Examples with UnassignedInfo org.opensearch.cluster.routing.UnassignedInfo used on opensource projects

Example 26 with UnassignedInfo

use of org.opensearch.cluster.routing.UnassignedInfo in project OpenSearch by opensearch-project.

the class ClusterAllocationExplainIT method testAssignedReplicaOnSpecificNode.

public void testAssignedReplicaOnSpecificNode() throws Exception {
    logger.info("--> starting 3 nodes");
    List<String> nodes = internalCluster().startNodes(3);
    String excludedNode = nodes.get(randomIntBetween(0, 2));
    prepareIndex(randomIndexState(), 1, 2, Settings.builder().put("index.routing.allocation.exclude._name", excludedNode).build(), ActiveShardCount.from(2));
    boolean includeYesDecisions = randomBoolean();
    boolean includeDiskInfo = randomBoolean();
    ClusterAllocationExplanation explanation = runExplain(false, replicaNode().getId(), includeYesDecisions, includeDiskInfo);
    ShardId shardId = explanation.getShard();
    boolean isPrimary = explanation.isPrimary();
    ShardRoutingState shardRoutingState = explanation.getShardState();
    DiscoveryNode currentNode = explanation.getCurrentNode();
    UnassignedInfo unassignedInfo = explanation.getUnassignedInfo();
    ClusterInfo clusterInfo = explanation.getClusterInfo();
    AllocateUnassignedDecision allocateDecision = explanation.getShardAllocationDecision().getAllocateDecision();
    MoveDecision moveDecision = explanation.getShardAllocationDecision().getMoveDecision();
    // verify shard info
    assertEquals("idx", shardId.getIndexName());
    assertEquals(0, shardId.getId());
    assertFalse(isPrimary);
    // verify current node info
    assertEquals(ShardRoutingState.STARTED, shardRoutingState);
    assertNotNull(currentNode);
    assertEquals(replicaNode().getName(), currentNode.getName());
    // verify unassigned info
    assertNull(unassignedInfo);
    // verify cluster info
    verifyClusterInfo(clusterInfo, includeDiskInfo, 3);
    // verify decision objects
    assertFalse(allocateDecision.isDecisionTaken());
    assertTrue(moveDecision.isDecisionTaken());
    assertEquals(AllocationDecision.NO, moveDecision.getAllocationDecision());
    assertEquals("rebalancing is not allowed", moveDecision.getExplanation());
    assertTrue(moveDecision.canRemain());
    assertFalse(moveDecision.forceMove());
    assertFalse(moveDecision.canRebalanceCluster());
    assertNotNull(moveDecision.getCanRemainDecision());
    assertNull(moveDecision.getTargetNode());
    // verifying cluster rebalance decision object
    assertNotNull(moveDecision.getClusterRebalanceDecision());
    assertEquals(Decision.Type.NO, moveDecision.getClusterRebalanceDecision().type());
    // verify node decisions
    assertEquals(2, moveDecision.getNodeDecisions().size());
    for (NodeAllocationResult result : moveDecision.getNodeDecisions()) {
        assertNotNull(result.getNode());
        assertEquals(1, result.getWeightRanking());
        assertEquals(AllocationDecision.NO, result.getNodeDecision());
        if (includeYesDecisions) {
            assertThat(result.getCanAllocateDecision().getDecisions().size(), greaterThan(1));
        } else {
            assertEquals(1, result.getCanAllocateDecision().getDecisions().size());
        }
        for (Decision d : result.getCanAllocateDecision().getDecisions()) {
            if (d.type() == Decision.Type.NO) {
                assertThat(d.label(), is(oneOf("filter", "same_shard")));
            }
            assertNotNull(d.getExplanation());
        }
    }
    // verify JSON output
    try (XContentParser parser = getParser(explanation)) {
        verifyShardInfo(parser, false, includeDiskInfo, ShardRoutingState.STARTED);
        parser.nextToken();
        assertEquals("can_remain_on_current_node", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.YES.toString(), parser.text());
        parser.nextToken();
        assertEquals("can_rebalance_cluster", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.NO.toString(), parser.text());
        parser.nextToken();
        assertEquals("can_rebalance_cluster_decisions", parser.currentName());
        verifyDeciders(parser, AllocationDecision.NO);
        parser.nextToken();
        assertEquals("can_rebalance_to_other_node", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.NO.toString(), parser.text());
        parser.nextToken();
        assertEquals("rebalance_explanation", parser.currentName());
        parser.nextToken();
        assertEquals("rebalancing is not allowed", parser.text());
        verifyNodeDecisions(parser, allNodeDecisions(AllocationDecision.NO, false), includeYesDecisions, false);
        assertEquals(Token.END_OBJECT, parser.nextToken());
    }
}

Example 27 with UnassignedInfo

use of org.opensearch.cluster.routing.UnassignedInfo in project OpenSearch by opensearch-project.

the class ClusterAllocationExplainIT method testCannotAllocateStaleReplicaExplanation.

public void testCannotAllocateStaleReplicaExplanation() throws Exception {
    logger.info("--> starting 3 nodes");
    final String masterNode = internalCluster().startNode();
    // start replica node first, so it's path will be used first when we start a node after
    // stopping all of them at end of test.
    final String replicaNode = internalCluster().startNode();
    Settings replicaDataPathSettings = internalCluster().dataPathSettings(replicaNode);
    final String primaryNode = internalCluster().startNode();
    prepareIndex(IndexMetadata.State.OPEN, 1, 1, Settings.builder().put("index.routing.allocation.include._name", primaryNode).put("index.routing.allocation.exclude._name", masterNode).build(), ActiveShardCount.ONE);
    client().admin().indices().prepareUpdateSettings("idx").setSettings(Settings.builder().put("index.routing.allocation.include._name", (String) null)).get();
    ensureGreen();
    assertThat(replicaNode().getName(), equalTo(replicaNode));
    assertThat(primaryNodeName(), equalTo(primaryNode));
    logger.info("--> stop node with the replica shard");
    internalCluster().stopRandomNode(InternalTestCluster.nameFilter(replicaNode));
    final IndexMetadata.State indexState = randomIndexState();
    if (indexState == IndexMetadata.State.OPEN) {
        logger.info("--> index more data, now the replica is stale");
        indexData();
    } else {
        logger.info("--> close the index, now the replica is stale");
        assertAcked(client().admin().indices().prepareClose("idx"));
        final ClusterHealthResponse clusterHealthResponse = client().admin().cluster().prepareHealth("idx").setTimeout(TimeValue.timeValueSeconds(30)).setWaitForActiveShards(ActiveShardCount.ONE).setWaitForNoInitializingShards(true).setWaitForEvents(Priority.LANGUID).get();
        assertThat(clusterHealthResponse.getStatus().value(), lessThanOrEqualTo(ClusterHealthStatus.YELLOW.value()));
    }
    logger.info("--> stop the node with the primary");
    internalCluster().stopRandomNode(InternalTestCluster.nameFilter(primaryNode));
    logger.info("--> restart the node with the stale replica");
    String restartedNode = internalCluster().startDataOnlyNode(replicaDataPathSettings);
    // wait for the master to finish processing join.
    ensureClusterSizeConsistency();
    // wait until the system has fetched shard data and we know there is no valid shard copy
    assertBusy(() -> {
        ClusterAllocationExplanation explanation = client().admin().cluster().prepareAllocationExplain().setIndex("idx").setShard(0).setPrimary(true).get().getExplanation();
        assertTrue(explanation.getShardAllocationDecision().getAllocateDecision().isDecisionTaken());
        assertEquals(AllocationDecision.NO_VALID_SHARD_COPY, explanation.getShardAllocationDecision().getAllocateDecision().getAllocationDecision());
    });
    boolean includeYesDecisions = randomBoolean();
    boolean includeDiskInfo = randomBoolean();
    ClusterAllocationExplanation explanation = runExplain(true, includeYesDecisions, includeDiskInfo);
    ShardId shardId = explanation.getShard();
    boolean isPrimary = explanation.isPrimary();
    ShardRoutingState shardRoutingState = explanation.getShardState();
    DiscoveryNode currentNode = explanation.getCurrentNode();
    UnassignedInfo unassignedInfo = explanation.getUnassignedInfo();
    AllocateUnassignedDecision allocateDecision = explanation.getShardAllocationDecision().getAllocateDecision();
    MoveDecision moveDecision = explanation.getShardAllocationDecision().getMoveDecision();
    // verify shard info
    assertEquals("idx", shardId.getIndexName());
    assertEquals(0, shardId.getId());
    assertTrue(isPrimary);
    // verify current node info
    assertEquals(ShardRoutingState.UNASSIGNED, shardRoutingState);
    assertNull(currentNode);
    // verify unassigned info
    assertNotNull(unassignedInfo);
    // verify decision object
    assertTrue(allocateDecision.isDecisionTaken());
    assertFalse(moveDecision.isDecisionTaken());
    assertEquals(AllocationDecision.NO_VALID_SHARD_COPY, allocateDecision.getAllocationDecision());
    assertEquals(2, allocateDecision.getNodeDecisions().size());
    for (NodeAllocationResult nodeAllocationResult : allocateDecision.getNodeDecisions()) {
        if (nodeAllocationResult.getNode().getName().equals(restartedNode)) {
            assertNotNull(nodeAllocationResult.getShardStoreInfo());
            assertNotNull(nodeAllocationResult.getShardStoreInfo().getAllocationId());
            assertFalse(nodeAllocationResult.getShardStoreInfo().isInSync());
            assertNull(nodeAllocationResult.getShardStoreInfo().getStoreException());
        } else {
            assertNotNull(nodeAllocationResult.getShardStoreInfo());
            assertNull(nodeAllocationResult.getShardStoreInfo().getAllocationId());
            assertFalse(nodeAllocationResult.getShardStoreInfo().isInSync());
            assertNull(nodeAllocationResult.getShardStoreInfo().getStoreException());
        }
    }
    // verify JSON output
    try (XContentParser parser = getParser(explanation)) {
        verifyShardInfo(parser, true, includeDiskInfo, ShardRoutingState.UNASSIGNED);
        parser.nextToken();
        assertEquals("can_allocate", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.NO_VALID_SHARD_COPY.toString(), parser.text());
        parser.nextToken();
        assertEquals("allocate_explanation", parser.currentName());
        parser.nextToken();
        assertEquals("cannot allocate because all found copies of the shard are either stale or corrupt", parser.text());
        verifyStaleShardCopyNodeDecisions(parser, 2, Collections.singleton(restartedNode));
    }
}

Example 28 with UnassignedInfo

use of org.opensearch.cluster.routing.UnassignedInfo in project OpenSearch by opensearch-project.

the class ClusterAllocationExplainIT method testRebalancingNotAllowed.

public void testRebalancingNotAllowed() throws Exception {
    logger.info("--> starting a single node");
    internalCluster().startNode();
    ensureStableCluster(1);
    prepareIndex(5, 0);
    logger.info("--> disabling rebalancing on the index");
    client().admin().indices().prepareUpdateSettings("idx").setSettings(Settings.builder().put("index.routing.rebalance.enable", "none")).get();
    logger.info("--> starting another node, with rebalancing disabled, it should get no shards");
    internalCluster().startNode();
    ensureStableCluster(2);
    boolean includeYesDecisions = randomBoolean();
    boolean includeDiskInfo = randomBoolean();
    ClusterAllocationExplanation explanation = runExplain(true, includeYesDecisions, includeDiskInfo);
    ShardId shardId = explanation.getShard();
    boolean isPrimary = explanation.isPrimary();
    ShardRoutingState shardRoutingState = explanation.getShardState();
    DiscoveryNode currentNode = explanation.getCurrentNode();
    UnassignedInfo unassignedInfo = explanation.getUnassignedInfo();
    ClusterInfo clusterInfo = explanation.getClusterInfo();
    AllocateUnassignedDecision allocateDecision = explanation.getShardAllocationDecision().getAllocateDecision();
    MoveDecision moveDecision = explanation.getShardAllocationDecision().getMoveDecision();
    // verify shard info
    assertEquals("idx", shardId.getIndexName());
    assertEquals(0, shardId.getId());
    assertTrue(isPrimary);
    // verify current node info
    assertEquals(ShardRoutingState.STARTED, shardRoutingState);
    assertNotNull(currentNode);
    // verify unassigned info
    assertNull(unassignedInfo);
    // verify cluster info
    verifyClusterInfo(clusterInfo, includeDiskInfo, 2);
    // verify decision object
    assertFalse(allocateDecision.isDecisionTaken());
    assertTrue(moveDecision.isDecisionTaken());
    assertEquals(AllocationDecision.NO, moveDecision.getAllocationDecision());
    assertEquals("rebalancing is not allowed, even though there is at least one node on which the shard can be allocated", moveDecision.getExplanation());
    assertTrue(moveDecision.canRemain());
    assertFalse(moveDecision.forceMove());
    assertFalse(moveDecision.canRebalanceCluster());
    assertNotNull(moveDecision.getCanRemainDecision());
    assertNull(moveDecision.getTargetNode());
    assertEquals(2, moveDecision.getCurrentNodeRanking());
    // verifying cluster rebalance decision object
    assertNotNull(moveDecision.getClusterRebalanceDecision());
    assertEquals(Decision.Type.NO, moveDecision.getClusterRebalanceDecision().type());
    for (Decision d : moveDecision.getClusterRebalanceDecision().getDecisions()) {
        if (d.label().equals("enable")) {
            assertEquals(Decision.Type.NO, d.type());
            assertEquals("no rebalancing is allowed due to index setting [index.routing.rebalance.enable=none]", d.getExplanation());
        } else {
            assertEquals(Decision.Type.YES, d.type());
            assertNotNull(d.getExplanation());
        }
    }
    // verify node decisions
    assertEquals(1, moveDecision.getNodeDecisions().size());
    NodeAllocationResult result = moveDecision.getNodeDecisions().get(0);
    assertNotNull(result.getNode());
    assertEquals(1, result.getWeightRanking());
    assertEquals(AllocationDecision.YES, result.getNodeDecision());
    if (includeYesDecisions) {
        assertThat(result.getCanAllocateDecision().getDecisions().size(), greaterThan(0));
    } else {
        assertEquals(0, result.getCanAllocateDecision().getDecisions().size());
    }
    for (Decision d : result.getCanAllocateDecision().getDecisions()) {
        assertEquals(Decision.Type.YES, d.type());
        assertNotNull(d.getExplanation());
    }
    // verify JSON output
    try (XContentParser parser = getParser(explanation)) {
        verifyShardInfo(parser, true, includeDiskInfo, ShardRoutingState.STARTED);
        parser.nextToken();
        assertEquals("can_remain_on_current_node", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.YES.toString(), parser.text());
        parser.nextToken();
        assertEquals("can_rebalance_cluster", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.NO.toString(), parser.text());
        parser.nextToken();
        assertEquals("can_rebalance_cluster_decisions", parser.currentName());
        verifyDeciders(parser, AllocationDecision.NO);
        parser.nextToken();
        assertEquals("can_rebalance_to_other_node", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.NO.toString(), parser.text());
        parser.nextToken();
        assertEquals("rebalance_explanation", parser.currentName());
        parser.nextToken();
        assertEquals("rebalancing is not allowed, even though there is at least one node on which the shard can be allocated", parser.text());
        verifyNodeDecisions(parser, allNodeDecisions(AllocationDecision.YES, true), includeYesDecisions, false);
        assertEquals(Token.END_OBJECT, parser.nextToken());
    }
}

Example 29 with UnassignedInfo

use of org.opensearch.cluster.routing.UnassignedInfo in project OpenSearch by opensearch-project.

the class ClusterAllocationExplainIT method testWorseBalance.

public void testWorseBalance() throws Exception {
    logger.info("--> starting a single node");
    internalCluster().startNode();
    ensureStableCluster(1);
    prepareIndex(5, 0);
    logger.info("--> setting balancing threshold really high, so it won't be met");
    client().admin().cluster().prepareUpdateSettings().setTransientSettings(Settings.builder().put("cluster.routing.allocation.balance.threshold", 1000.0f)).get();
    logger.info("--> starting another node, with the rebalance threshold so high, it should not get any shards");
    internalCluster().startNode();
    ensureStableCluster(2);
    boolean includeYesDecisions = randomBoolean();
    boolean includeDiskInfo = randomBoolean();
    ClusterAllocationExplanation explanation = runExplain(true, includeYesDecisions, includeDiskInfo);
    ShardId shardId = explanation.getShard();
    boolean isPrimary = explanation.isPrimary();
    ShardRoutingState shardRoutingState = explanation.getShardState();
    DiscoveryNode currentNode = explanation.getCurrentNode();
    UnassignedInfo unassignedInfo = explanation.getUnassignedInfo();
    ClusterInfo clusterInfo = explanation.getClusterInfo();
    AllocateUnassignedDecision allocateDecision = explanation.getShardAllocationDecision().getAllocateDecision();
    MoveDecision moveDecision = explanation.getShardAllocationDecision().getMoveDecision();
    // verify shard info
    assertEquals("idx", shardId.getIndexName());
    assertEquals(0, shardId.getId());
    assertTrue(isPrimary);
    // verify current node info
    assertEquals(ShardRoutingState.STARTED, shardRoutingState);
    assertNotNull(currentNode);
    // verify unassigned info
    assertNull(unassignedInfo);
    // verify cluster info
    verifyClusterInfo(clusterInfo, includeDiskInfo, 2);
    // verify decision object
    assertFalse(allocateDecision.isDecisionTaken());
    assertTrue(moveDecision.isDecisionTaken());
    assertEquals(AllocationDecision.NO, moveDecision.getAllocationDecision());
    assertEquals("cannot rebalance as no target node exists that can both allocate this shard and improve the cluster balance", moveDecision.getExplanation());
    assertTrue(moveDecision.canRemain());
    assertFalse(moveDecision.forceMove());
    assertTrue(moveDecision.canRebalanceCluster());
    assertNotNull(moveDecision.getCanRemainDecision());
    assertNull(moveDecision.getTargetNode());
    assertEquals(1, moveDecision.getCurrentNodeRanking());
    // verifying cluster rebalance decision object
    assertNotNull(moveDecision.getClusterRebalanceDecision());
    assertEquals(Decision.Type.YES, moveDecision.getClusterRebalanceDecision().type());
    for (Decision d : moveDecision.getClusterRebalanceDecision().getDecisions()) {
        assertEquals(Decision.Type.YES, d.type());
        assertNotNull(d.getExplanation());
    }
    // verify node decisions
    assertEquals(1, moveDecision.getNodeDecisions().size());
    NodeAllocationResult result = moveDecision.getNodeDecisions().get(0);
    assertNotNull(result.getNode());
    assertEquals(1, result.getWeightRanking());
    assertEquals(AllocationDecision.WORSE_BALANCE, result.getNodeDecision());
    if (includeYesDecisions) {
        assertThat(result.getCanAllocateDecision().getDecisions().size(), greaterThan(0));
    } else {
        assertEquals(0, result.getCanAllocateDecision().getDecisions().size());
    }
    for (Decision d : result.getCanAllocateDecision().getDecisions()) {
        assertEquals(Decision.Type.YES, d.type());
        assertNotNull(d.getExplanation());
    }
    // verify JSON output
    try (XContentParser parser = getParser(explanation)) {
        verifyShardInfo(parser, true, includeDiskInfo, ShardRoutingState.STARTED);
        parser.nextToken();
        assertEquals("can_remain_on_current_node", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.YES.toString(), parser.text());
        parser.nextToken();
        assertEquals("can_rebalance_cluster", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.YES.toString(), parser.text());
        parser.nextToken();
        assertEquals("can_rebalance_to_other_node", parser.currentName());
        parser.nextToken();
        assertEquals(AllocationDecision.NO.toString(), parser.text());
        parser.nextToken();
        assertEquals("rebalance_explanation", parser.currentName());
        parser.nextToken();
        assertEquals("cannot rebalance as no target node exists that can both allocate this shard and improve the cluster balance", parser.text());
        verifyNodeDecisions(parser, allNodeDecisions(AllocationDecision.WORSE_BALANCE, true), includeYesDecisions, false);
        assertEquals(Token.END_OBJECT, parser.nextToken());
    }
}

Example 30 with UnassignedInfo

use of org.opensearch.cluster.routing.UnassignedInfo in project OpenSearch by opensearch-project.

the class SharedClusterSnapshotRestoreIT method testUnrestorableFilesDuringRestore.

/**
 * Test that restoring a snapshot whose files can't be downloaded at all is not stuck or
 * does not hang indefinitely.
 */
public void testUnrestorableFilesDuringRestore() throws Exception {
    final String indexName = "unrestorable-files";
    final int maxRetries = randomIntBetween(1, 10);
    Settings createIndexSettings = Settings.builder().put(SETTING_ALLOCATION_MAX_RETRY.getKey(), maxRetries).put(IndexMetadata.INDEX_NUMBER_OF_SHARDS_SETTING.getKey(), 1).build();
    Settings repositorySettings = Settings.builder().put("random", randomAlphaOfLength(10)).put("max_failure_number", 10000000L).put("use_lucene_corruption", false).put("random_data_file_io_exception_rate", 1.0).build();
    Consumer<UnassignedInfo> checkUnassignedInfo = unassignedInfo -> {
        assertThat(unassignedInfo.getReason(), equalTo(UnassignedInfo.Reason.ALLOCATION_FAILED));
        assertThat(unassignedInfo.getNumFailedAllocations(), anyOf(equalTo(maxRetries), equalTo(1)));
    };
    unrestorableUseCase(indexName, createIndexSettings, repositorySettings, Settings.EMPTY, checkUnassignedInfo, () -> {
    });
}