use of org.opensearch.action.admin.cluster.reroute.ClusterRerouteResponse in project OpenSearch by opensearch-project.
the class ShrinkIndexIT method testCreateShrinkIndexFails.
/**
* Tests that we can manually recover from a failed allocation due to shards being moved away etc.
*/
public void testCreateShrinkIndexFails() throws Exception {
internalCluster().ensureAtLeastNumDataNodes(2);
prepareCreate("source").setSettings(Settings.builder().put(indexSettings()).put("number_of_shards", randomIntBetween(2, 7)).put("number_of_replicas", 0)).get();
for (int i = 0; i < 20; i++) {
client().prepareIndex("source").setSource("{\"foo\" : \"bar\", \"i\" : " + i + "}", XContentType.JSON).get();
}
ImmutableOpenMap<String, DiscoveryNode> dataNodes = client().admin().cluster().prepareState().get().getState().nodes().getDataNodes();
assertTrue("at least 2 nodes but was: " + dataNodes.size(), dataNodes.size() >= 2);
DiscoveryNode[] discoveryNodes = dataNodes.values().toArray(DiscoveryNode.class);
String spareNode = discoveryNodes[0].getName();
String mergeNode = discoveryNodes[1].getName();
// ensure all shards are allocated otherwise the ensure green below might not succeed since we require the merge node
// if we change the setting too quickly we will end up with one replica unassigned which can't be assigned anymore due
// to the require._name below.
ensureGreen();
// relocate all shards to one node such that we can merge it.
client().admin().indices().prepareUpdateSettings("source").setSettings(Settings.builder().put("index.routing.allocation.require._name", mergeNode).put("index.blocks.write", true)).get();
ensureGreen();
// now merge source into a single shard index
client().admin().indices().prepareResizeIndex("source", "target").setWaitForActiveShards(ActiveShardCount.NONE).setSettings(Settings.builder().put("index.routing.allocation.exclude._name", // we manually exclude the merge node to forcefully fuck it up
mergeNode).put("index.number_of_replicas", 0).put("index.allocation.max_retries", 1).build()).get();
client().admin().cluster().prepareHealth("target").setWaitForEvents(Priority.LANGUID).get();
// now we move all shards away from the merge node
client().admin().indices().prepareUpdateSettings("source").setSettings(Settings.builder().put("index.routing.allocation.require._name", spareNode).put("index.blocks.write", true)).get();
ensureGreen("source");
client().admin().indices().prepareUpdateSettings(// erase the forcefully fuckup!
"target").setSettings(Settings.builder().putNull("index.routing.allocation.exclude._name")).get();
// wait until it fails
assertBusy(() -> {
ClusterStateResponse clusterStateResponse = client().admin().cluster().prepareState().get();
RoutingTable routingTables = clusterStateResponse.getState().routingTable();
assertTrue(routingTables.index("target").shard(0).getShards().get(0).unassigned());
assertEquals(UnassignedInfo.Reason.ALLOCATION_FAILED, routingTables.index("target").shard(0).getShards().get(0).unassignedInfo().getReason());
assertEquals(1, routingTables.index("target").shard(0).getShards().get(0).unassignedInfo().getNumFailedAllocations());
});
client().admin().indices().prepareUpdateSettings(// now relocate them all to the right node
"source").setSettings(Settings.builder().put("index.routing.allocation.require._name", mergeNode)).get();
ensureGreen("source");
final InternalClusterInfoService infoService = (InternalClusterInfoService) internalCluster().getInstance(ClusterInfoService.class, internalCluster().getMasterName());
infoService.refresh();
// kick off a retry and wait until it's done!
ClusterRerouteResponse clusterRerouteResponse = client().admin().cluster().prepareReroute().setRetryFailed(true).get();
long expectedShardSize = clusterRerouteResponse.getState().routingTable().index("target").shard(0).getShards().get(0).getExpectedShardSize();
// we support the expected shard size in the allocator to sum up over the source index shards
assertTrue("expected shard size must be set but wasn't: " + expectedShardSize, expectedShardSize > 0);
ensureGreen();
assertHitCount(client().prepareSearch("target").setSize(100).setQuery(new TermsQueryBuilder("foo", "bar")).get(), 20);
}
use of org.opensearch.action.admin.cluster.reroute.ClusterRerouteResponse in project OpenSearch by opensearch-project.
the class RelocationIT method testRelocateWhileContinuouslyIndexingAndWaitingForRefresh.
public void testRelocateWhileContinuouslyIndexingAndWaitingForRefresh() throws Exception {
logger.info("--> starting [node1] ...");
final String node1 = internalCluster().startNode();
logger.info("--> creating test index ...");
prepareCreate("test", // we
Settings.builder().put("index.number_of_shards", 1).put("index.number_of_replicas", 0).put("index.refresh_interval", -1)).get();
logger.info("--> index 10 docs");
for (int i = 0; i < 10; i++) {
client().prepareIndex("test").setId(Integer.toString(i)).setSource("field", "value" + i).execute().actionGet();
}
logger.info("--> flush so we have an actual index");
client().admin().indices().prepareFlush().execute().actionGet();
logger.info("--> index more docs so we have something in the translog");
final List<ActionFuture<IndexResponse>> pendingIndexResponses = new ArrayList<>();
for (int i = 10; i < 20; i++) {
pendingIndexResponses.add(client().prepareIndex("test").setId(Integer.toString(i)).setRefreshPolicy(WriteRequest.RefreshPolicy.WAIT_UNTIL).setSource("field", "value" + i).execute());
}
logger.info("--> start another node");
final String node2 = internalCluster().startNode();
ClusterHealthResponse clusterHealthResponse = client().admin().cluster().prepareHealth().setWaitForEvents(Priority.LANGUID).setWaitForNodes("2").execute().actionGet();
assertThat(clusterHealthResponse.isTimedOut(), equalTo(false));
logger.info("--> relocate the shard from node1 to node2");
ActionFuture<ClusterRerouteResponse> relocationListener = client().admin().cluster().prepareReroute().add(new MoveAllocationCommand("test", 0, node1, node2)).execute();
logger.info("--> index 100 docs while relocating");
for (int i = 20; i < 120; i++) {
pendingIndexResponses.add(client().prepareIndex("test").setId(Integer.toString(i)).setRefreshPolicy(WriteRequest.RefreshPolicy.WAIT_UNTIL).setSource("field", "value" + i).execute());
}
relocationListener.actionGet();
clusterHealthResponse = client().admin().cluster().prepareHealth().setWaitForEvents(Priority.LANGUID).setWaitForNoRelocatingShards(true).setTimeout(ACCEPTABLE_RELOCATION_TIME).execute().actionGet();
assertThat(clusterHealthResponse.isTimedOut(), equalTo(false));
logger.info("--> verifying count");
assertBusy(() -> {
client().admin().indices().prepareRefresh().execute().actionGet();
assertTrue(pendingIndexResponses.stream().allMatch(ActionFuture::isDone));
}, 1, TimeUnit.MINUTES);
assertThat(client().prepareSearch("test").setSize(0).execute().actionGet().getHits().getTotalHits().value, equalTo(120L));
}
use of org.opensearch.action.admin.cluster.reroute.ClusterRerouteResponse in project OpenSearch by opensearch-project.
the class ClusterRerouteIT method testRerouteExplain.
public void testRerouteExplain() {
Settings commonSettings = Settings.builder().build();
logger.info("--> starting a node");
String node_1 = internalCluster().startNode(commonSettings);
assertThat(cluster().size(), equalTo(1));
ClusterHealthResponse healthResponse = client().admin().cluster().prepareHealth().setWaitForNodes("1").execute().actionGet();
assertThat(healthResponse.isTimedOut(), equalTo(false));
logger.info("--> create an index with 1 shard");
createIndex("test", Settings.builder().put(IndexMetadata.SETTING_NUMBER_OF_SHARDS, 1).put(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, 0).build());
if (randomBoolean()) {
assertAcked(client().admin().indices().prepareClose("test"));
}
ensureGreen("test");
logger.info("--> disable allocation");
Settings newSettings = Settings.builder().put(CLUSTER_ROUTING_ALLOCATION_ENABLE_SETTING.getKey(), Allocation.NONE.name()).build();
client().admin().cluster().prepareUpdateSettings().setTransientSettings(newSettings).execute().actionGet();
logger.info("--> starting a second node");
String node_2 = internalCluster().startNode(commonSettings);
assertThat(cluster().size(), equalTo(2));
healthResponse = client().admin().cluster().prepareHealth().setWaitForNodes("2").execute().actionGet();
assertThat(healthResponse.isTimedOut(), equalTo(false));
logger.info("--> try to move the shard from node1 to node2");
MoveAllocationCommand cmd = new MoveAllocationCommand("test", 0, node_1, node_2);
ClusterRerouteResponse resp = client().admin().cluster().prepareReroute().add(cmd).setExplain(true).execute().actionGet();
RoutingExplanations e = resp.getExplanations();
assertThat(e.explanations().size(), equalTo(1));
RerouteExplanation explanation = e.explanations().get(0);
assertThat(explanation.command().name(), equalTo(cmd.name()));
assertThat(((MoveAllocationCommand) explanation.command()).shardId(), equalTo(cmd.shardId()));
assertThat(((MoveAllocationCommand) explanation.command()).fromNode(), equalTo(cmd.fromNode()));
assertThat(((MoveAllocationCommand) explanation.command()).toNode(), equalTo(cmd.toNode()));
assertThat(explanation.decisions().type(), equalTo(Decision.Type.YES));
}
use of org.opensearch.action.admin.cluster.reroute.ClusterRerouteResponse in project OpenSearch by opensearch-project.
the class ClusterRerouteIT method testMessageLogging.
public void testMessageLogging() throws Exception {
final Settings settings = Settings.builder().put(EnableAllocationDecider.CLUSTER_ROUTING_ALLOCATION_ENABLE_SETTING.getKey(), Allocation.NONE.name()).put(EnableAllocationDecider.CLUSTER_ROUTING_REBALANCE_ENABLE_SETTING.getKey(), EnableAllocationDecider.Rebalance.NONE.name()).build();
final String nodeName1 = internalCluster().startNode(settings);
assertThat(cluster().size(), equalTo(1));
ClusterHealthResponse healthResponse = client().admin().cluster().prepareHealth().setWaitForNodes("1").execute().actionGet();
assertThat(healthResponse.isTimedOut(), equalTo(false));
final String nodeName2 = internalCluster().startNode(settings);
assertThat(cluster().size(), equalTo(2));
healthResponse = client().admin().cluster().prepareHealth().setWaitForNodes("2").execute().actionGet();
assertThat(healthResponse.isTimedOut(), equalTo(false));
final String indexName = "test_index";
client().admin().indices().prepareCreate(indexName).setWaitForActiveShards(ActiveShardCount.NONE).setSettings(Settings.builder().put(IndexMetadata.INDEX_NUMBER_OF_SHARDS_SETTING.getKey(), 2).put(IndexMetadata.INDEX_NUMBER_OF_REPLICAS_SETTING.getKey(), 1)).execute().actionGet();
Logger actionLogger = LogManager.getLogger(TransportClusterRerouteAction.class);
try (MockLogAppender dryRunMockLog = MockLogAppender.createForLoggers(actionLogger)) {
dryRunMockLog.addExpectation(new MockLogAppender.UnseenEventExpectation("no completed message logged on dry run", TransportClusterRerouteAction.class.getName(), Level.INFO, "allocated an empty primary*"));
AllocationCommand dryRunAllocation = new AllocateEmptyPrimaryAllocationCommand(indexName, 0, nodeName1, true);
ClusterRerouteResponse dryRunResponse = client().admin().cluster().prepareReroute().setExplain(randomBoolean()).setDryRun(true).add(dryRunAllocation).execute().actionGet();
// during a dry run, messages exist but are not logged or exposed
assertThat(dryRunResponse.getExplanations().getYesDecisionMessages(), hasSize(1));
assertThat(dryRunResponse.getExplanations().getYesDecisionMessages().get(0), containsString("allocated an empty primary"));
dryRunMockLog.assertAllExpectationsMatched();
}
try (MockLogAppender allocateMockLog = MockLogAppender.createForLoggers(actionLogger)) {
allocateMockLog.addExpectation(new MockLogAppender.SeenEventExpectation("message for first allocate empty primary", TransportClusterRerouteAction.class.getName(), Level.INFO, "allocated an empty primary*" + nodeName1 + "*"));
allocateMockLog.addExpectation(new MockLogAppender.UnseenEventExpectation("no message for second allocate empty primary", TransportClusterRerouteAction.class.getName(), Level.INFO, "allocated an empty primary*" + nodeName2 + "*"));
AllocationCommand yesDecisionAllocation = new AllocateEmptyPrimaryAllocationCommand(indexName, 0, nodeName1, true);
AllocationCommand noDecisionAllocation = new AllocateEmptyPrimaryAllocationCommand("noexist", 1, nodeName2, true);
ClusterRerouteResponse response = client().admin().cluster().prepareReroute().setExplain(// so we get a NO decision back rather than an exception
true).add(yesDecisionAllocation).add(noDecisionAllocation).execute().actionGet();
assertThat(response.getExplanations().getYesDecisionMessages(), hasSize(1));
assertThat(response.getExplanations().getYesDecisionMessages().get(0), containsString("allocated an empty primary"));
assertThat(response.getExplanations().getYesDecisionMessages().get(0), containsString(nodeName1));
allocateMockLog.assertAllExpectationsMatched();
}
}
Aggregations