Search in sources :

Example 71 with RoutingTable

use of org.opensearch.cluster.routing.RoutingTable in project OpenSearch by opensearch-project.

the class ActiveShardCountTests method initializeWithNewIndex.

private ClusterState initializeWithNewIndex(final String indexName, final int numShards, final int numReplicas) {
    // initial index creation and new routing table info
    final IndexMetadata indexMetadata = IndexMetadata.builder(indexName).settings(settings(Version.CURRENT).put(IndexMetadata.SETTING_INDEX_UUID, UUIDs.randomBase64UUID())).numberOfShards(numShards).numberOfReplicas(numReplicas).build();
    final Metadata metadata = Metadata.builder().put(indexMetadata, true).build();
    final RoutingTable routingTable = RoutingTable.builder().addAsNew(indexMetadata).build();
    return ClusterState.builder(new ClusterName("test_cluster")).metadata(metadata).routingTable(routingTable).build();
}
Also used : IndexShardRoutingTable(org.opensearch.cluster.routing.IndexShardRoutingTable) IndexRoutingTable(org.opensearch.cluster.routing.IndexRoutingTable) RoutingTable(org.opensearch.cluster.routing.RoutingTable) Metadata(org.opensearch.cluster.metadata.Metadata) IndexMetadata(org.opensearch.cluster.metadata.IndexMetadata) ClusterName(org.opensearch.cluster.ClusterName) IndexMetadata(org.opensearch.cluster.metadata.IndexMetadata)

Example 72 with RoutingTable

use of org.opensearch.cluster.routing.RoutingTable in project OpenSearch by opensearch-project.

the class ActiveShardCountTests method startAllShards.

private ClusterState startAllShards(final ClusterState clusterState, final String indexName) {
    RoutingTable routingTable = clusterState.routingTable();
    IndexRoutingTable indexRoutingTable = routingTable.index(indexName);
    IndexRoutingTable.Builder newIndexRoutingTable = IndexRoutingTable.builder(indexRoutingTable.getIndex());
    for (final ObjectCursor<IndexShardRoutingTable> shardEntry : indexRoutingTable.getShards().values()) {
        final IndexShardRoutingTable shardRoutingTable = shardEntry.value;
        for (ShardRouting shardRouting : shardRoutingTable.getShards()) {
            if (shardRouting.primary()) {
                assertTrue(shardRouting.active());
            } else {
                if (shardRouting.active() == false) {
                    shardRouting = shardRouting.initialize(randomAlphaOfLength(8), null, shardRouting.getExpectedShardSize()).moveToStarted();
                }
            }
            newIndexRoutingTable.addShard(shardRouting);
        }
    }
    routingTable = RoutingTable.builder(routingTable).add(newIndexRoutingTable).build();
    return ClusterState.builder(clusterState).routingTable(routingTable).build();
}
Also used : IndexRoutingTable(org.opensearch.cluster.routing.IndexRoutingTable) IndexShardRoutingTable(org.opensearch.cluster.routing.IndexShardRoutingTable) IndexShardRoutingTable(org.opensearch.cluster.routing.IndexShardRoutingTable) IndexRoutingTable(org.opensearch.cluster.routing.IndexRoutingTable) RoutingTable(org.opensearch.cluster.routing.RoutingTable) ShardRouting(org.opensearch.cluster.routing.ShardRouting)

Example 73 with RoutingTable

use of org.opensearch.cluster.routing.RoutingTable in project OpenSearch by opensearch-project.

the class ClusterStateHealthTests method simulateClusterRecoveryStates.

private List<ClusterState> simulateClusterRecoveryStates(final String indexName, final boolean withPreviousAllocationIds, final boolean withPrimaryAllocationFailures) {
    final int numberOfShards = randomIntBetween(1, 5);
    final int numberOfReplicas = randomIntBetween(1, numberOfShards);
    // initial index creation and new routing table info
    IndexMetadata indexMetadata = IndexMetadata.builder(indexName).settings(settings(Version.CURRENT).put(IndexMetadata.SETTING_INDEX_UUID, UUIDs.randomBase64UUID())).numberOfShards(numberOfShards).numberOfReplicas(numberOfReplicas).state(IndexMetadata.State.OPEN).build();
    if (withPreviousAllocationIds) {
        final IndexMetadata.Builder idxMetaWithAllocationIds = IndexMetadata.builder(indexMetadata);
        boolean atLeastOne = false;
        for (int i = 0; i < numberOfShards; i++) {
            if (atLeastOne == false || randomBoolean()) {
                idxMetaWithAllocationIds.putInSyncAllocationIds(i, Sets.newHashSet(UUIDs.randomBase64UUID()));
                atLeastOne = true;
            }
        }
        indexMetadata = idxMetaWithAllocationIds.build();
    }
    final Metadata metadata = Metadata.builder().put(indexMetadata, true).build();
    final RoutingTable routingTable = RoutingTable.builder().addAsRecovery(indexMetadata).build();
    ClusterState clusterState = ClusterState.builder(new ClusterName("test_cluster")).metadata(metadata).routingTable(routingTable).build();
    return generateClusterStates(clusterState, indexName, numberOfReplicas, withPrimaryAllocationFailures);
}
Also used : ClusterState(org.opensearch.cluster.ClusterState) IndexShardRoutingTable(org.opensearch.cluster.routing.IndexShardRoutingTable) IndexRoutingTable(org.opensearch.cluster.routing.IndexRoutingTable) RoutingTable(org.opensearch.cluster.routing.RoutingTable) Metadata(org.opensearch.cluster.metadata.Metadata) IndexMetadata(org.opensearch.cluster.metadata.IndexMetadata) ClusterName(org.opensearch.cluster.ClusterName) IndexMetadata(org.opensearch.cluster.metadata.IndexMetadata)

Example 74 with RoutingTable

use of org.opensearch.cluster.routing.RoutingTable in project OpenSearch by opensearch-project.

the class ClusterStateHealthTests method simulateIndexCreationStates.

private List<ClusterState> simulateIndexCreationStates(final String indexName, final boolean withPrimaryAllocationFailures) {
    final int numberOfShards = randomIntBetween(1, 5);
    final int numberOfReplicas = randomIntBetween(1, numberOfShards);
    // initial index creation and new routing table info
    final IndexMetadata indexMetadata = IndexMetadata.builder(indexName).settings(settings(Version.CURRENT).put(IndexMetadata.SETTING_INDEX_UUID, UUIDs.randomBase64UUID())).numberOfShards(numberOfShards).numberOfReplicas(numberOfReplicas).build();
    final Metadata metadata = Metadata.builder().put(indexMetadata, true).build();
    final RoutingTable routingTable = RoutingTable.builder().addAsNew(indexMetadata).build();
    ClusterState clusterState = ClusterState.builder(new ClusterName("test_cluster")).metadata(metadata).routingTable(routingTable).build();
    return generateClusterStates(clusterState, indexName, numberOfReplicas, withPrimaryAllocationFailures);
}
Also used : ClusterState(org.opensearch.cluster.ClusterState) IndexShardRoutingTable(org.opensearch.cluster.routing.IndexShardRoutingTable) IndexRoutingTable(org.opensearch.cluster.routing.IndexRoutingTable) RoutingTable(org.opensearch.cluster.routing.RoutingTable) Metadata(org.opensearch.cluster.metadata.Metadata) IndexMetadata(org.opensearch.cluster.metadata.IndexMetadata) ClusterName(org.opensearch.cluster.ClusterName) IndexMetadata(org.opensearch.cluster.metadata.IndexMetadata)

Example 75 with RoutingTable

use of org.opensearch.cluster.routing.RoutingTable in project OpenSearch by opensearch-project.

the class DiskThresholdDeciderTests method testDiskThresholdWithAbsoluteSizes.

public void testDiskThresholdWithAbsoluteSizes() {
    Settings diskSettings = Settings.builder().put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_THRESHOLD_ENABLED_SETTING.getKey(), true).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_LOW_DISK_WATERMARK_SETTING.getKey(), "30b").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), "9b").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_FLOOD_STAGE_WATERMARK_SETTING.getKey(), "5b").build();
    ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
    // 90% used
    usagesBuilder.put("node1", new DiskUsage("node1", "n1", "/dev/null", 100, 10));
    // 90% used
    usagesBuilder.put("node2", new DiskUsage("node2", "n2", "/dev/null", 100, 10));
    // 40% used
    usagesBuilder.put("node3", new DiskUsage("node3", "n3", "/dev/null", 100, 60));
    // 20% used
    usagesBuilder.put("node4", new DiskUsage("node4", "n4", "/dev/null", 100, 80));
    // 15% used
    usagesBuilder.put("node5", new DiskUsage("node5", "n5", "/dev/null", 100, 85));
    ImmutableOpenMap<String, DiskUsage> usages = usagesBuilder.build();
    ImmutableOpenMap.Builder<String, Long> shardSizesBuilder = ImmutableOpenMap.builder();
    // 10 bytes
    shardSizesBuilder.put("[test][0][p]", 10L);
    shardSizesBuilder.put("[test][0][r]", 10L);
    ImmutableOpenMap<String, Long> shardSizes = shardSizesBuilder.build();
    final ClusterInfo clusterInfo = new DevNullClusterInfo(usages, usages, shardSizes);
    ClusterSettings clusterSettings = new ClusterSettings(Settings.EMPTY, ClusterSettings.BUILT_IN_CLUSTER_SETTINGS);
    AllocationDeciders deciders = new AllocationDeciders(new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, clusterSettings), makeDecider(diskSettings))));
    ClusterInfoService cis = () -> {
        logger.info("--> calling fake getClusterInfo");
        return clusterInfo;
    };
    AllocationService strategy = new AllocationService(deciders, new TestGatewayAllocator(), new BalancedShardsAllocator(Settings.EMPTY), cis, EmptySnapshotsInfoService.INSTANCE);
    Metadata metadata = Metadata.builder().put(IndexMetadata.builder("test").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(2)).build();
    RoutingTable initialRoutingTable = RoutingTable.builder().addAsNew(metadata.index("test")).build();
    ClusterState clusterState = ClusterState.builder(ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metadata(metadata).routingTable(initialRoutingTable).build();
    logger.info("--> adding node1 and node2 node");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Primary should initialize, even though both nodes are over the limit initialize
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
    String nodeWithPrimary, nodeWithoutPrimary;
    if (clusterState.getRoutingNodes().node("node1").size() == 1) {
        nodeWithPrimary = "node1";
        nodeWithoutPrimary = "node2";
    } else {
        nodeWithPrimary = "node2";
        nodeWithoutPrimary = "node1";
    }
    logger.info("--> nodeWithPrimary: {}", nodeWithPrimary);
    logger.info("--> nodeWithoutPrimary: {}", nodeWithoutPrimary);
    // Make node without the primary now habitable to replicas
    usagesBuilder = ImmutableOpenMap.builder(usages);
    // 65% used
    usagesBuilder.put(nodeWithoutPrimary, new DiskUsage(nodeWithoutPrimary, "", "/dev/null", 100, 35));
    usages = usagesBuilder.build();
    final ClusterInfo clusterInfo2 = new DevNullClusterInfo(usages, usages, shardSizes);
    cis = () -> {
        logger.info("--> calling fake getClusterInfo");
        return clusterInfo2;
    };
    strategy = new AllocationService(deciders, new TestGatewayAllocator(), new BalancedShardsAllocator(Settings.EMPTY), cis, EmptySnapshotsInfoService.INSTANCE);
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Now the replica should be able to initialize
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(2));
    logger.info("--> start the shards (primaries)");
    clusterState = startInitializingShardsAndReroute(strategy, clusterState);
    logShardStates(clusterState);
    // Assert that we're able to start the primary and replica, since they were both initializing
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).size(), equalTo(2));
    // Assert that node1 got a single shard (the primary), even though its disk usage is too high
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(1));
    // Assert that node2 got a single shard (a replica)
    assertThat(clusterState.getRoutingNodes().node("node2").size(), equalTo(1));
    // Assert that one replica is still unassigned
    // assertThat(clusterState.routingNodes().shardsWithState(ShardRoutingState.UNASSIGNED).size(), equalTo(1));
    logger.info("--> adding node3");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node3"))).build();
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Assert that the replica is initialized now that node3 is available with enough space
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).size(), equalTo(2));
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.INITIALIZING).size(), equalTo(1));
    logger.info("--> start the shards (replicas)");
    clusterState = startInitializingShardsAndReroute(strategy, clusterState);
    logShardStates(clusterState);
    // Assert that all replicas could be started
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).size(), equalTo(3));
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node2").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node3").size(), equalTo(1));
    logger.info("--> changing decider settings");
    // Set the low threshold to 60 instead of 70
    // Set the high threshold to 70 instead of 80
    // node2 now should not have new shards allocated to it, but shards can remain
    diskSettings = Settings.builder().put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_THRESHOLD_ENABLED_SETTING.getKey(), true).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_LOW_DISK_WATERMARK_SETTING.getKey(), "40b").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), "30b").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_FLOOD_STAGE_WATERMARK_SETTING.getKey(), "20b").build();
    deciders = new AllocationDeciders(new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, clusterSettings), makeDecider(diskSettings))));
    strategy = new AllocationService(deciders, new TestGatewayAllocator(), new BalancedShardsAllocator(Settings.EMPTY), cis, EmptySnapshotsInfoService.INSTANCE);
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Shards remain started
    assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(3));
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node2").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node3").size(), equalTo(1));
    logger.info("--> changing settings again");
    // Set the low threshold to 50 instead of 60
    // Set the high threshold to 60 instead of 70
    // node2 now should not have new shards allocated to it, and shards cannot remain
    diskSettings = Settings.builder().put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_THRESHOLD_ENABLED_SETTING.getKey(), true).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_LOW_DISK_WATERMARK_SETTING.getKey(), "50b").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), "40b").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_FLOOD_STAGE_WATERMARK_SETTING.getKey(), "30b").build();
    deciders = new AllocationDeciders(new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, clusterSettings), makeDecider(diskSettings))));
    strategy = new AllocationService(deciders, new TestGatewayAllocator(), new BalancedShardsAllocator(Settings.EMPTY), cis, EmptySnapshotsInfoService.INSTANCE);
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Shards remain started
    assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(3));
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(1));
    // Shard hasn't been moved off of node2 yet because there's nowhere for it to go
    assertThat(clusterState.getRoutingNodes().node("node2").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node3").size(), equalTo(1));
    logger.info("--> adding node4");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node4"))).build();
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Shards remain started
    assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(2));
    // One shard is relocating off of node1
    assertThat(clusterState.getRoutingNodes().shardsWithState(RELOCATING).size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
    logger.info("--> apply INITIALIZING shards");
    clusterState = startInitializingShardsAndReroute(strategy, clusterState);
    logShardStates(clusterState);
    // primary shard already has been relocated away
    assertThat(clusterState.getRoutingNodes().node(nodeWithPrimary).size(), equalTo(0));
    // node with increased space still has its shard
    assertThat(clusterState.getRoutingNodes().node(nodeWithoutPrimary).size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node3").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node4").size(), equalTo(1));
    logger.info("--> adding node5");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node5"))).build();
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Shards remain started on node3 and node4
    assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(2));
    // One shard is relocating off of node2 now
    assertThat(clusterState.getRoutingNodes().shardsWithState(RELOCATING).size(), equalTo(1));
    // Initializing on node5
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
    logger.info("--> apply INITIALIZING shards");
    clusterState = startInitializingShardsAndReroute(strategy, clusterState);
    logger.info("--> final cluster state:");
    logShardStates(clusterState);
    // Node1 still has no shards because it has no space for them
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(0));
    // Node5 is available now, so the shard is moved off of node2
    assertThat(clusterState.getRoutingNodes().node("node2").size(), equalTo(0));
    assertThat(clusterState.getRoutingNodes().node("node3").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node4").size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node5").size(), equalTo(1));
}
Also used : TestGatewayAllocator(org.opensearch.test.gateway.TestGatewayAllocator) ClusterState(org.opensearch.cluster.ClusterState) ClusterSettings(org.opensearch.common.settings.ClusterSettings) ClusterInfoService(org.opensearch.cluster.ClusterInfoService) BalancedShardsAllocator(org.opensearch.cluster.routing.allocation.allocator.BalancedShardsAllocator) Metadata(org.opensearch.cluster.metadata.Metadata) IndexMetadata(org.opensearch.cluster.metadata.IndexMetadata) Matchers.containsString(org.hamcrest.Matchers.containsString) DiskUsage(org.opensearch.cluster.DiskUsage) ImmutableOpenMap(org.opensearch.common.collect.ImmutableOpenMap) ClusterInfo(org.opensearch.cluster.ClusterInfo) IndexShardRoutingTable(org.opensearch.cluster.routing.IndexShardRoutingTable) IndexRoutingTable(org.opensearch.cluster.routing.IndexRoutingTable) RoutingTable(org.opensearch.cluster.routing.RoutingTable) Settings(org.opensearch.common.settings.Settings) DiskThresholdSettings(org.opensearch.cluster.routing.allocation.DiskThresholdSettings) ClusterSettings(org.opensearch.common.settings.ClusterSettings) AllocationService(org.opensearch.cluster.routing.allocation.AllocationService) HashSet(java.util.HashSet) IntHashSet(com.carrotsearch.hppc.IntHashSet)

Aggregations

RoutingTable (org.opensearch.cluster.routing.RoutingTable)227 ClusterState (org.opensearch.cluster.ClusterState)193 IndexMetadata (org.opensearch.cluster.metadata.IndexMetadata)190 Metadata (org.opensearch.cluster.metadata.Metadata)187 ShardRouting (org.opensearch.cluster.routing.ShardRouting)81 IndexShardRoutingTable (org.opensearch.cluster.routing.IndexShardRoutingTable)58 IndexRoutingTable (org.opensearch.cluster.routing.IndexRoutingTable)55 RoutingNodes (org.opensearch.cluster.routing.RoutingNodes)42 AllocationService (org.opensearch.cluster.routing.allocation.AllocationService)42 ShardId (org.opensearch.index.shard.ShardId)35 DiscoveryNodes (org.opensearch.cluster.node.DiscoveryNodes)34 DiscoveryNode (org.opensearch.cluster.node.DiscoveryNode)33 Settings (org.opensearch.common.settings.Settings)33 Index (org.opensearch.index.Index)30 HashSet (java.util.HashSet)29 TestGatewayAllocator (org.opensearch.test.gateway.TestGatewayAllocator)29 ImmutableOpenMap (org.opensearch.common.collect.ImmutableOpenMap)28 ClusterSettings (org.opensearch.common.settings.ClusterSettings)28 RoutingNode (org.opensearch.cluster.routing.RoutingNode)24 BalancedShardsAllocator (org.opensearch.cluster.routing.allocation.allocator.BalancedShardsAllocator)23