Examples with AllocationService org.elasticsearch.cluster.routing.allocation.AllocationService used on opensource projects

Example 26 with AllocationService

use of org.elasticsearch.cluster.routing.allocation.AllocationService in project elasticsearch by elastic.

the class RareClusterStateIT method testAssignmentWithJustAddedNodes.

public void testAssignmentWithJustAddedNodes() throws Exception {
    internalCluster().startNode();
    final String index = "index";
    prepareCreate(index).setSettings(IndexMetaData.SETTING_NUMBER_OF_SHARDS, 1, IndexMetaData.SETTING_NUMBER_OF_REPLICAS, 0).get();
    ensureGreen(index);
    // close to have some unassigned started shards shards..
    client().admin().indices().prepareClose(index).get();
    final String masterName = internalCluster().getMasterName();
    final ClusterService clusterService = internalCluster().clusterService(masterName);
    final AllocationService allocationService = internalCluster().getInstance(AllocationService.class, masterName);
    clusterService.submitStateUpdateTask("test-inject-node-and-reroute", new ClusterStateUpdateTask() {

        @Override
        public ClusterState execute(ClusterState currentState) throws Exception {
            // inject a node
            ClusterState.Builder builder = ClusterState.builder(currentState);
            builder.nodes(DiscoveryNodes.builder(currentState.nodes()).add(new DiscoveryNode("_non_existent", buildNewFakeTransportAddress(), emptyMap(), emptySet(), Version.CURRENT)));
            // open index
            final IndexMetaData indexMetaData = IndexMetaData.builder(currentState.metaData().index(index)).state(IndexMetaData.State.OPEN).build();
            builder.metaData(MetaData.builder(currentState.metaData()).put(indexMetaData, true));
            builder.blocks(ClusterBlocks.builder().blocks(currentState.blocks()).removeIndexBlocks(index));
            ClusterState updatedState = builder.build();
            RoutingTable.Builder routingTable = RoutingTable.builder(updatedState.routingTable());
            routingTable.addAsRecovery(updatedState.metaData().index(index));
            updatedState = ClusterState.builder(updatedState).routingTable(routingTable.build()).build();
            return allocationService.reroute(updatedState, "reroute");
        }

        @Override
        public void onFailure(String source, Exception e) {
        }
    });
    ensureGreen(index);
    // remove the extra node
    clusterService.submitStateUpdateTask("test-remove-injected-node", new ClusterStateUpdateTask() {

        @Override
        public ClusterState execute(ClusterState currentState) throws Exception {
            ClusterState.Builder builder = ClusterState.builder(currentState);
            builder.nodes(DiscoveryNodes.builder(currentState.nodes()).remove("_non_existent"));
            currentState = builder.build();
            return allocationService.deassociateDeadNodes(currentState, true, "reroute");
        }

        @Override
        public void onFailure(String source, Exception e) {
        }
    });
}

Example 27 with AllocationService

use of org.elasticsearch.cluster.routing.allocation.AllocationService in project elasticsearch by elastic.

the class TransportReplicationActionTests method testNoRerouteOnStaleClusterState.

/**
     * When relocating a primary shard, there is a cluster state update at the end of relocation where the active primary is switched from
     * the relocation source to the relocation target. If relocation source receives and processes this cluster state
     * before the relocation target, there is a time span where relocation source believes active primary to be on
     * relocation target and relocation target believes active primary to be on relocation source. This results in replication
     * requests being sent back and forth.
     * <p>
     * This test checks that replication request is not routed back from relocation target to relocation source in case of
     * stale index routing table on relocation target.
     */
public void testNoRerouteOnStaleClusterState() throws InterruptedException, ExecutionException {
    final String index = "test";
    final ShardId shardId = new ShardId(index, "_na_", 0);
    ClusterState state = state(index, true, ShardRoutingState.RELOCATING);
    String relocationTargetNode = state.getRoutingTable().shardRoutingTable(shardId).primaryShard().relocatingNodeId();
    state = ClusterState.builder(state).nodes(DiscoveryNodes.builder(state.nodes()).localNodeId(relocationTargetNode)).build();
    setState(clusterService, state);
    logger.debug("--> relocation ongoing state:\n{}", clusterService.state());
    Request request = new Request(shardId).timeout("1ms").routedBasedOnClusterVersion(clusterService.state().version() + 1);
    PlainActionFuture<TestResponse> listener = new PlainActionFuture<>();
    TestAction.ReroutePhase reroutePhase = action.new ReroutePhase(null, request, listener);
    reroutePhase.run();
    assertListenerThrows("cluster state too old didn't cause a timeout", listener, UnavailableShardsException.class);
    assertTrue(request.isRetrySet.compareAndSet(true, false));
    request = new Request(shardId).routedBasedOnClusterVersion(clusterService.state().version() + 1);
    listener = new PlainActionFuture<>();
    reroutePhase = action.new ReroutePhase(null, request, listener);
    reroutePhase.run();
    assertFalse("cluster state too old didn't cause a retry", listener.isDone());
    assertTrue(request.isRetrySet.get());
    // finish relocation
    ShardRouting relocationTarget = clusterService.state().getRoutingTable().shardRoutingTable(shardId).shardsWithState(ShardRoutingState.INITIALIZING).get(0);
    AllocationService allocationService = ESAllocationTestCase.createAllocationService();
    ClusterState updatedState = allocationService.applyStartedShards(state, Collections.singletonList(relocationTarget));
    setState(clusterService, updatedState);
    logger.debug("--> relocation complete state:\n{}", clusterService.state());
    IndexShardRoutingTable shardRoutingTable = clusterService.state().routingTable().index(index).shard(shardId.id());
    final String primaryNodeId = shardRoutingTable.primaryShard().currentNodeId();
    final List<CapturingTransport.CapturedRequest> capturedRequests = transport.getCapturedRequestsByTargetNodeAndClear().get(primaryNodeId);
    assertThat(capturedRequests, notNullValue());
    assertThat(capturedRequests.size(), equalTo(1));
    assertThat(capturedRequests.get(0).action, equalTo("testAction[p]"));
    assertIndexShardCounter(0);
}

Example 28 with AllocationService

use of org.elasticsearch.cluster.routing.allocation.AllocationService in project elasticsearch by elastic.

the class DiskThresholdDeciderTests method testUnknownDiskUsage.

public void testUnknownDiskUsage() {
    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(), 0.7).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), 0.85).build();
    ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
    // 50% used
    usagesBuilder.put("node2", new DiskUsage("node2", "node2", "/dev/null", 100, 50));
    // 100% used
    usagesBuilder.put("node3", new DiskUsage("node3", "node3", "/dev/null", 100, 0));
    ImmutableOpenMap<String, DiskUsage> usages = usagesBuilder.build();
    ImmutableOpenMap.Builder<String, Long> shardSizesBuilder = ImmutableOpenMap.builder();
    // 10 bytes
    shardSizesBuilder.put("[test][0][p]", 10L);
    // 10 bytes
    shardSizesBuilder.put("[test][0][r]", 10L);
    ImmutableOpenMap<String, Long> shardSizes = shardSizesBuilder.build();
    final ClusterInfo clusterInfo = new DevNullClusterInfo(usages, usages, shardSizes);
    AllocationDeciders deciders = new AllocationDeciders(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, new ClusterSettings(Settings.EMPTY, ClusterSettings.BUILT_IN_CLUSTER_SETTINGS)), makeDecider(diskSettings))));
    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(0)).build();
    RoutingTable routingTable = RoutingTable.builder().addAsNew(metaData.index("test")).build();
    ClusterState clusterState = ClusterState.builder(ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(routingTable).build();
    logger.info("--> adding node1");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(// node3 is added because DiskThresholdDecider automatically ignore single-node clusters
    newNode("node3"))).build();
    routingTable = strategy.reroute(clusterState, "reroute").routingTable();
    clusterState = ClusterState.builder(clusterState).routingTable(routingTable).build();
    // Shard can be allocated to node1, even though it only has 25% free,
    // because it's a primary that's never been allocated before
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
    logger.info("--> start the shards (primaries)");
    routingTable = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING)).routingTable();
    clusterState = ClusterState.builder(clusterState).routingTable(routingTable).build();
    logShardStates(clusterState);
    // A single shard is started on node1, even though it normally would not
    // be allowed, because it's a primary that hasn't been allocated, and node1
    // is still below the high watermark (unlike node3)
    assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(1));
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(1));
}

Example 29 with AllocationService

use of org.elasticsearch.cluster.routing.allocation.AllocationService in project elasticsearch by elastic.

the class DiskThresholdDeciderTests method testDiskThreshold.

public void testDiskThreshold() {
    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(), 0.7).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), 0.8).build();
    ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
    // 90% used
    usagesBuilder.put("node1", new DiskUsage("node1", "node1", "/dev/null", 100, 10));
    // 65% used
    usagesBuilder.put("node2", new DiskUsage("node2", "node2", "/dev/null", 100, 35));
    // 40% used
    usagesBuilder.put("node3", new DiskUsage("node3", "node3", "/dev/null", 100, 60));
    // 20% used
    usagesBuilder.put("node4", new DiskUsage("node4", "node4", "/dev/null", 100, 80));
    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(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, clusterSettings), makeDecider(diskSettings))));
    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)).build();
    final 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 two nodes");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Primary shard should be initializing, replica should not
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
    logger.info("--> start the shards (primaries)");
    clusterState = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
    logShardStates(clusterState);
    // Assert that we're able to start the primary
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).size(), equalTo(1));
    // Assert that node1 didn't get any shards because its disk usage is too high
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(0));
    logger.info("--> start the shards (replicas)");
    clusterState = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
    logShardStates(clusterState);
    // Assert that the replica couldn't be started since node1 doesn't have enough space
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).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(1));
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.INITIALIZING).size(), equalTo(1));
    logger.info("--> start the shards (replicas)");
    clusterState = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
    logShardStates(clusterState);
    // Assert that the replica couldn't be started since node1 doesn't have enough space
    assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).size(), equalTo(2));
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(0));
    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(), "60%").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), 0.7).build();
    deciders = new AllocationDeciders(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, clusterSettings), makeDecider(diskSettings))));
    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);
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Shards remain started
    assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(2));
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(0));
    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(), 0.5).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), 0.6).build();
    deciders = new AllocationDeciders(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, clusterSettings), makeDecider(diskSettings))));
    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);
    clusterState = strategy.reroute(clusterState, "reroute");
    logShardStates(clusterState);
    // Shards remain started
    assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(2));
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(0));
    // 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(1));
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(1));
    logger.info("--> apply INITIALIZING shards");
    clusterState = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
    logShardStates(clusterState);
    assertThat(clusterState.getRoutingNodes().node("node1").size(), equalTo(0));
    // Node4 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));
}

Example 30 with AllocationService

use of org.elasticsearch.cluster.routing.allocation.AllocationService in project elasticsearch by elastic.

the class EnableAllocationTests method testClusterEnableNone.

public void testClusterEnableNone() {
    AllocationService strategy = createAllocationService(Settings.builder().put(CLUSTER_ROUTING_ALLOCATION_ENABLE_SETTING.getKey(), Allocation.NONE.name()).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(org.elasticsearch.cluster.ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(routingTable).build();
    logger.info("--> adding two nodes and do rerouting");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
    routingTable = strategy.reroute(clusterState, "reroute").routingTable();
    clusterState = ClusterState.builder(clusterState).routingTable(routingTable).build();
    assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(0));
}