use of org.elasticsearch.cluster.ClusterInfoService in project elasticsearch by elastic.
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").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(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(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 = new ClusterInfoService() {
@Override
public ClusterInfo getClusterInfo() {
logger.info("--> calling fake getClusterInfo");
return clusterInfo2;
}
@Override
public void addListener(Listener listener) {
// noop
}
};
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);
// Now the replica should be able to initialize
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(2));
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 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 = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
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").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(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").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(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 = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
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 = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
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));
}
use of org.elasticsearch.cluster.ClusterInfoService in project elasticsearch by elastic.
the class DiskThresholdDeciderTests method testDiskThresholdWithShardSizes.
public void testDiskThresholdWithShardSizes() {
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(), "71%").build();
ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
// 69% used
usagesBuilder.put("node1", new DiskUsage("node1", "n1", "/dev/null", 100, 31));
// 99% used
usagesBuilder.put("node2", new DiskUsage("node2", "n2", "/dev/null", 100, 1));
ImmutableOpenMap<String, DiskUsage> usages = usagesBuilder.build();
ImmutableOpenMap.Builder<String, Long> shardSizesBuilder = ImmutableOpenMap.builder();
// 10 bytes
shardSizesBuilder.put("[test][0][p]", 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(// node2 is added because DiskThresholdDecider automatically ignore single-node clusters
newNode("node2"))).build();
routingTable = strategy.reroute(clusterState, "reroute").routingTable();
clusterState = ClusterState.builder(clusterState).routingTable(routingTable).build();
logger.info("--> start the shards (primaries)");
routingTable = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING)).routingTable();
clusterState = ClusterState.builder(clusterState).routingTable(routingTable).build();
logShardStates(clusterState);
// Shard can't be allocated to node1 (or node2) because it would cause too much usage
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(0));
// No shards are started, no nodes have enough disk for allocation
assertThat(clusterState.getRoutingNodes().shardsWithState(STARTED).size(), equalTo(0));
}
use of org.elasticsearch.cluster.ClusterInfoService 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"));
}
use of org.elasticsearch.cluster.ClusterInfoService in project elasticsearch by elastic.
the class DiskThresholdDeciderTests method testShardRelocationsTakenIntoAccount.
public void testShardRelocationsTakenIntoAccount() {
Settings diskSettings = Settings.builder().put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_DISK_THRESHOLD_ENABLED_SETTING.getKey(), true).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_INCLUDE_RELOCATIONS_SETTING.getKey(), true).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_LOW_DISK_WATERMARK_SETTING.getKey(), 0.7).put(DiskThresholdSettings.CLUSTER_ROUTING_ALLOCATION_HIGH_DISK_WATERMARK_SETTING.getKey(), 0.8).build();
ImmutableOpenMap.Builder<String, DiskUsage> usagesBuilder = ImmutableOpenMap.builder();
// 60% used
usagesBuilder.put("node1", new DiskUsage("node1", "n1", "/dev/null", 100, 40));
// 60% used
usagesBuilder.put("node2", new DiskUsage("node2", "n2", "/dev/null", 100, 40));
// 60% used
usagesBuilder.put("node3", new DiskUsage("node3", "n3", "/dev/null", 100, 40));
ImmutableOpenMap<String, DiskUsage> usages = usagesBuilder.build();
ImmutableOpenMap.Builder<String, Long> shardSizesBuilder = ImmutableOpenMap.builder();
// 14 bytes
shardSizesBuilder.put("[test][0][p]", 14L);
shardSizesBuilder.put("[test][0][r]", 14L);
// 1 bytes
shardSizesBuilder.put("[test2][0][p]", 1L);
shardSizesBuilder.put("[test2][0][r]", 1L);
ImmutableOpenMap<String, Long> shardSizes = shardSizesBuilder.build();
final ClusterInfo clusterInfo = new DevNullClusterInfo(usages, usages, shardSizes);
DiskThresholdDecider decider = makeDecider(diskSettings);
AllocationDeciders deciders = new AllocationDeciders(Settings.EMPTY, new HashSet<>(Arrays.asList(new SameShardAllocationDecider(Settings.EMPTY, new ClusterSettings(Settings.EMPTY, ClusterSettings.BUILT_IN_CLUSTER_SETTINGS)), decider)));
ClusterInfoService cis = new ClusterInfoService() {
@Override
public ClusterInfo getClusterInfo() {
logger.info("--> calling fake getClusterInfo");
return clusterInfo;
}
@Override
public void addListener(Listener listener) {
// noop
}
};
AllocationService strategy = new AllocationService(Settings.builder().put("cluster.routing.allocation.node_concurrent_recoveries", 10).put(ClusterRebalanceAllocationDecider.CLUSTER_ROUTING_ALLOCATION_ALLOW_REBALANCE_SETTING.getKey(), "always").put("cluster.routing.allocation.cluster_concurrent_rebalance", -1).build(), deciders, new TestGatewayAllocator(), new BalancedShardsAllocator(Settings.EMPTY), cis);
MetaData metaData = MetaData.builder().put(IndexMetaData.builder("test").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(1)).put(IndexMetaData.builder("test2").settings(settings(Version.CURRENT)).numberOfShards(1).numberOfReplicas(1)).build();
RoutingTable initialRoutingTable = RoutingTable.builder().addAsNew(metaData.index("test")).addAsNew(metaData.index("test2")).build();
ClusterState clusterState = ClusterState.builder(ClusterName.CLUSTER_NAME_SETTING.getDefault(Settings.EMPTY)).metaData(metaData).routingTable(initialRoutingTable).build();
logger.info("--> adding two nodes");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder().add(newNode("node1")).add(newNode("node2"))).build();
clusterState = strategy.reroute(clusterState, "reroute");
logShardStates(clusterState);
// shards should be initializing
assertThat(clusterState.getRoutingNodes().shardsWithState(INITIALIZING).size(), equalTo(4));
logger.info("--> start the shards");
clusterState = strategy.applyStartedShards(clusterState, clusterState.getRoutingNodes().shardsWithState(INITIALIZING));
logShardStates(clusterState);
// Assert that we're able to start the primary and replicas
assertThat(clusterState.getRoutingNodes().shardsWithState(ShardRoutingState.STARTED).size(), equalTo(4));
logger.info("--> adding node3");
clusterState = ClusterState.builder(clusterState).nodes(DiscoveryNodes.builder(clusterState.nodes()).add(newNode("node3"))).build();
AllocationCommand relocate1 = new MoveAllocationCommand("test", 0, "node2", "node3");
AllocationCommands cmds = new AllocationCommands(relocate1);
clusterState = strategy.reroute(clusterState, cmds, false, false).getClusterState();
logShardStates(clusterState);
AllocationCommand relocate2 = new MoveAllocationCommand("test2", 0, "node2", "node3");
cmds = new AllocationCommands(relocate2);
try {
// The shard for the "test" index is already being relocated to
// node3, which will put it over the low watermark when it
// completes, with shard relocations taken into account this should
// throw an exception about not being able to complete
strategy.reroute(clusterState, cmds, false, false);
fail("should not have been able to reroute the shard");
} catch (IllegalArgumentException e) {
assertThat("can't be allocated because there isn't enough room: " + e.getMessage(), e.getMessage(), containsString("the node is above the low watermark cluster setting " + "[cluster.routing.allocation.disk.watermark.low=0.7], using more disk space than the maximum " + "allowed [70.0%], actual free: [26.0%]"));
}
}
use of org.elasticsearch.cluster.ClusterInfoService 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));
}
}
Aggregations