Search in sources :

Example 21 with RoutingNode

use of org.elasticsearch.cluster.routing.RoutingNode in project elasticsearch by elastic.

the class IndicesClusterStateService method updateFailedShardsCache.

/**
     * Removes shard entries from the failed shards cache that are no longer allocated to this node by the master.
     * Sends shard failures for shards that are marked as actively allocated to this node but don't actually exist on the node.
     * Resends shard failures for shards that are still marked as allocated to this node but previously failed.
     *
     * @param state new cluster state
     */
private void updateFailedShardsCache(final ClusterState state) {
    RoutingNode localRoutingNode = state.getRoutingNodes().node(state.nodes().getLocalNodeId());
    if (localRoutingNode == null) {
        failedShardsCache.clear();
        return;
    }
    DiscoveryNode masterNode = state.nodes().getMasterNode();
    // remove items from cache which are not in our routing table anymore and resend failures that have not executed on master yet
    for (Iterator<Map.Entry<ShardId, ShardRouting>> iterator = failedShardsCache.entrySet().iterator(); iterator.hasNext(); ) {
        ShardRouting failedShardRouting = iterator.next().getValue();
        ShardRouting matchedRouting = localRoutingNode.getByShardId(failedShardRouting.shardId());
        if (matchedRouting == null || matchedRouting.isSameAllocation(failedShardRouting) == false) {
            iterator.remove();
        } else {
            if (masterNode != null) {
                // TODO: can we remove this? Is resending shard failures the responsibility of shardStateAction?
                String message = "master " + masterNode + " has not removed previously failed shard. resending shard failure";
                logger.trace("[{}] re-sending failed shard [{}], reason [{}]", matchedRouting.shardId(), matchedRouting, message);
                shardStateAction.localShardFailed(matchedRouting, message, null, SHARD_STATE_ACTION_LISTENER, state);
            }
        }
    }
}
Also used : DiscoveryNode(org.elasticsearch.cluster.node.DiscoveryNode) RoutingNode(org.elasticsearch.cluster.routing.RoutingNode) ShardRouting(org.elasticsearch.cluster.routing.ShardRouting)

Example 22 with RoutingNode

use of org.elasticsearch.cluster.routing.RoutingNode in project elasticsearch by elastic.

the class IndicesClusterStateService method createOrUpdateShards.

private void createOrUpdateShards(final ClusterState state) {
    RoutingNode localRoutingNode = state.getRoutingNodes().node(state.nodes().getLocalNodeId());
    if (localRoutingNode == null) {
        return;
    }
    DiscoveryNodes nodes = state.nodes();
    RoutingTable routingTable = state.routingTable();
    for (final ShardRouting shardRouting : localRoutingNode) {
        ShardId shardId = shardRouting.shardId();
        if (failedShardsCache.containsKey(shardId) == false) {
            AllocatedIndex<? extends Shard> indexService = indicesService.indexService(shardId.getIndex());
            assert indexService != null : "index " + shardId.getIndex() + " should have been created by createIndices";
            Shard shard = indexService.getShardOrNull(shardId.id());
            if (shard == null) {
                assert shardRouting.initializing() : shardRouting + " should have been removed by failMissingShards";
                createShard(nodes, routingTable, shardRouting, state);
            } else {
                updateShard(nodes, shardRouting, shard, routingTable, state);
            }
        }
    }
}
Also used : ShardId(org.elasticsearch.index.shard.ShardId) RoutingNode(org.elasticsearch.cluster.routing.RoutingNode) IndexShardRoutingTable(org.elasticsearch.cluster.routing.IndexShardRoutingTable) RoutingTable(org.elasticsearch.cluster.routing.RoutingTable) ShardRouting(org.elasticsearch.cluster.routing.ShardRouting) IndexShard(org.elasticsearch.index.shard.IndexShard) DiscoveryNodes(org.elasticsearch.cluster.node.DiscoveryNodes)

Example 23 with RoutingNode

use of org.elasticsearch.cluster.routing.RoutingNode in project elasticsearch by elastic.

the class DiskThresholdDeciderTests method testCanRemainWithShardRelocatingAway.

public void testCanRemainWithShardRelocatingAway() {
    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(), "60%").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), "70%").build();
    // We have an index with 2 primary shards each taking 40 bytes. Each node has 100 bytes available
    ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
    // 80% used
    usagesBuilder.put("node1", new DiskUsage("node1", "n1", "/dev/null", 100, 20));
    // 0% used
    usagesBuilder.put("node2", new DiskUsage("node2", "n2", "/dev/null", 100, 100));
    ImmutableOpenMap<String, DiskUsage> usages = usagesBuilder.build();
    ImmutableOpenMap.Builder<String, Long> shardSizesBuilder = ImmutableOpenMap.builder();
    shardSizesBuilder.put("[test][0][p]", 40L);
    shardSizesBuilder.put("[test][1][p]", 40L);
    shardSizesBuilder.put("[foo][0][p]", 10L);
    ImmutableOpenMap<String, Long> shardSizes = shardSizesBuilder.build();
    final ClusterInfo clusterInfo = new DevNullClusterInfo(usages, usages, shardSizes);
    DiskThresholdDecider diskThresholdDecider = makeDecider(diskSettings);
    MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(2).numberOfReplicas(0)).put(IndexMetaData.builder("foo").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(0)).build();
    RoutingTable initialRoutingTable = RoutingTable.builder().addAsNew(metaData.index("test")).addAsNew(metaData.index("foo")).build();
    DiscoveryNode discoveryNode1 = new DiscoveryNode("node1", buildNewFakeTransportAddress(), emptyMap(), MASTER_DATA_ROLES, Version.CURRENT);
    DiscoveryNode discoveryNode2 = new DiscoveryNode("node2", buildNewFakeTransportAddress(), emptyMap(), MASTER_DATA_ROLES, Version.CURRENT);
    DiscoveryNodes discoveryNodes = DiscoveryNodes.builder().add(discoveryNode1).add(discoveryNode2).build();
    ClusterState baseClusterState = ClusterState.builder(ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(initialRoutingTable).nodes(discoveryNodes).build();
    // Two shards consuming each 80% of disk space while 70% is allowed, so shard 0 isn't allowed here
    ShardRouting firstRouting = TestShardRouting.newShardRouting("test", 0, "node1", null, true, ShardRoutingState.STARTED);
    ShardRouting secondRouting = TestShardRouting.newShardRouting("test", 1, "node1", null, true, ShardRoutingState.STARTED);
    RoutingNode firstRoutingNode = new RoutingNode("node1", discoveryNode1, firstRouting, secondRouting);
    RoutingTable.Builder builder = RoutingTable.builder().add(IndexRoutingTable.builder(firstRouting.index()).addIndexShard(new IndexShardRoutingTable.Builder(firstRouting.shardId()).addShard(firstRouting).build()).addIndexShard(new IndexShardRoutingTable.Builder(secondRouting.shardId()).addShard(secondRouting).build()));
    ClusterState clusterState = ClusterState.builder(baseClusterState).routingTable(builder.build()).build();
    RoutingAllocation routingAllocation = new RoutingAllocation(null, new RoutingNodes(clusterState), clusterState, clusterInfo, System.nanoTime(), false);
    routingAllocation.debugDecision(true);
    Decision decision = diskThresholdDecider.canRemain(firstRouting, firstRoutingNode, routingAllocation);
    assertThat(decision.type(), equalTo(Decision.Type.NO));
    assertThat(((Decision.Single) decision).getExplanation(), containsString("the shard cannot remain on this node because it is above the high watermark cluster setting " + "[cluster.routing.allocation.disk.watermark.high=70%] and there is less than the required [30.0%] free disk on node, " + "actual free: [20.0%]"));
    // Two shards consuming each 80% of disk space while 70% is allowed, but one is relocating, so shard 0 can stay
    firstRouting = TestShardRouting.newShardRouting("test", 0, "node1", null, true, ShardRoutingState.STARTED);
    secondRouting = TestShardRouting.newShardRouting("test", 1, "node1", "node2", true, ShardRoutingState.RELOCATING);
    ShardRouting fooRouting = TestShardRouting.newShardRouting("foo", 0, "node1", null, true, ShardRoutingState.UNASSIGNED);
    firstRoutingNode = new RoutingNode("node1", discoveryNode1, firstRouting, secondRouting);
    builder = RoutingTable.builder().add(IndexRoutingTable.builder(firstRouting.index()).addIndexShard(new IndexShardRoutingTable.Builder(firstRouting.shardId()).addShard(firstRouting).build()).addIndexShard(new IndexShardRoutingTable.Builder(secondRouting.shardId()).addShard(secondRouting).build()));
    clusterState = ClusterState.builder(baseClusterState).routingTable(builder.build()).build();
    routingAllocation = new RoutingAllocation(null, new RoutingNodes(clusterState), clusterState, clusterInfo, System.nanoTime(), false);
    routingAllocation.debugDecision(true);
    decision = diskThresholdDecider.canRemain(firstRouting, firstRoutingNode, routingAllocation);
    assertThat(decision.type(), equalTo(Decision.Type.YES));
    assertEquals("there is enough disk on this node for the shard to remain, free: [60b]", ((Decision.Single) decision).getExplanation());
    decision = diskThresholdDecider.canAllocate(fooRouting, firstRoutingNode, routingAllocation);
    assertThat(decision.type(), equalTo(Decision.Type.NO));
    if (fooRouting.recoverySource().getType() == RecoverySource.Type.EMPTY_STORE) {
        assertThat(((Decision.Single) decision).getExplanation(), containsString("the node is above the high watermark cluster setting [cluster.routing.allocation.disk.watermark.high=70%], using " + "more disk space than the maximum allowed [70.0%], actual free: [20.0%]"));
    } else {
        assertThat(((Decision.Single) decision).getExplanation(), containsString("the node is above the low watermark cluster setting [cluster.routing.allocation.disk.watermark.low=60%], using more " + "disk space than the maximum allowed [60.0%], actual free: [20.0%]"));
    }
    // Creating AllocationService instance and the services it depends on...
    ClusterInfoService cis = new ClusterInfoService() {

        @Override
        public ClusterInfo getClusterInfo() {
            logger.info("--> calling fake getClusterInfo");
            return clusterInfo;
        }

        @Override
        public void addListener(Listener listener) {
        // noop
        }
    };
    AllocationDeciders deciders = new AllocationDeciders(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, new ClusterSettings(Settings.EMPTY, ClusterSettings.BUILT_IN_CLUSTER_SETTINGS)), diskThresholdDecider)));
    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);
    // Ensure that the reroute call doesn't alter the routing table, since the first primary is relocating away
    // and therefor we will have sufficient disk space on node1.
    ClusterState result = strategy.reroute(clusterState, "reroute");
    assertThat(result, equalTo(clusterState));
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().state(), equalTo(STARTED));
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().currentNodeId(), equalTo("node1"));
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().relocatingNodeId(), nullValue());
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().state(), equalTo(RELOCATING));
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().currentNodeId(), equalTo("node1"));
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().relocatingNodeId(), equalTo("node2"));
}
Also used : IndexShardRoutingTable(org.elasticsearch.cluster.routing.IndexShardRoutingTable) DiscoveryNode(org.elasticsearch.cluster.node.DiscoveryNode) ClusterSettings(org.elasticsearch.common.settings.ClusterSettings) RoutingNodes(org.elasticsearch.cluster.routing.RoutingNodes) DevNullClusterInfo(org.elasticsearch.cluster.MockInternalClusterInfoService.DevNullClusterInfo) Matchers.containsString(org.hamcrest.Matchers.containsString) DiskUsage(org.elasticsearch.cluster.DiskUsage) ImmutableOpenMap(org.elasticsearch.common.collect.ImmutableOpenMap) RoutingNode(org.elasticsearch.cluster.routing.RoutingNode) MetaData(org.elasticsearch.cluster.metadata.MetaData) IndexMetaData(org.elasticsearch.cluster.metadata.IndexMetaData) DiskThresholdSettings(org.elasticsearch.cluster.routing.allocation.DiskThresholdSettings) Settings(org.elasticsearch.common.settings.Settings) ClusterSettings(org.elasticsearch.common.settings.ClusterSettings) DiscoveryNodes(org.elasticsearch.cluster.node.DiscoveryNodes) AllocationService(org.elasticsearch.cluster.routing.allocation.AllocationService) TestGatewayAllocator(org.elasticsearch.test.gateway.TestGatewayAllocator) ClusterState(org.elasticsearch.cluster.ClusterState) ClusterInfoService(org.elasticsearch.cluster.ClusterInfoService) BalancedShardsAllocator(org.elasticsearch.cluster.routing.allocation.allocator.BalancedShardsAllocator) DevNullClusterInfo(org.elasticsearch.cluster.MockInternalClusterInfoService.DevNullClusterInfo) ClusterInfo(org.elasticsearch.cluster.ClusterInfo) IndexShardRoutingTable(org.elasticsearch.cluster.routing.IndexShardRoutingTable) IndexRoutingTable(org.elasticsearch.cluster.routing.IndexRoutingTable) RoutingTable(org.elasticsearch.cluster.routing.RoutingTable) ShardRouting(org.elasticsearch.cluster.routing.ShardRouting) TestShardRouting(org.elasticsearch.cluster.routing.TestShardRouting) RoutingAllocation(org.elasticsearch.cluster.routing.allocation.RoutingAllocation)

Example 24 with RoutingNode

use of org.elasticsearch.cluster.routing.RoutingNode in project elasticsearch by elastic.

the class RebalanceAfterActiveTests method testRebalanceOnlyAfterAllShardsAreActive.

public void testRebalanceOnlyAfterAllShardsAreActive() {
    final long[] sizes = new long[5];
    for (int i = 0; i < sizes.length; i++) {
        sizes[i] = randomIntBetween(0, Integer.MAX_VALUE);
    }
    AllocationService strategy = createAllocationService(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(), new ClusterInfoService() {

        @Override
        public ClusterInfo getClusterInfo() {
            return new ClusterInfo() {

                @Override
                public Long getShardSize(ShardRouting shardRouting) {
                    if (shardRouting.getIndexName().equals("test")) {
                        return sizes[shardRouting.getId()];
                    }
                    return null;
                }
            };
        }

        @Override
        public void addListener(Listener listener) {
        }
    });
    logger.info("Building initial routing table");
    MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(5).numberOfReplicas(1)).build();
    RoutingTable initialRoutingTable = RoutingTable.builder().addAsNew(metaData.index("test")).build();
    ClusterState clusterState = ClusterState.builder(org.elasticsearch.cluster.ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(initialRoutingTable).build();
    assertThat(clusterState.routingTable().index("test").shards().size(), equalTo(5));
    for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
        assertThat(clusterState.routingTable().index("test").shard(i).shards().size(), equalTo(2));
        assertThat(clusterState.routingTable().index("test").shard(i).shards().get(0).state(), equalTo(UNASSIGNED));
        assertThat(clusterState.routingTable().index("test").shard(i).shards().get(1).state(), equalTo(UNASSIGNED));
        assertThat(clusterState.routingTable().index("test").shard(i).shards().get(0).currentNodeId(), nullValue());
        assertThat(clusterState.routingTable().index("test").shard(i).shards().get(1).currentNodeId(), nullValue());
    }
    logger.info("start two nodes and fully start the shards");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
    clusterState = strategy.reroute(clusterState, "reroute");
    for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
        assertThat(clusterState.routingTable().index("test").shard(i).shards().size(), equalTo(2));
        assertThat(clusterState.routingTable().index("test").shard(i).primaryShard().state(), equalTo(INITIALIZING));
        assertThat(clusterState.routingTable().index("test").shard(i).replicaShards().get(0).state(), equalTo(UNASSIGNED));
    }
    logger.info("start all the primary shards, replicas will start initializing");
    RoutingNodes routingNodes = clusterState.getRoutingNodes();
    clusterState = strategy.applyStartedShards(clusterState, routingNodes.shardsWithState(INITIALIZING));
    routingNodes = clusterState.getRoutingNodes();
    for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
        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).replicaShards().get(0).state(), equalTo(INITIALIZING));
        assertEquals(clusterState.routingTable().index("test").shard(i).replicaShards().get(0).getExpectedShardSize(), sizes[i]);
    }
    logger.info("now, start 8 more nodes, and check that no rebalancing/relocation have happened");
    clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node3")).add(newNode("node4")).add(newNode("node5")).add(newNode("node6")).add(newNode("node7")).add(newNode("node8")).add(newNode("node9")).add(newNode("node10"))).build();
    clusterState = strategy.reroute(clusterState, "reroute");
    routingNodes = clusterState.getRoutingNodes();
    for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
        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).replicaShards().get(0).state(), equalTo(INITIALIZING));
        assertEquals(clusterState.routingTable().index("test").shard(i).replicaShards().get(0).getExpectedShardSize(), sizes[i]);
    }
    logger.info("start the replica shards, rebalancing should start");
    routingNodes = clusterState.getRoutingNodes();
    clusterState = strategy.applyStartedShards(clusterState, routingNodes.shardsWithState(INITIALIZING));
    routingNodes = clusterState.getRoutingNodes();
    // we only allow one relocation at a time
    assertThat(clusterState.routingTable().shardsWithState(STARTED).size(), equalTo(5));
    assertThat(clusterState.routingTable().shardsWithState(RELOCATING).size(), equalTo(5));
    for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
        int num = 0;
        for (ShardRouting routing : clusterState.routingTable().index("test").shard(i).shards()) {
            if (routing.state() == RELOCATING || routing.state() == INITIALIZING) {
                assertEquals(routing.getExpectedShardSize(), sizes[i]);
                num++;
            }
        }
        assertTrue(num > 0);
    }
    logger.info("complete relocation, other half of relocation should happen");
    routingNodes = clusterState.getRoutingNodes();
    clusterState = strategy.applyStartedShards(clusterState, routingNodes.shardsWithState(INITIALIZING));
    routingNodes = clusterState.getRoutingNodes();
    // we now only relocate 3, since 2 remain where they are!
    assertThat(clusterState.routingTable().shardsWithState(STARTED).size(), equalTo(7));
    assertThat(clusterState.routingTable().shardsWithState(RELOCATING).size(), equalTo(3));
    for (int i = 0; i < clusterState.routingTable().index("test").shards().size(); i++) {
        for (ShardRouting routing : clusterState.routingTable().index("test").shard(i).shards()) {
            if (routing.state() == RELOCATING || routing.state() == INITIALIZING) {
                assertEquals(routing.getExpectedShardSize(), sizes[i]);
            }
        }
    }
    logger.info("complete relocation, that's it!");
    routingNodes = clusterState.getRoutingNodes();
    clusterState = strategy.applyStartedShards(clusterState, routingNodes.shardsWithState(INITIALIZING));
    routingNodes = clusterState.getRoutingNodes();
    assertThat(clusterState.routingTable().shardsWithState(STARTED).size(), equalTo(10));
    // make sure we have an even relocation
    for (RoutingNode routingNode : routingNodes) {
        assertThat(routingNode.size(), equalTo(1));
    }
}
Also used : ClusterState(org.elasticsearch.cluster.ClusterState) ClusterInfoService(org.elasticsearch.cluster.ClusterInfoService) RoutingNodes(org.elasticsearch.cluster.routing.RoutingNodes) ClusterInfo(org.elasticsearch.cluster.ClusterInfo) RoutingTable(org.elasticsearch.cluster.routing.RoutingTable) RoutingNode(org.elasticsearch.cluster.routing.RoutingNode) MetaData(org.elasticsearch.cluster.metadata.MetaData) IndexMetaData(org.elasticsearch.cluster.metadata.IndexMetaData) ShardRouting(org.elasticsearch.cluster.routing.ShardRouting)

Example 25 with RoutingNode

use of org.elasticsearch.cluster.routing.RoutingNode in project elasticsearch by elastic.

the class DiskThresholdDeciderTests method testForSingleDataNode.

public void testForSingleDataNode() {
    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(), "60%").put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), "70%").build();
    ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
    // 0% used
    usagesBuilder.put("node1", new DiskUsage("node1", "n1", "/dev/null", 100, 100));
    // 80% used
    usagesBuilder.put("node2", new DiskUsage("node2", "n2", "/dev/null", 100, 20));
    // 0% used
    usagesBuilder.put("node3", new DiskUsage("node3", "n3", "/dev/null", 100, 100));
    ImmutableOpenMap<String, DiskUsage> usages = usagesBuilder.build();
    // We have an index with 1 primary shards each taking 40 bytes. Each node has 100 bytes available
    ImmutableOpenMap.Builder<String, Long> shardSizes = ImmutableOpenMap.builder();
    shardSizes.put("[test][0][p]", 40L);
    shardSizes.put("[test][1][p]", 40L);
    final ClusterInfo clusterInfo = new DevNullClusterInfo(usages, usages, shardSizes.build());
    DiskThresholdDecider diskThresholdDecider = makeDecider(diskSettings);
    MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(2).numberOfReplicas(0)).build();
    RoutingTable initialRoutingTable = RoutingTable.builder().addAsNew(metaData.index("test")).build();
    logger.info("--> adding one master node, one data node");
    DiscoveryNode discoveryNode1 = new DiscoveryNode("", "node1", buildNewFakeTransportAddress(), emptyMap(), singleton(DiscoveryNode.Role.MASTER), Version.CURRENT);
    DiscoveryNode discoveryNode2 = new DiscoveryNode("", "node2", buildNewFakeTransportAddress(), emptyMap(), singleton(DiscoveryNode.Role.DATA), Version.CURRENT);
    DiscoveryNodes discoveryNodes = DiscoveryNodes.builder().add(discoveryNode1).add(discoveryNode2).build();
    ClusterState baseClusterState = ClusterState.builder(ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(initialRoutingTable).nodes(discoveryNodes).build();
    // Two shards consumes 80% of disk space in data node, but we have only one data node, shards should remain.
    ShardRouting firstRouting = TestShardRouting.newShardRouting("test", 0, "node2", null, true, ShardRoutingState.STARTED);
    ShardRouting secondRouting = TestShardRouting.newShardRouting("test", 1, "node2", null, true, ShardRoutingState.STARTED);
    RoutingNode firstRoutingNode = new RoutingNode("node2", discoveryNode2, firstRouting, secondRouting);
    RoutingTable.Builder builder = RoutingTable.builder().add(IndexRoutingTable.builder(firstRouting.index()).addIndexShard(new IndexShardRoutingTable.Builder(firstRouting.shardId()).addShard(firstRouting).build()).addIndexShard(new IndexShardRoutingTable.Builder(secondRouting.shardId()).addShard(secondRouting).build()));
    ClusterState clusterState = ClusterState.builder(baseClusterState).routingTable(builder.build()).build();
    RoutingAllocation routingAllocation = new RoutingAllocation(null, new RoutingNodes(clusterState), clusterState, clusterInfo, System.nanoTime(), false);
    routingAllocation.debugDecision(true);
    Decision decision = diskThresholdDecider.canRemain(firstRouting, firstRoutingNode, routingAllocation);
    // Two shards should start happily
    assertThat(decision.type(), equalTo(Decision.Type.YES));
    assertThat(((Decision.Single) decision).getExplanation(), containsString("there is only a single data node present"));
    ClusterInfoService cis = new ClusterInfoService() {

        @Override
        public ClusterInfo getClusterInfo() {
            logger.info("--> calling fake getClusterInfo");
            return clusterInfo;
        }

        @Override
        public void addListener(Listener listener) {
        }
    };
    AllocationDeciders deciders = new AllocationDeciders(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, new ClusterSettings(Settings.EMPTY, ClusterSettings.BUILT_IN_CLUSTER_SETTINGS)), diskThresholdDecider)));
    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);
    ClusterState result = strategy.reroute(clusterState, "reroute");
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().state(), equalTo(STARTED));
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().currentNodeId(), equalTo("node2"));
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().relocatingNodeId(), nullValue());
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().state(), equalTo(STARTED));
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().currentNodeId(), equalTo("node2"));
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().relocatingNodeId(), nullValue());
    // Add another datanode, it should relocate.
    logger.info("--> adding node3");
    DiscoveryNode discoveryNode3 = new DiscoveryNode("", "node3", buildNewFakeTransportAddress(), emptyMap(), singleton(DiscoveryNode.Role.DATA), Version.CURRENT);
    ClusterState updateClusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(discoveryNode3)).build();
    firstRouting = TestShardRouting.newShardRouting("test", 0, "node2", null, true, ShardRoutingState.STARTED);
    secondRouting = TestShardRouting.newShardRouting("test", 1, "node2", "node3", true, ShardRoutingState.RELOCATING);
    firstRoutingNode = new RoutingNode("node2", discoveryNode2, firstRouting, secondRouting);
    builder = RoutingTable.builder().add(IndexRoutingTable.builder(firstRouting.index()).addIndexShard(new IndexShardRoutingTable.Builder(firstRouting.shardId()).addShard(firstRouting).build()).addIndexShard(new IndexShardRoutingTable.Builder(secondRouting.shardId()).addShard(secondRouting).build()));
    clusterState = ClusterState.builder(updateClusterState).routingTable(builder.build()).build();
    routingAllocation = new RoutingAllocation(null, new RoutingNodes(clusterState), clusterState, clusterInfo, System.nanoTime(), false);
    routingAllocation.debugDecision(true);
    decision = diskThresholdDecider.canRemain(firstRouting, firstRoutingNode, routingAllocation);
    assertThat(decision.type(), equalTo(Decision.Type.YES));
    assertThat(((Decision.Single) decision).getExplanation(), containsString("there is enough disk on this node for the shard to remain, free: [60b]"));
    result = strategy.reroute(clusterState, "reroute");
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().state(), equalTo(STARTED));
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().currentNodeId(), equalTo("node2"));
    assertThat(result.routingTable().index("test").getShards().get(0).primaryShard().relocatingNodeId(), nullValue());
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().state(), equalTo(RELOCATING));
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().currentNodeId(), equalTo("node2"));
    assertThat(result.routingTable().index("test").getShards().get(1).primaryShard().relocatingNodeId(), equalTo("node3"));
}
Also used : IndexShardRoutingTable(org.elasticsearch.cluster.routing.IndexShardRoutingTable) DiscoveryNode(org.elasticsearch.cluster.node.DiscoveryNode) ClusterSettings(org.elasticsearch.common.settings.ClusterSettings) RoutingNodes(org.elasticsearch.cluster.routing.RoutingNodes) DevNullClusterInfo(org.elasticsearch.cluster.MockInternalClusterInfoService.DevNullClusterInfo) Matchers.containsString(org.hamcrest.Matchers.containsString) DiskUsage(org.elasticsearch.cluster.DiskUsage) ImmutableOpenMap(org.elasticsearch.common.collect.ImmutableOpenMap) RoutingNode(org.elasticsearch.cluster.routing.RoutingNode) MetaData(org.elasticsearch.cluster.metadata.MetaData) IndexMetaData(org.elasticsearch.cluster.metadata.IndexMetaData) DiskThresholdSettings(org.elasticsearch.cluster.routing.allocation.DiskThresholdSettings) Settings(org.elasticsearch.common.settings.Settings) ClusterSettings(org.elasticsearch.common.settings.ClusterSettings) DiscoveryNodes(org.elasticsearch.cluster.node.DiscoveryNodes) AllocationService(org.elasticsearch.cluster.routing.allocation.AllocationService) TestGatewayAllocator(org.elasticsearch.test.gateway.TestGatewayAllocator) ClusterState(org.elasticsearch.cluster.ClusterState) ClusterInfoService(org.elasticsearch.cluster.ClusterInfoService) BalancedShardsAllocator(org.elasticsearch.cluster.routing.allocation.allocator.BalancedShardsAllocator) DevNullClusterInfo(org.elasticsearch.cluster.MockInternalClusterInfoService.DevNullClusterInfo) ClusterInfo(org.elasticsearch.cluster.ClusterInfo) IndexShardRoutingTable(org.elasticsearch.cluster.routing.IndexShardRoutingTable) IndexRoutingTable(org.elasticsearch.cluster.routing.IndexRoutingTable) RoutingTable(org.elasticsearch.cluster.routing.RoutingTable) ShardRouting(org.elasticsearch.cluster.routing.ShardRouting) TestShardRouting(org.elasticsearch.cluster.routing.TestShardRouting) RoutingAllocation(org.elasticsearch.cluster.routing.allocation.RoutingAllocation)

Aggregations

RoutingNode (org.elasticsearch.cluster.routing.RoutingNode)61 ShardRouting (org.elasticsearch.cluster.routing.ShardRouting)45 ClusterState (org.elasticsearch.cluster.ClusterState)28 IndexMetaData (org.elasticsearch.cluster.metadata.IndexMetaData)23 DiscoveryNode (org.elasticsearch.cluster.node.DiscoveryNode)20 RoutingTable (org.elasticsearch.cluster.routing.RoutingTable)16 RoutingNodes (org.elasticsearch.cluster.routing.RoutingNodes)15 MetaData (org.elasticsearch.cluster.metadata.MetaData)13 Settings (org.elasticsearch.common.settings.Settings)12 Decision (org.elasticsearch.cluster.routing.allocation.decider.Decision)10 ShardId (org.elasticsearch.index.shard.ShardId)10 DiscoveryNodes (org.elasticsearch.cluster.node.DiscoveryNodes)9 IndexShardRoutingTable (org.elasticsearch.cluster.routing.IndexShardRoutingTable)9 Matchers.containsString (org.hamcrest.Matchers.containsString)9 UnassignedInfo (org.elasticsearch.cluster.routing.UnassignedInfo)8 RoutingAllocation (org.elasticsearch.cluster.routing.allocation.RoutingAllocation)8 ClusterSettings (org.elasticsearch.common.settings.ClusterSettings)8 IndexShard (org.elasticsearch.index.shard.IndexShard)8 ArrayList (java.util.ArrayList)7 HashSet (java.util.HashSet)7