use of org.opensearch.cluster.action.shard.ShardStateAction in project OpenSearch by opensearch-project.
the class TransportVerifyShardBeforeCloseActionTests method setUp.
@Override
@Before
public void setUp() throws Exception {
super.setUp();
indexShard = mock(IndexShard.class);
when(indexShard.getActiveOperationsCount()).thenReturn(IndexShard.OPERATIONS_BLOCKED);
final ShardId shardId = new ShardId("index", "_na_", randomIntBetween(0, 3));
when(indexShard.shardId()).thenReturn(shardId);
clusterService = createClusterService(threadPool);
clusterBlock = MetadataIndexStateService.createIndexClosingBlock();
setState(clusterService, new ClusterState.Builder(clusterService.state()).blocks(ClusterBlocks.builder().blocks(clusterService.state().blocks()).addIndexBlock("index", clusterBlock).build()).build());
transport = new CapturingTransport();
TransportService transportService = transport.createTransportService(Settings.EMPTY, threadPool, TransportService.NOOP_TRANSPORT_INTERCEPTOR, x -> clusterService.localNode(), null, Collections.emptySet());
transportService.start();
transportService.acceptIncomingRequests();
ShardStateAction shardStateAction = new ShardStateAction(clusterService, transportService, null, null, threadPool);
action = new TransportVerifyShardBeforeCloseAction(Settings.EMPTY, transportService, clusterService, mock(IndicesService.class), mock(ThreadPool.class), shardStateAction, mock(ActionFilters.class));
}
use of org.opensearch.cluster.action.shard.ShardStateAction in project OpenSearch by opensearch-project.
the class TransportWriteActionTests method testReplicaProxy.
public void testReplicaProxy() throws InterruptedException, ExecutionException {
CapturingTransport transport = new CapturingTransport();
TransportService transportService = transport.createTransportService(clusterService.getSettings(), threadPool, TransportService.NOOP_TRANSPORT_INTERCEPTOR, x -> clusterService.localNode(), null, Collections.emptySet());
transportService.start();
transportService.acceptIncomingRequests();
ShardStateAction shardStateAction = new ShardStateAction(clusterService, transportService, null, null, threadPool);
TestAction action = new TestAction(Settings.EMPTY, "internal:testAction", transportService, clusterService, shardStateAction, threadPool);
final String index = "test";
final ShardId shardId = new ShardId(index, "_na_", 0);
ClusterState state = ClusterStateCreationUtils.stateWithActivePrimary(index, true, 1 + randomInt(3), randomInt(2));
logger.info("using state: {}", state);
ClusterServiceUtils.setState(clusterService, state);
final long primaryTerm = state.metadata().index(index).primaryTerm(0);
ReplicationOperation.Replicas<TestRequest> proxy = action.newReplicasProxy();
// check that at unknown node fails
PlainActionFuture<ReplicaResponse> listener = new PlainActionFuture<>();
ShardRoutingState routingState = randomFrom(ShardRoutingState.INITIALIZING, ShardRoutingState.STARTED, ShardRoutingState.RELOCATING);
proxy.performOn(TestShardRouting.newShardRouting(shardId, "NOT THERE", routingState == ShardRoutingState.RELOCATING ? state.nodes().iterator().next().getId() : null, false, routingState), new TestRequest(), primaryTerm, randomNonNegativeLong(), randomNonNegativeLong(), listener);
assertTrue(listener.isDone());
assertListenerThrows("non existent node should throw a NoNodeAvailableException", listener, NoNodeAvailableException.class);
final IndexShardRoutingTable shardRoutings = state.routingTable().shardRoutingTable(shardId);
final ShardRouting replica = randomFrom(shardRoutings.replicaShards().stream().filter(ShardRouting::assignedToNode).collect(Collectors.toList()));
listener = new PlainActionFuture<>();
proxy.performOn(replica, new TestRequest(), primaryTerm, randomNonNegativeLong(), randomNonNegativeLong(), listener);
assertFalse(listener.isDone());
CapturingTransport.CapturedRequest[] captures = transport.getCapturedRequestsAndClear();
assertThat(captures, arrayWithSize(1));
if (randomBoolean()) {
final TransportReplicationAction.ReplicaResponse response = new TransportReplicationAction.ReplicaResponse(randomLong(), randomLong());
transport.handleResponse(captures[0].requestId, response);
assertTrue(listener.isDone());
assertThat(listener.get(), equalTo(response));
} else if (randomBoolean()) {
transport.handleRemoteError(captures[0].requestId, new OpenSearchException("simulated"));
assertTrue(listener.isDone());
assertListenerThrows("listener should reflect remote error", listener, OpenSearchException.class);
} else {
transport.handleError(captures[0].requestId, new TransportException("simulated"));
assertTrue(listener.isDone());
assertListenerThrows("listener should reflect remote error", listener, TransportException.class);
}
AtomicReference<Object> failure = new AtomicReference<>();
AtomicBoolean success = new AtomicBoolean();
proxy.failShardIfNeeded(replica, primaryTerm, "test", new OpenSearchException("simulated"), ActionListener.wrap(r -> success.set(true), failure::set));
CapturingTransport.CapturedRequest[] shardFailedRequests = transport.getCapturedRequestsAndClear();
// A write replication action proxy should fail the shard
assertEquals(1, shardFailedRequests.length);
CapturingTransport.CapturedRequest shardFailedRequest = shardFailedRequests[0];
ShardStateAction.FailedShardEntry shardEntry = (ShardStateAction.FailedShardEntry) shardFailedRequest.request;
// the shard the request was sent to and the shard to be failed should be the same
assertEquals(shardEntry.getShardId(), replica.shardId());
assertEquals(shardEntry.getAllocationId(), replica.allocationId().getId());
if (randomBoolean()) {
// simulate success
transport.handleResponse(shardFailedRequest.requestId, TransportResponse.Empty.INSTANCE);
assertTrue(success.get());
assertNull(failure.get());
} else if (randomBoolean()) {
// simulate the primary has been demoted
transport.handleRemoteError(shardFailedRequest.requestId, new ShardStateAction.NoLongerPrimaryShardException(replica.shardId(), "shard-failed-test"));
assertFalse(success.get());
assertNotNull(failure.get());
} else {
// simulated a node closing exception
transport.handleRemoteError(shardFailedRequest.requestId, new NodeClosedException(state.nodes().getLocalNode()));
assertFalse(success.get());
assertNotNull(failure.get());
}
}
use of org.opensearch.cluster.action.shard.ShardStateAction in project OpenSearch by opensearch-project.
the class TransportReplicationActionTests method setUp.
@Override
@Before
public void setUp() throws Exception {
super.setUp();
forceExecute = randomBoolean();
transport = new CapturingTransport();
clusterService = createClusterService(threadPool);
transportService = transport.createTransportService(clusterService.getSettings(), threadPool, TransportService.NOOP_TRANSPORT_INTERCEPTOR, x -> clusterService.localNode(), null, Collections.emptySet());
transportService.start();
transportService.acceptIncomingRequests();
shardStateAction = new ShardStateAction(clusterService, transportService, null, null, threadPool);
action = new TestAction(Settings.EMPTY, "internal:testAction", transportService, clusterService, shardStateAction, threadPool);
}
use of org.opensearch.cluster.action.shard.ShardStateAction in project OpenSearch by opensearch-project.
the class ClusterDisruptionIT method testSendingShardFailure.
// simulate handling of sending shard failure during an isolation
public void testSendingShardFailure() throws Exception {
List<String> nodes = startCluster(3);
String masterNode = internalCluster().getMasterName();
List<String> nonMasterNodes = nodes.stream().filter(node -> !node.equals(masterNode)).collect(Collectors.toList());
String nonMasterNode = randomFrom(nonMasterNodes);
assertAcked(prepareCreate("test").setSettings(Settings.builder().put(IndexMetadata.SETTING_NUMBER_OF_SHARDS, 3).put(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, 2)));
ensureGreen();
String nonMasterNodeId = internalCluster().clusterService(nonMasterNode).localNode().getId();
// fail a random shard
ShardRouting failedShard = randomFrom(clusterService().state().getRoutingNodes().node(nonMasterNodeId).shardsWithState(ShardRoutingState.STARTED));
ShardStateAction service = internalCluster().getInstance(ShardStateAction.class, nonMasterNode);
CountDownLatch latch = new CountDownLatch(1);
AtomicBoolean success = new AtomicBoolean();
String isolatedNode = randomBoolean() ? masterNode : nonMasterNode;
TwoPartitions partitions = isolateNode(isolatedNode);
// we cannot use the NetworkUnresponsive disruption type here as it will swallow the "shard failed" request, calling neither
// onSuccess nor onFailure on the provided listener.
NetworkDisruption networkDisruption = new NetworkDisruption(partitions, NetworkDisruption.DISCONNECT);
setDisruptionScheme(networkDisruption);
networkDisruption.startDisrupting();
service.localShardFailed(failedShard, "simulated", new CorruptIndexException("simulated", (String) null), new ActionListener<Void>() {
@Override
public void onResponse(final Void aVoid) {
success.set(true);
latch.countDown();
}
@Override
public void onFailure(Exception e) {
success.set(false);
latch.countDown();
assert false;
}
});
if (isolatedNode.equals(nonMasterNode)) {
assertNoMaster(nonMasterNode);
} else {
ensureStableCluster(2, nonMasterNode);
}
// heal the partition
networkDisruption.removeAndEnsureHealthy(internalCluster());
// the cluster should stabilize
ensureStableCluster(3);
latch.await();
// the listener should be notified
assertTrue(success.get());
// the failed shard should be gone
List<ShardRouting> shards = clusterService().state().getRoutingTable().allShards("test");
for (ShardRouting shard : shards) {
assertThat(shard.allocationId(), not(equalTo(failedShard.allocationId())));
}
}
use of org.opensearch.cluster.action.shard.ShardStateAction in project OpenSearch by opensearch-project.
the class TransportResyncReplicationActionTests method testResyncDoesNotBlockOnPrimaryAction.
public void testResyncDoesNotBlockOnPrimaryAction() throws Exception {
try (ClusterService clusterService = createClusterService(threadPool)) {
final String indexName = randomAlphaOfLength(5);
setState(clusterService, state(indexName, true, ShardRoutingState.STARTED));
setState(clusterService, ClusterState.builder(clusterService.state()).blocks(ClusterBlocks.builder().addGlobalBlock(NoMasterBlockService.NO_MASTER_BLOCK_ALL).addIndexBlock(indexName, IndexMetadata.INDEX_WRITE_BLOCK)));
try (MockNioTransport transport = new MockNioTransport(Settings.EMPTY, Version.CURRENT, threadPool, new NetworkService(emptyList()), PageCacheRecycler.NON_RECYCLING_INSTANCE, new NamedWriteableRegistry(emptyList()), new NoneCircuitBreakerService())) {
final MockTransportService transportService = new MockTransportService(Settings.EMPTY, transport, threadPool, NOOP_TRANSPORT_INTERCEPTOR, x -> clusterService.localNode(), null, Collections.emptySet());
transportService.start();
transportService.acceptIncomingRequests();
final ShardStateAction shardStateAction = new ShardStateAction(clusterService, transportService, null, null, threadPool);
final IndexMetadata indexMetadata = clusterService.state().metadata().index(indexName);
final Index index = indexMetadata.getIndex();
final ShardId shardId = new ShardId(index, 0);
final IndexShardRoutingTable shardRoutingTable = clusterService.state().routingTable().shardRoutingTable(shardId);
final ShardRouting primaryShardRouting = clusterService.state().routingTable().shardRoutingTable(shardId).primaryShard();
final String allocationId = primaryShardRouting.allocationId().getId();
final long primaryTerm = indexMetadata.primaryTerm(shardId.id());
final AtomicInteger acquiredPermits = new AtomicInteger();
final IndexShard indexShard = mock(IndexShard.class);
when(indexShard.indexSettings()).thenReturn(new IndexSettings(indexMetadata, Settings.EMPTY));
when(indexShard.shardId()).thenReturn(shardId);
when(indexShard.routingEntry()).thenReturn(primaryShardRouting);
when(indexShard.getPendingPrimaryTerm()).thenReturn(primaryTerm);
when(indexShard.getOperationPrimaryTerm()).thenReturn(primaryTerm);
when(indexShard.getActiveOperationsCount()).then(i -> acquiredPermits.get());
doAnswer(invocation -> {
ActionListener<Releasable> callback = (ActionListener<Releasable>) invocation.getArguments()[0];
acquiredPermits.incrementAndGet();
callback.onResponse(acquiredPermits::decrementAndGet);
return null;
}).when(indexShard).acquirePrimaryOperationPermit(any(ActionListener.class), anyString(), any(), eq(true));
when(indexShard.getReplicationGroup()).thenReturn(new ReplicationGroup(shardRoutingTable, clusterService.state().metadata().index(index).inSyncAllocationIds(shardId.id()), shardRoutingTable.getAllAllocationIds(), 0));
final IndexService indexService = mock(IndexService.class);
when(indexService.getShard(eq(shardId.id()))).thenReturn(indexShard);
final IndicesService indexServices = mock(IndicesService.class);
when(indexServices.indexServiceSafe(eq(index))).thenReturn(indexService);
final TransportResyncReplicationAction action = new TransportResyncReplicationAction(Settings.EMPTY, transportService, clusterService, indexServices, threadPool, shardStateAction, new ActionFilters(new HashSet<>()), new IndexingPressureService(Settings.EMPTY, clusterService), new SystemIndices(emptyMap()));
assertThat(action.globalBlockLevel(), nullValue());
assertThat(action.indexBlockLevel(), nullValue());
final Task task = mock(Task.class);
when(task.getId()).thenReturn(randomNonNegativeLong());
final byte[] bytes = "{}".getBytes(Charset.forName("UTF-8"));
final ResyncReplicationRequest request = new ResyncReplicationRequest(shardId, 42L, 100, new Translog.Operation[] { new Translog.Index("type", "id", 0, primaryTerm, 0L, bytes, null, -1) });
final PlainActionFuture<ResyncReplicationResponse> listener = new PlainActionFuture<>();
action.sync(request, task, allocationId, primaryTerm, listener);
assertThat(listener.get().getShardInfo().getFailed(), equalTo(0));
assertThat(listener.isDone(), is(true));
}
}
}
Aggregations