Examples with ReplicationGroup org.elasticsearch.index.shard.ReplicationGroup used on opensource projects

Example 1 with ReplicationGroup

use of org.elasticsearch.index.shard.ReplicationGroup in project crate by crate.

the class ReplicationOperationTests method testNoLongerPrimary.

@Test
public void testNoLongerPrimary() throws Exception {
    final String index = "test";
    final ShardId shardId = new ShardId(index, "_na_", 0);
    ClusterState initialState = stateWithActivePrimary(index, true, 1 + randomInt(2), randomInt(2));
    IndexMetadata indexMetadata = initialState.getMetadata().index(index);
    final long primaryTerm = indexMetadata.primaryTerm(0);
    final IndexShardRoutingTable indexShardRoutingTable = initialState.getRoutingTable().shardRoutingTable(shardId);
    ShardRouting primaryShard = indexShardRoutingTable.primaryShard();
    if (primaryShard.relocating() && randomBoolean()) {
        // simulate execution of the replication phase on the relocation target node after relocation source was marked as relocated
        initialState = ClusterState.builder(initialState).nodes(DiscoveryNodes.builder(initialState.nodes()).localNodeId(primaryShard.relocatingNodeId())).build();
        primaryShard = primaryShard.getTargetRelocatingShard();
    }
    // add an in-sync allocation id that doesn't have a corresponding routing entry
    final Set<String> staleAllocationIds = Set.of(randomAlphaOfLength(10));
    final Set<String> inSyncAllocationIds = Sets.union(indexMetadata.inSyncAllocationIds(0), staleAllocationIds);
    final Set<String> trackedShards = new HashSet<>();
    addTrackingInfo(indexShardRoutingTable, primaryShard, trackedShards, new HashSet<>());
    trackedShards.addAll(staleAllocationIds);
    final ReplicationGroup replicationGroup = new ReplicationGroup(indexShardRoutingTable, inSyncAllocationIds, trackedShards);
    final Set<ShardRouting> expectedReplicas = getExpectedReplicas(shardId, initialState, trackedShards);
    final Map<ShardRouting, Exception> expectedFailures = new HashMap<>();
    if (expectedReplicas.isEmpty()) {
        return;
    }
    final ShardRouting failedReplica = randomFrom(new ArrayList<>(expectedReplicas));
    expectedFailures.put(failedReplica, new CorruptIndexException("simulated", (String) null));
    Request request = new Request(shardId);
    PlainActionFuture<TestPrimary.Result> listener = new PlainActionFuture<>();
    final boolean testPrimaryDemotedOnStaleShardCopies = randomBoolean();
    final Exception shardActionFailure;
    if (randomBoolean()) {
        shardActionFailure = new NodeClosedException(new DiscoveryNode("foo", buildNewFakeTransportAddress(), Version.CURRENT));
    } else if (randomBoolean()) {
        DiscoveryNode node = new DiscoveryNode("foo", buildNewFakeTransportAddress(), Version.CURRENT);
        shardActionFailure = new SendRequestTransportException(node, ShardStateAction.SHARD_FAILED_ACTION_NAME, new NodeClosedException(node));
    } else {
        shardActionFailure = new ShardStateAction.NoLongerPrimaryShardException(failedReplica.shardId(), "the king is dead");
    }
    final TestReplicaProxy replicasProxy = new TestReplicaProxy(expectedFailures) {

        @Override
        public void failShardIfNeeded(ShardRouting replica, long primaryTerm, String message, Exception exception, ActionListener<Void> shardActionListener) {
            if (testPrimaryDemotedOnStaleShardCopies) {
                super.failShardIfNeeded(replica, primaryTerm, message, exception, shardActionListener);
            } else {
                assertThat(replica, equalTo(failedReplica));
                shardActionListener.onFailure(shardActionFailure);
            }
        }

        @Override
        public void markShardCopyAsStaleIfNeeded(ShardId shardId, String allocationId, long primaryTerm, ActionListener<Void> shardActionListener) {
            if (testPrimaryDemotedOnStaleShardCopies) {
                shardActionListener.onFailure(shardActionFailure);
            } else {
                super.markShardCopyAsStaleIfNeeded(shardId, allocationId, primaryTerm, shardActionListener);
            }
        }
    };
    AtomicBoolean primaryFailed = new AtomicBoolean();
    final TestPrimary primary = new TestPrimary(primaryShard, () -> replicationGroup) {

        @Override
        public void failShard(String message, Exception exception) {
            assertTrue(primaryFailed.compareAndSet(false, true));
        }
    };
    final TestReplicationOperation op = new TestReplicationOperation(request, primary, listener, replicasProxy, primaryTerm);
    op.execute();
    assertThat("request was not processed on primary", request.processedOnPrimary.get(), equalTo(true));
    assertTrue("listener is not marked as done", listener.isDone());
    if (shardActionFailure instanceof ShardStateAction.NoLongerPrimaryShardException) {
        assertTrue(primaryFailed.get());
    } else {
        assertFalse(primaryFailed.get());
    }
    assertListenerThrows("should throw exception to trigger retry", listener, ReplicationOperation.RetryOnPrimaryException.class);
}

Example 2 with ReplicationGroup

use of org.elasticsearch.index.shard.ReplicationGroup in project crate by crate.

the class ReplicationOperationTests method testWaitForActiveShards.

@Test
public void testWaitForActiveShards() throws Exception {
    final String index = "test";
    final ShardId shardId = new ShardId(index, "_na_", 0);
    final int assignedReplicas = randomInt(2);
    final int unassignedReplicas = randomInt(2);
    final int totalShards = 1 + assignedReplicas + unassignedReplicas;
    final int activeShardCount = randomIntBetween(0, totalShards);
    Request request = new Request(shardId).waitForActiveShards(activeShardCount == totalShards ? ActiveShardCount.ALL : ActiveShardCount.from(activeShardCount));
    final boolean passesActiveShardCheck = activeShardCount <= assignedReplicas + 1;
    ShardRoutingState[] replicaStates = new ShardRoutingState[assignedReplicas + unassignedReplicas];
    for (int i = 0; i < assignedReplicas; i++) {
        replicaStates[i] = randomFrom(ShardRoutingState.STARTED, ShardRoutingState.RELOCATING);
    }
    for (int i = assignedReplicas; i < replicaStates.length; i++) {
        replicaStates[i] = ShardRoutingState.UNASSIGNED;
    }
    final ClusterState state = state(index, true, ShardRoutingState.STARTED, replicaStates);
    logger.debug("using active shard count of [{}], assigned shards [{}], total shards [{}]." + " expecting op to [{}]. using state: \n{}", request.waitForActiveShards(), 1 + assignedReplicas, 1 + assignedReplicas + unassignedReplicas, passesActiveShardCheck ? "succeed" : "retry", state);
    final long primaryTerm = state.metadata().index(index).primaryTerm(shardId.id());
    final IndexShardRoutingTable shardRoutingTable = state.routingTable().index(index).shard(shardId.id());
    final Set<String> inSyncAllocationIds = state.metadata().index(index).inSyncAllocationIds(0);
    Set<String> trackedShards = new HashSet<>();
    addTrackingInfo(shardRoutingTable, null, trackedShards, new HashSet<>());
    final ReplicationGroup initialReplicationGroup = new ReplicationGroup(shardRoutingTable, inSyncAllocationIds, trackedShards);
    PlainActionFuture<TestPrimary.Result> listener = new PlainActionFuture<>();
    final ShardRouting primaryShard = shardRoutingTable.primaryShard();
    final TestReplicationOperation op = new TestReplicationOperation(request, new TestPrimary(primaryShard, () -> initialReplicationGroup), listener, new TestReplicaProxy(), logger, "test", primaryTerm);
    if (passesActiveShardCheck) {
        assertThat(op.checkActiveShardCount(), nullValue());
        op.execute();
        assertTrue("operations should have been performed, active shard count is met", request.processedOnPrimary.get());
    } else {
        assertThat(op.checkActiveShardCount(), notNullValue());
        op.execute();
        assertFalse("operations should not have been perform, active shard count is *NOT* met", request.processedOnPrimary.get());
        assertListenerThrows("should throw exception to trigger retry", listener, UnavailableShardsException.class);
    }
}

Example 3 with ReplicationGroup

use of org.elasticsearch.index.shard.ReplicationGroup in project crate by crate.

the class ReplicationOperationTests method testReplication.

@Test
public void testReplication() throws Exception {
    final String index = "test";
    final ShardId shardId = new ShardId(index, "_na_", 0);
    ClusterState initialState = stateWithActivePrimary(index, true, randomInt(5));
    IndexMetadata indexMetadata = initialState.getMetadata().index(index);
    final long primaryTerm = indexMetadata.primaryTerm(0);
    final IndexShardRoutingTable indexShardRoutingTable = initialState.getRoutingTable().shardRoutingTable(shardId);
    ShardRouting primaryShard = indexShardRoutingTable.primaryShard();
    if (primaryShard.relocating() && randomBoolean()) {
        // simulate execution of the replication phase on the relocation target node after relocation source was marked as relocated
        initialState = ClusterState.builder(initialState).nodes(DiscoveryNodes.builder(initialState.nodes()).localNodeId(primaryShard.relocatingNodeId())).build();
        primaryShard = primaryShard.getTargetRelocatingShard();
    }
    // add a few in-sync allocation ids that don't have corresponding routing entries
    final Set<String> staleAllocationIds = Set.of(generateRandomStringArray(4, 10, false));
    final Set<String> inSyncAllocationIds = Sets.union(indexMetadata.inSyncAllocationIds(0), staleAllocationIds);
    final Set<String> trackedShards = new HashSet<>();
    final Set<String> untrackedShards = new HashSet<>();
    addTrackingInfo(indexShardRoutingTable, primaryShard, trackedShards, untrackedShards);
    trackedShards.addAll(staleAllocationIds);
    final ReplicationGroup replicationGroup = new ReplicationGroup(indexShardRoutingTable, inSyncAllocationIds, trackedShards);
    final Set<ShardRouting> expectedReplicas = getExpectedReplicas(shardId, initialState, trackedShards);
    final Map<ShardRouting, Exception> simulatedFailures = new HashMap<>();
    final Map<ShardRouting, Exception> reportedFailures = new HashMap<>();
    for (ShardRouting replica : expectedReplicas) {
        if (randomBoolean()) {
            Exception t;
            boolean criticalFailure = randomBoolean();
            if (criticalFailure) {
                t = new CorruptIndexException("simulated", (String) null);
                reportedFailures.put(replica, t);
            } else {
                t = new IndexShardNotStartedException(shardId, IndexShardState.RECOVERING);
            }
            logger.debug("--> simulating failure on {} with [{}]", replica, t.getClass().getSimpleName());
            simulatedFailures.put(replica, t);
        }
    }
    Request request = new Request(shardId);
    PlainActionFuture<TestPrimary.Result> listener = new PlainActionFuture<>();
    final TestReplicaProxy replicasProxy = new TestReplicaProxy(simulatedFailures);
    final TestPrimary primary = new TestPrimary(primaryShard, () -> replicationGroup);
    final TestReplicationOperation op = new TestReplicationOperation(request, primary, listener, replicasProxy, primaryTerm);
    op.execute();
    assertThat("request was not processed on primary", request.processedOnPrimary.get(), equalTo(true));
    assertThat(request.processedOnReplicas, equalTo(expectedReplicas));
    assertThat(replicasProxy.failedReplicas, equalTo(simulatedFailures.keySet()));
    assertThat(replicasProxy.markedAsStaleCopies, equalTo(staleAllocationIds));
    assertThat("post replication operations not run on primary", request.runPostReplicationActionsOnPrimary.get(), equalTo(true));
    assertTrue("listener is not marked as done", listener.isDone());
    ShardInfo shardInfo = listener.actionGet().getShardInfo();
    assertThat(shardInfo.getFailed(), equalTo(reportedFailures.size()));
    assertThat(shardInfo.getFailures(), arrayWithSize(reportedFailures.size()));
    assertThat(shardInfo.getSuccessful(), equalTo(1 + expectedReplicas.size() - simulatedFailures.size()));
    final List<ShardRouting> unassignedShards = indexShardRoutingTable.shardsWithState(ShardRoutingState.UNASSIGNED);
    final int totalShards = 1 + expectedReplicas.size() + unassignedShards.size() + untrackedShards.size();
    assertThat(replicationGroup.toString(), shardInfo.getTotal(), equalTo(totalShards));
    assertThat(primary.knownLocalCheckpoints.remove(primaryShard.allocationId().getId()), equalTo(primary.localCheckpoint));
    assertThat(primary.knownLocalCheckpoints, equalTo(replicasProxy.generatedLocalCheckpoints));
    assertThat(primary.knownGlobalCheckpoints.remove(primaryShard.allocationId().getId()), equalTo(primary.globalCheckpoint));
    assertThat(primary.knownGlobalCheckpoints, equalTo(replicasProxy.generatedGlobalCheckpoints));
}

Example 4 with ReplicationGroup

use of org.elasticsearch.index.shard.ReplicationGroup in project crate by crate.

the class ReplicationOperation method handlePrimaryResult.

private void handlePrimaryResult(final PrimaryResultT primaryResult) {
    this.primaryResult = primaryResult;
    final ReplicaRequest replicaRequest = primaryResult.replicaRequest();
    if (replicaRequest != null) {
        if (logger.isTraceEnabled()) {
            logger.trace("[{}] op [{}] completed on primary for request [{}]", primary.routingEntry().shardId(), opType, request);
        }
        // we have to get the replication group after successfully indexing into the primary in order to honour recovery semantics.
        // we have to make sure that every operation indexed into the primary after recovery start will also be replicated
        // to the recovery target. If we used an old replication group, we may miss a recovery that has started since then.
        // we also have to make sure to get the global checkpoint before the replication group, to ensure that the global checkpoint
        // is valid for this replication group. If we would sample in the reverse, the global checkpoint might be based on a subset
        // of the sampled replication group, and advanced further than what the given replication group would allow it to.
        // This would entail that some shards could learn about a global checkpoint that would be higher than its local checkpoint.
        final long globalCheckpoint = primary.computedGlobalCheckpoint();
        // we have to capture the max_seq_no_of_updates after this request was completed on the primary to make sure the value of
        // max_seq_no_of_updates on replica when this request is executed is at least the value on the primary when it was executed on.
        final long maxSeqNoOfUpdatesOrDeletes = primary.maxSeqNoOfUpdatesOrDeletes();
        assert maxSeqNoOfUpdatesOrDeletes != SequenceNumbers.UNASSIGNED_SEQ_NO : "seqno_of_updates still uninitialized";
        final ReplicationGroup replicationGroup = primary.getReplicationGroup();
        markUnavailableShardsAsStale(replicaRequest, replicationGroup);
        performOnReplicas(replicaRequest, globalCheckpoint, maxSeqNoOfUpdatesOrDeletes, replicationGroup);
    }
    primaryResult.runPostReplicationActions(new ActionListener<>() {

        @Override
        public void onResponse(Void aVoid) {
            successfulShards.incrementAndGet();
            try {
                updateCheckPoints(primary.routingEntry(), primary::localCheckpoint, primary::globalCheckpoint);
            } finally {
                decPendingAndFinishIfNeeded();
            }
        }

        @Override
        public void onFailure(Exception e) {
            logger.trace("[{}] op [{}] post replication actions failed for [{}]", primary.routingEntry().shardId(), opType, request);
            // TODO: fail shard? This will otherwise have the local / global checkpoint info lagging, or possibly have replicas
            // go out of sync with the primary
            finishAsFailed(e);
        }
    });
}

Example 5 with ReplicationGroup

use of org.elasticsearch.index.shard.ReplicationGroup in project crate by crate.

the class TransportVerifyShardBeforeCloseActionTests method testUnavailableShardsMarkedAsStale.

@Test
public void testUnavailableShardsMarkedAsStale() throws Exception {
    String index = "test";
    ShardId shardId = new ShardId(index, "_na_", 0);
    int nbReplicas = randomIntBetween(1, 10);
    ShardRoutingState[] replicaStates = new ShardRoutingState[nbReplicas];
    Arrays.fill(replicaStates, ShardRoutingState.STARTED);
    final ClusterState clusterState = state(index, true, ShardRoutingState.STARTED, replicaStates);
    setState(clusterService, clusterState);
    IndexShardRoutingTable shardRoutingTable = clusterState.routingTable().index(index).shard(shardId.id());
    IndexMetadata indexMetaData = clusterState.getMetadata().index(index);
    ShardRouting primaryRouting = shardRoutingTable.primaryShard();
    long primaryTerm = indexMetaData.primaryTerm(0);
    Set<String> inSyncAllocationIds = indexMetaData.inSyncAllocationIds(0);
    Set<String> trackedShards = shardRoutingTable.getAllAllocationIds();
    List<ShardRouting> unavailableShards = randomSubsetOf(randomIntBetween(1, nbReplicas), shardRoutingTable.replicaShards());
    IndexShardRoutingTable.Builder shardRoutingTableBuilder = new IndexShardRoutingTable.Builder(shardRoutingTable);
    unavailableShards.forEach(shardRoutingTableBuilder::removeShard);
    shardRoutingTable = shardRoutingTableBuilder.build();
    ReplicationGroup replicationGroup = new ReplicationGroup(shardRoutingTable, inSyncAllocationIds, trackedShards);
    assertThat(replicationGroup.getUnavailableInSyncShards().size(), greaterThan(0));
    PlainActionFuture<PrimaryResult> listener = new PlainActionFuture<>();
    TransportVerifyShardBeforeCloseAction.ShardRequest request = new TransportVerifyShardBeforeCloseAction.ShardRequest(shardId, false, clusterBlock);
    ReplicationOperation.Replicas<TransportVerifyShardBeforeCloseAction.ShardRequest> proxy = action.newReplicasProxy();
    ReplicationOperation<TransportVerifyShardBeforeCloseAction.ShardRequest, TransportVerifyShardBeforeCloseAction.ShardRequest, PrimaryResult> operation = new ReplicationOperation<>(request, createPrimary(primaryRouting, replicationGroup), listener, proxy, logger, "test", primaryTerm);
    operation.execute();
    CapturingTransport.CapturedRequest[] capturedRequests = transport.getCapturedRequestsAndClear();
    assertThat(capturedRequests.length, equalTo(nbReplicas));
    for (CapturingTransport.CapturedRequest capturedRequest : capturedRequests) {
        String actionName = capturedRequest.action;
        if (actionName.startsWith(ShardStateAction.SHARD_FAILED_ACTION_NAME)) {
            assertThat(capturedRequest.request, instanceOf(ShardStateAction.FailedShardEntry.class));
            String allocationId = ((ShardStateAction.FailedShardEntry) capturedRequest.request).getAllocationId();
            assertTrue(unavailableShards.stream().anyMatch(shardRouting -> shardRouting.allocationId().getId().equals(allocationId)));
            transport.handleResponse(capturedRequest.requestId, TransportResponse.Empty.INSTANCE);
        } else if (actionName.startsWith(TransportVerifyShardBeforeCloseAction.NAME)) {
            assertThat(capturedRequest.request, instanceOf(ConcreteShardRequest.class));
            String allocationId = ((ConcreteShardRequest) capturedRequest.request).getTargetAllocationID();
            assertFalse(unavailableShards.stream().anyMatch(shardRouting -> shardRouting.allocationId().getId().equals(allocationId)));
            assertTrue(inSyncAllocationIds.stream().anyMatch(inSyncAllocationId -> inSyncAllocationId.equals(allocationId)));
            transport.handleResponse(capturedRequest.requestId, new TransportReplicationAction.ReplicaResponse(0L, 0L));
        } else {
            fail("Test does not support action " + capturedRequest.action);
        }
    }
    ReplicationResponse.ShardInfo shardInfo = listener.get().getShardInfo();
    assertThat(shardInfo.getFailed(), equalTo(0));
    assertThat(shardInfo.getFailures(), arrayWithSize(0));
    assertThat(shardInfo.getSuccessful(), equalTo(1 + nbReplicas - unavailableShards.size()));
}