
assertThat(clusterState.routingTable().index("test").shard(i).replicaShards().get(0).currentNodeId(), anyOf(equalTo("node2"), equalTo("node1")));
}
logger.info("Start the shards (backups)");
routingNodes = clusterState.getRoutingNodes();
newState = strategy.applyStartedShards(clusterState, routingNodes.shardsWithState(INITIALIZING));
assertThat(newState, not(equalTo(clusterState)));
clusterState = newState;
assertThat(clusterState.routingTable().index("test").shards().size(), equalTo(2));
for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
assertThat(clusterState.routingTable().index("test").shard(i).size(), equalTo(2));
assertThat(clusterState.routingTable().index("test").shard(i).shards().size(), equalTo(2));
assertThat(clusterState.routingTable().index("test").shard(i).primaryShard().state(), equalTo(STARTED));
assertThat(clusterState.routingTable().index("test").shard(i).primaryShard().currentNodeId(), anyOf(equalTo("node1"), equalTo("node2")));
assertThat(clusterState.routingTable().index("test").shard(i).replicaShards().size(), equalTo(1));
assertThat(clusterState.routingTable().index("test").shard(i).replicaShards().get(0).state(), equalTo(STARTED));
assertThat(clusterState.routingTable().index("test").shard(i).replicaShards().get(0).currentNodeId(), anyOf(equalTo("node2"), equalTo("node1")));
}
logger.info("Adding third node and reroute");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node3"))).build();
newState = strategy.reroute(clusterState, "reroute");
assertThat(newState, not(equalTo(clusterState)));
clusterState = newState;
routingNodes = clusterState.getRoutingNodes();
assertThat(clusterState.routingTable().index("test").shards().size(), equalTo(2));
assertThat(routingNodes.node("node1").numberOfShardsWithState(STARTED, RELOCATING), equalTo(2));
assertThat(routingNodes.node("node1").numberOfShardsWithState(STARTED), lessThan(3));
assertThat(routingNodes.node("node2").numberOfShardsWithState(STARTED, RELOCATING), equalTo(2));
assertThat(routingNodes.node("node2").numberOfShardsWithState(STARTED), lessThan(3));
assertThat(routingNodes.node("node3").numberOfShardsWithState(INITIALIZING), equalTo(1));
logger.info("Fail the shards on node 3");
ShardRouting shardToFail = routingNodes.node("node3").iterator().next();
newState = strategy.applyFailedShard(clusterState, shardToFail);
assertThat(newState, not(equalTo(clusterState)));
clusterState = newState;
routingNodes = clusterState.getRoutingNodes();
assertThat(clusterState.routingTable().index("test").shards().size(), equalTo(2));
assertThat(routingNodes.node("node1").numberOfShardsWithState(STARTED, RELOCATING), equalTo(2));
assertThat(routingNodes.node("node1").numberOfShardsWithState(STARTED), lessThan(3));
assertThat(routingNodes.node("node2").numberOfShardsWithState(STARTED, RELOCATING), equalTo(2));
assertThat(routingNodes.node("node2").numberOfShardsWithState(STARTED), lessThan(3));
assertThat(routingNodes.node("node3").numberOfShardsWithState(INITIALIZING), equalTo(1));
// make sure the failedShard is not INITIALIZING again on node3
assertThat(routingNodes.node("node3").iterator().next().shardId(), not(equalTo(shardToFail.shardId())));
}
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")));
}
Aggregations