Examples with ClusterInfoService org.elasticsearch.cluster.ClusterInfoService used on opensource projects

Example 16 with ClusterInfoService

use of org.elasticsearch.cluster.ClusterInfoService 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 17 with ClusterInfoService

use of org.elasticsearch.cluster.ClusterInfoService in project crate by crate.

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.builder().put(IndexMetadata.SETTING_VERSION_CREATED, Version.CURRENT).put(IndexMetadata.SETTING_NUMBER_OF_SHARDS, 2).put(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, 0).put(IndexMetadata.SETTING_AUTO_EXPAND_REPLICAS, "false"))).put(IndexMetadata.builder("foo").settings(Settings.builder().put(IndexMetadata.SETTING_VERSION_CREATED, Version.CURRENT).put(IndexMetadata.SETTING_NUMBER_OF_SHARDS, 1).put(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, 0).put(IndexMetadata.SETTING_AUTO_EXPAND_REPLICAS, "false"))).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());
    routingAllocation.debugDecision(true);
    Decision decision = diskThresholdDecider.canRemain(firstRouting, firstRoutingNode, routingAllocation);
    assertThat(decision.type(), equalTo(Decision.Type.NO));
    assertThat(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, 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());
    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.getExplanation());
    decision = diskThresholdDecider.canAllocate(fooRouting, firstRoutingNode, routingAllocation);
    assertThat(decision.type(), equalTo(Decision.Type.NO));
    if (fooRouting.recoverySource().getType() == RecoverySource.Type.EMPTY_STORE) {
        assertThat(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.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 = () -> {
        logger.info("--> calling fake getClusterInfo");
        return clusterInfo;
    };
    AllocationDeciders deciders = new AllocationDeciders(new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, new ClusterSettings(Settings.EMPTY, ClusterSettings.BUILT_IN_CLUSTER_SETTINGS)), diskThresholdDecider)));
    AllocationService strategy = new AllocationService(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"));
}