public void addDeletion(SearchRequestBuilder searchRequest) {
searchRequest
.addSort("_doc", SortOrder.ASC)
.setScroll(TimeValue.timeValueMinutes(5))
.setSize(100)
// load only doc ids, not _source fields
.setFetchSource(false);
// this search is synchronous. An optimization would be to be non-blocking,
// but it requires to tracking pending requests in close().
// Same semaphore can't be reused because of potential deadlock (requires to acquire
// two locks)
SearchResponse searchResponse = searchRequest.get();
while (true) {
SearchHit[] hits = searchResponse.getHits().getHits();
for (SearchHit hit : hits) {
DeleteRequestBuilder deleteRequestBuilder =
client.prepareDelete(hit.index(), hit.type(), hit.getId());
SearchHitField routing = hit.field("_routing");
if (routing != null) {
deleteRequestBuilder.setRouting(routing.getValue());
}
add(deleteRequestBuilder.request());
}
String scrollId = searchResponse.getScrollId();
searchResponse =
client.prepareSearchScroll(scrollId).setScroll(TimeValue.timeValueMinutes(5)).get();
if (hits.length == 0) {
client.nativeClient().prepareClearScroll().addScrollId(scrollId).get();
break;
}
}
}
public static void main(String[] args) throws Exception {
NodesStressTest test =
new NodesStressTest()
.numberOfNodes(2)
.indexThreads(5)
.indexIterations(10 * 1000)
.searcherThreads(5)
.searchIterations(10 * 1000)
.sleepBeforeClose(TimeValue.timeValueMinutes(10))
.sleepAfterDone(TimeValue.timeValueMinutes(10))
.build(EMPTY_SETTINGS);
test.start();
}
@SuppressWarnings({"unchecked"})
@Inject
public SimpleRiver(
RiverName riverName, RiverSettings settings, Client client, ThreadPool threadPool) {
super(riverName, settings);
this.client = client;
if (settings.settings().containsKey("simple")) {
Map<String, Object> simpleSettings = (Map<String, Object>) settings.settings().get("simple");
simpleNumber = XContentMapValues.nodeIntegerValue(simpleSettings.get("number"), 100);
fieldName = XContentMapValues.nodeStringValue(simpleSettings.get("field"), "test");
poll =
XContentMapValues.nodeTimeValue(
simpleSettings.get("poll"), TimeValue.timeValueMinutes(60));
}
logger.info(
"creating simple stream river for [{} numbers] with field [{}]", simpleNumber, fieldName);
if (settings.settings().containsKey("index")) {
Map<String, Object> indexSettings = (Map<String, Object>) settings.settings().get("index");
indexName = XContentMapValues.nodeStringValue(indexSettings.get("index"), riverName.name());
typeName = XContentMapValues.nodeStringValue(indexSettings.get("type"), "simple_type");
bulkSize = XContentMapValues.nodeIntegerValue(indexSettings.get("bulk_size"), 100);
bulkThreshold = XContentMapValues.nodeIntegerValue(indexSettings.get("bulk_threshold"), 10);
} else {
indexName = riverName.name();
typeName = "simple_type";
bulkSize = 100;
bulkThreshold = 10;
}
}
public void testCloseAndReopenOrDeleteWithActiveScroll() throws IOException {
createIndex("test");
for (int i = 0; i < 100; i++) {
client()
.prepareIndex("test", "type1", Integer.toString(i))
.setSource(jsonBuilder().startObject().field("field", i).endObject())
.execute()
.actionGet();
}
refresh();
SearchResponse searchResponse =
client()
.prepareSearch()
.setQuery(matchAllQuery())
.setSize(35)
.setScroll(TimeValue.timeValueMinutes(2))
.addSort("field", SortOrder.ASC)
.execute()
.actionGet();
long counter = 0;
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(35));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
if (randomBoolean()) {
client().admin().indices().prepareClose("test").get();
client().admin().indices().prepareOpen("test").get();
ensureGreen("test");
} else {
client().admin().indices().prepareDelete("test").get();
}
}
public ScrollChunk nextChunk() {
final SearchResponse search;
// make sure to return the initial hits, see
// https://github.com/Graylog2/graylog2-server/issues/2126
if (firstResponse == null) {
search =
client
.prepareSearchScroll(scrollId)
.setScroll(TimeValue.timeValueMinutes(1))
.execute()
.actionGet();
} else {
search = firstResponse;
firstResponse = null;
}
final SearchHits hits = search.getHits();
if (hits.getHits().length == 0) {
// scroll exhausted
LOG.debug("[{}] Reached end of scroll results.", queryHash, getOriginalQuery());
return null;
}
LOG.debug(
"[{}][{}] New scroll id {}, number of hits in chunk: {}",
queryHash,
chunkId,
search.getScrollId(),
hits.getHits().length);
scrollId = search.getScrollId(); // save the id for the next request.
return new ScrollChunk(hits, fields, chunkId++);
}
public static void main(String[] args) throws Exception {
System.setProperty("es.logger.prefix", "");
Settings settings =
settingsBuilder()
.put("index.shard.check_index", true)
.put("gateway.type", "none")
.put("path.data", "data/data1,data/data2")
.build();
RollingRestartStressTest test =
new RollingRestartStressTest()
.settings(settings)
.numberOfNodes(4)
.numberOfShards(5)
.numberOfReplicas(1)
.initialNumberOfDocs(1000)
.textTokens(150)
.numberOfFields(10)
.cleanNodeData(false)
.indexers(5)
.indexerThrottle(TimeValue.timeValueMillis(50))
.period(TimeValue.timeValueMinutes(3));
test.run();
}
@Test
public void get_with_time_value_timeout_is_not_yet_implemented() {
try {
esTester.client().prepareSearchScroll("scrollId").get(TimeValue.timeValueMinutes(1));
fail();
} catch (Exception e) {
assertThat(e).isInstanceOf(IllegalStateException.class).hasMessage("Not yet implemented");
}
}
@Override
public boolean hasNext() {
if (hits.isEmpty()) {
SearchScrollRequestBuilder esRequest =
esClient
.prepareSearchScroll(scrollId)
.setScroll(TimeValue.timeValueMinutes(SCROLL_TIME_IN_MINUTES));
Collections.addAll(hits, esRequest.get().getHits().getHits());
}
return !hits.isEmpty();
}
@Test
@TestLogging(value = "cluster.service:TRACE")
public void testDeleteCreateInOneBulk() throws Exception {
internalCluster()
.startNodesAsync(
2, Settings.builder().put(DiscoveryModule.DISCOVERY_TYPE_KEY, "zen").build())
.get();
assertFalse(client().admin().cluster().prepareHealth().setWaitForNodes("2").get().isTimedOut());
prepareCreate("test")
.setSettings(IndexMetaData.SETTING_AUTO_EXPAND_REPLICAS, true)
.addMapping("type")
.get();
ensureGreen("test");
// now that the cluster is stable, remove publishing timeout
assertAcked(
client()
.admin()
.cluster()
.prepareUpdateSettings()
.setTransientSettings(Settings.builder().put(DiscoverySettings.PUBLISH_TIMEOUT, "0")));
Set<String> nodes = new HashSet<>(Arrays.asList(internalCluster().getNodeNames()));
nodes.remove(internalCluster().getMasterName());
// block none master node.
BlockClusterStateProcessing disruption =
new BlockClusterStateProcessing(nodes.iterator().next(), getRandom());
internalCluster().setDisruptionScheme(disruption);
logger.info("--> indexing a doc");
index("test", "type", "1");
refresh();
disruption.startDisrupting();
logger.info("--> delete index and recreate it");
assertFalse(
client()
.admin()
.indices()
.prepareDelete("test")
.setTimeout("200ms")
.get()
.isAcknowledged());
assertFalse(
prepareCreate("test")
.setTimeout("200ms")
.setSettings(IndexMetaData.SETTING_AUTO_EXPAND_REPLICAS, true)
.get()
.isAcknowledged());
logger.info("--> letting cluster proceed");
disruption.stopDisrupting();
ensureGreen(TimeValue.timeValueMinutes(30), "test");
assertHitCount(client().prepareSearch("test").get(), 0);
}
public void scanThis(
AnyExecutor<T> any, QueryBuilder query, long keepTimeInMinutes, int pageSize) {
SearchRequestBuilder srb =
client
.prepareSearch(getIndexName())
.setQuery(query)
.setSize(pageSize)
.setSearchType(SearchType.SCAN)
.setScroll(TimeValue.timeValueMinutes(keepTimeInMinutes));
SearchResponse rsp = srb.execute().actionGet();
try {
int counter = 0;
while (true) {
rsp =
client
.prepareSearchScroll(rsp.scrollId())
.setScroll(TimeValue.timeValueMinutes(keepTimeInMinutes))
.execute()
.actionGet();
long currentResults = rsp.hits().hits().length;
logger.info(
"("
+ counter++
+ ") scanquery with "
+ pageSize
+ " page size and "
+ currentResults
+ " hits");
if (currentResults == 0) break;
for (T t : collectObjects(rsp)) {
any.execute(t);
}
}
} catch (Exception ex) {
logger.error("Cannot run scanThis", ex);
}
}
@Test
public void no_trace_logs() {
logTester.setLevel(LoggerLevel.DEBUG);
SearchResponse response =
esTester
.client()
.prepareSearch(FakeIndexDefinition.INDEX)
.setSearchType(SearchType.SCAN)
.setScroll(TimeValue.timeValueMinutes(1))
.get();
logTester.clear();
esTester.client().prepareSearchScroll(response.getScrollId()).get();
assertThat(logTester.logs()).isEmpty();
}
public static long parseHumanDTToMills(String humanDateTime) {
long datetime = 0;
if (humanDateTime.indexOf("ms") != -1) {
datetime =
TimeValue.timeValueMillis(NumberUtils.toLong(humanDateTime.replaceAll("ms", "")))
.getMillis();
} else if (humanDateTime.indexOf("h") != -1) {
datetime =
TimeValue.timeValueHours(NumberUtils.toLong(humanDateTime.replaceAll("h", "")))
.getMillis();
} else if (humanDateTime.indexOf("m") != -1) {
datetime =
TimeValue.timeValueMinutes(NumberUtils.toLong(humanDateTime.replaceAll("m", "")))
.getMillis();
} else if (humanDateTime.indexOf("s") != -1) {
datetime =
TimeValue.timeValueSeconds(NumberUtils.toLong(humanDateTime.replaceAll("s", "")))
.getMillis();
}
return datetime;
}
private void cleanFailedShards(final ClusterChangedEvent event) {
RoutingTable routingTable = event.state().routingTable();
RoutingNodes.RoutingNodeIterator routingNode =
event.state().readOnlyRoutingNodes().routingNodeIter(event.state().nodes().localNodeId());
if (routingNode == null) {
failedShards.clear();
return;
}
DiscoveryNodes nodes = event.state().nodes();
long now = System.currentTimeMillis();
String localNodeId = nodes.localNodeId();
Iterator<Map.Entry<ShardId, FailedShard>> iterator = failedShards.entrySet().iterator();
shards:
while (iterator.hasNext()) {
Map.Entry<ShardId, FailedShard> entry = iterator.next();
FailedShard failedShard = entry.getValue();
IndexRoutingTable indexRoutingTable = routingTable.index(entry.getKey().getIndex());
if (indexRoutingTable != null) {
IndexShardRoutingTable shardRoutingTable = indexRoutingTable.shard(entry.getKey().id());
if (shardRoutingTable != null) {
for (ShardRouting shardRouting : shardRoutingTable.assignedShards()) {
if (localNodeId.equals(shardRouting.currentNodeId())) {
// we have a timeout here just to make sure we don't have dangled failed shards for
// some reason
// its just another safely layer
if (shardRouting.version() == failedShard.version
&& ((now - failedShard.timestamp) < TimeValue.timeValueMinutes(60).millis())) {
// It's the same failed shard - keep it if it hasn't timed out
continue shards;
} else {
// Different version or expired, remove it
break;
}
}
}
}
}
iterator.remove();
}
}
void assertNewReplicasWork(String indexName) throws Exception {
final int numReplicas = 1;
final long startTime = System.currentTimeMillis();
logger.debug("--> creating [{}] replicas for index [{}]", numReplicas, indexName);
assertAcked(
client()
.admin()
.indices()
.prepareUpdateSettings(indexName)
.setSettings(Settings.builder().put("number_of_replicas", numReplicas))
.execute()
.actionGet());
ensureGreen(TimeValue.timeValueMinutes(2), indexName);
logger.debug(
"--> index [{}] is green, took [{}]",
indexName,
TimeValue.timeValueMillis(System.currentTimeMillis() - startTime));
logger.debug(
"--> recovery status:\n{}",
XContentHelper.toString(client().admin().indices().prepareRecoveries(indexName).get()));
// TODO: do something with the replicas! query? index?
}
@Override
public void start() {
logger.info("starting simple stream");
bulkProcessor =
BulkProcessor.builder(
client,
new BulkProcessor.Listener() {
@Override
public void beforeBulk(long executionId, BulkRequest request) {
logger.info(
"Going to execute new bulk composed of {} actions",
request.numberOfActions());
}
@Override
public void afterBulk(
long executionId, BulkRequest request, BulkResponse response) {
logger.info("Executed bulk composed of {} actions", request.numberOfActions());
}
@Override
public void afterBulk(long executionId, BulkRequest request, Throwable failure) {
logger.warn("Error executing bulk", failure);
}
})
.setBulkActions(bulkSize)
.setFlushInterval(TimeValue.timeValueMinutes(5))
.setConcurrentRequests(bulkThreshold)
.build();
thread =
EsExecutors.daemonThreadFactory(settings.globalSettings(), "Simple processor")
.newThread(new SimpleConnector());
thread.start();
}
private void waitForClusterState(long clusterStateVersion) {
ClusterStateObserver observer =
new ClusterStateObserver(
clusterService, TimeValue.timeValueMinutes(5), logger, threadPool.getThreadContext());
final ClusterState clusterState = observer.observedState();
if (clusterState.getVersion() >= clusterStateVersion) {
logger.trace(
"node has cluster state with version higher than {} (current: {})",
clusterStateVersion,
clusterState.getVersion());
return;
} else {
logger.trace(
"waiting for cluster state version {} (current: {})",
clusterStateVersion,
clusterState.getVersion());
final PlainActionFuture<Void> future = new PlainActionFuture<>();
observer.waitForNextChange(
new ClusterStateObserver.Listener() {
@Override
public void onNewClusterState(ClusterState state) {
future.onResponse(null);
}
@Override
public void onClusterServiceClose() {
future.onFailure(new NodeClosedException(clusterService.localNode()));
}
@Override
public void onTimeout(TimeValue timeout) {
future.onFailure(
new IllegalStateException(
"cluster state never updated to version " + clusterStateVersion));
}
},
new ClusterStateObserver.ValidationPredicate() {
@Override
protected boolean validate(ClusterState newState) {
return newState.getVersion() >= clusterStateVersion;
}
});
try {
future.get();
logger.trace(
"successfully waited for cluster state with version {} (current: {})",
clusterStateVersion,
observer.observedState().getVersion());
} catch (Exception e) {
logger.debug(
(Supplier<?>)
() ->
new ParameterizedMessage(
"failed waiting for cluster state with version {} (current: {})",
clusterStateVersion,
observer.observedState()),
e);
throw ExceptionsHelper.convertToRuntime(e);
}
}
}
@SuppressWarnings({"unchecked"})
@Inject
public RabbitmqRiver(
RiverName riverName, RiverSettings settings, Client client, ScriptService scriptService) {
super(riverName, settings);
this.client = client;
if (settings.settings().containsKey("rabbitmq")) {
Map<String, Object> rabbitSettings =
(Map<String, Object>) settings.settings().get("rabbitmq");
if (rabbitSettings.containsKey("addresses")) {
List<Address> addresses = new ArrayList<Address>();
for (Map<String, Object> address :
(List<Map<String, Object>>) rabbitSettings.get("addresses")) {
addresses.add(
new Address(
XContentMapValues.nodeStringValue(address.get("host"), "localhost"),
XContentMapValues.nodeIntegerValue(address.get("port"), AMQP.PROTOCOL.PORT)));
}
rabbitAddresses = addresses.toArray(new Address[addresses.size()]);
} else {
String rabbitHost =
XContentMapValues.nodeStringValue(rabbitSettings.get("host"), "localhost");
int rabbitPort =
XContentMapValues.nodeIntegerValue(rabbitSettings.get("port"), AMQP.PROTOCOL.PORT);
rabbitAddresses = new Address[] {new Address(rabbitHost, rabbitPort)};
}
rabbitUser = XContentMapValues.nodeStringValue(rabbitSettings.get("user"), "guest");
rabbitPassword = XContentMapValues.nodeStringValue(rabbitSettings.get("pass"), "guest");
rabbitVhost = XContentMapValues.nodeStringValue(rabbitSettings.get("vhost"), "/");
rabbitQueue = XContentMapValues.nodeStringValue(rabbitSettings.get("queue"), "elasticsearch");
rabbitExchange =
XContentMapValues.nodeStringValue(rabbitSettings.get("exchange"), "elasticsearch");
rabbitRoutingKey =
XContentMapValues.nodeStringValue(rabbitSettings.get("routing_key"), "elasticsearch");
rabbitExchangeDeclare =
XContentMapValues.nodeBooleanValue(rabbitSettings.get("exchange_declare"), true);
if (rabbitExchangeDeclare) {
rabbitExchangeType =
XContentMapValues.nodeStringValue(rabbitSettings.get("exchange_type"), "direct");
rabbitExchangeDurable =
XContentMapValues.nodeBooleanValue(rabbitSettings.get("exchange_durable"), true);
} else {
rabbitExchangeType = "direct";
rabbitExchangeDurable = true;
}
rabbitQueueDeclare =
XContentMapValues.nodeBooleanValue(rabbitSettings.get("queue_declare"), true);
if (rabbitQueueDeclare) {
rabbitQueueDurable =
XContentMapValues.nodeBooleanValue(rabbitSettings.get("queue_durable"), true);
rabbitQueueAutoDelete =
XContentMapValues.nodeBooleanValue(rabbitSettings.get("queue_auto_delete"), false);
if (rabbitSettings.containsKey("args")) {
rabbitQueueArgs = (Map<String, Object>) rabbitSettings.get("args");
}
} else {
rabbitQueueDurable = true;
rabbitQueueAutoDelete = false;
}
rabbitQueueBind = XContentMapValues.nodeBooleanValue(rabbitSettings.get("queue_bind"), true);
rabbitQosPrefetchSize =
XContentMapValues.nodeIntegerValue(rabbitSettings.get("qos_prefetch_size"), 0);
rabbitQosPrefetchCount =
XContentMapValues.nodeIntegerValue(rabbitSettings.get("qos_prefetch_count"), 10);
rabbitHeartbeat =
TimeValue.parseTimeValue(
XContentMapValues.nodeStringValue(rabbitSettings.get("heartbeat"), "30m"),
TimeValue.timeValueMinutes(30));
} else {
rabbitAddresses = new Address[] {new Address("localhost", AMQP.PROTOCOL.PORT)};
rabbitUser = "******";
rabbitPassword = "******";
rabbitVhost = "/";
rabbitQueue = "elasticsearch";
rabbitQueueAutoDelete = false;
rabbitQueueDurable = true;
rabbitExchange = "elasticsearch";
rabbitExchangeType = "direct";
rabbitExchangeDurable = true;
rabbitRoutingKey = "elasticsearch";
rabbitExchangeDeclare = true;
rabbitQueueDeclare = true;
rabbitQueueBind = true;
rabbitQosPrefetchSize = 0;
rabbitQosPrefetchCount = 10;
rabbitHeartbeat = TimeValue.timeValueMinutes(30);
}
if (settings.settings().containsKey("index")) {
Map<String, Object> indexSettings = (Map<String, Object>) settings.settings().get("index");
bulkSize = XContentMapValues.nodeIntegerValue(indexSettings.get("bulk_size"), 100);
if (indexSettings.containsKey("bulk_timeout")) {
bulkTimeout =
TimeValue.parseTimeValue(
XContentMapValues.nodeStringValue(indexSettings.get("bulk_timeout"), "10ms"),
TimeValue.timeValueMillis(10));
} else {
bulkTimeout = TimeValue.timeValueMillis(10);
}
ordered = XContentMapValues.nodeBooleanValue(indexSettings.get("ordered"), false);
} else {
bulkSize = 100;
bulkTimeout = TimeValue.timeValueMillis(10);
ordered = false;
}
if (settings.settings().containsKey("bulk_script_filter")) {
Map<String, Object> scriptSettings =
(Map<String, Object>) settings.settings().get("bulk_script_filter");
if (scriptSettings.containsKey("script")) {
String scriptLang = "native";
if (scriptSettings.containsKey("script_lang")) {
scriptLang = scriptSettings.get("script_lang").toString();
}
Map<String, Object> scriptParams = null;
if (scriptSettings.containsKey("script_params")) {
scriptParams = (Map<String, Object>) scriptSettings.get("script_params");
} else {
scriptParams = Maps.newHashMap();
}
bulkScript =
scriptService.executable(
scriptLang, scriptSettings.get("script").toString(), scriptParams);
} else {
bulkScript = null;
}
} else {
bulkScript = null;
}
if (settings.settings().containsKey("script_filter")) {
Map<String, Object> scriptSettings =
(Map<String, Object>) settings.settings().get("script_filter");
if (scriptSettings.containsKey("script")) {
String scriptLang = "mvel";
if (scriptSettings.containsKey("script_lang")) {
scriptLang = scriptSettings.get("script_lang").toString();
}
Map<String, Object> scriptParams = null;
if (scriptSettings.containsKey("script_params")) {
scriptParams = (Map<String, Object>) scriptSettings.get("script_params");
} else {
scriptParams = Maps.newHashMap();
}
script =
scriptService.executable(
scriptLang, scriptSettings.get("script").toString(), scriptParams);
} else {
script = null;
}
} else {
script = null;
}
}
public class GatewayService extends AbstractLifecycleComponent implements ClusterStateListener {
public static final Setting<Integer> EXPECTED_NODES_SETTING =
Setting.intSetting("gateway.expected_nodes", -1, -1, Property.NodeScope);
public static final Setting<Integer> EXPECTED_DATA_NODES_SETTING =
Setting.intSetting("gateway.expected_data_nodes", -1, -1, Property.NodeScope);
public static final Setting<Integer> EXPECTED_MASTER_NODES_SETTING =
Setting.intSetting("gateway.expected_master_nodes", -1, -1, Property.NodeScope);
public static final Setting<TimeValue> RECOVER_AFTER_TIME_SETTING =
Setting.positiveTimeSetting(
"gateway.recover_after_time", TimeValue.timeValueMillis(0), Property.NodeScope);
public static final Setting<Integer> RECOVER_AFTER_NODES_SETTING =
Setting.intSetting("gateway.recover_after_nodes", -1, -1, Property.NodeScope);
public static final Setting<Integer> RECOVER_AFTER_DATA_NODES_SETTING =
Setting.intSetting("gateway.recover_after_data_nodes", -1, -1, Property.NodeScope);
public static final Setting<Integer> RECOVER_AFTER_MASTER_NODES_SETTING =
Setting.intSetting("gateway.recover_after_master_nodes", 0, 0, Property.NodeScope);
public static final ClusterBlock STATE_NOT_RECOVERED_BLOCK =
new ClusterBlock(
1,
"state not recovered / initialized",
true,
true,
RestStatus.SERVICE_UNAVAILABLE,
ClusterBlockLevel.ALL);
public static final TimeValue DEFAULT_RECOVER_AFTER_TIME_IF_EXPECTED_NODES_IS_SET =
TimeValue.timeValueMinutes(5);
private final Gateway gateway;
private final ThreadPool threadPool;
private final AllocationService allocationService;
private final ClusterService clusterService;
private final TimeValue recoverAfterTime;
private final int recoverAfterNodes;
private final int expectedNodes;
private final int recoverAfterDataNodes;
private final int expectedDataNodes;
private final int recoverAfterMasterNodes;
private final int expectedMasterNodes;
private final AtomicBoolean recovered = new AtomicBoolean();
private final AtomicBoolean scheduledRecovery = new AtomicBoolean();
@Inject
public GatewayService(
Settings settings,
AllocationService allocationService,
ClusterService clusterService,
ThreadPool threadPool,
GatewayMetaState metaState,
TransportNodesListGatewayMetaState listGatewayMetaState,
Discovery discovery,
IndicesService indicesService) {
super(settings);
this.gateway =
new Gateway(
settings, clusterService, metaState, listGatewayMetaState, discovery, indicesService);
this.allocationService = allocationService;
this.clusterService = clusterService;
this.threadPool = threadPool;
// allow to control a delay of when indices will get created
this.expectedNodes = EXPECTED_NODES_SETTING.get(this.settings);
this.expectedDataNodes = EXPECTED_DATA_NODES_SETTING.get(this.settings);
this.expectedMasterNodes = EXPECTED_MASTER_NODES_SETTING.get(this.settings);
if (RECOVER_AFTER_TIME_SETTING.exists(this.settings)) {
recoverAfterTime = RECOVER_AFTER_TIME_SETTING.get(this.settings);
} else if (expectedNodes >= 0 || expectedDataNodes >= 0 || expectedMasterNodes >= 0) {
recoverAfterTime = DEFAULT_RECOVER_AFTER_TIME_IF_EXPECTED_NODES_IS_SET;
} else {
recoverAfterTime = null;
}
this.recoverAfterNodes = RECOVER_AFTER_NODES_SETTING.get(this.settings);
this.recoverAfterDataNodes = RECOVER_AFTER_DATA_NODES_SETTING.get(this.settings);
// default the recover after master nodes to the minimum master nodes in the discovery
if (RECOVER_AFTER_MASTER_NODES_SETTING.exists(this.settings)) {
recoverAfterMasterNodes = RECOVER_AFTER_MASTER_NODES_SETTING.get(this.settings);
} else {
// TODO: change me once the minimum_master_nodes is changed too
recoverAfterMasterNodes = settings.getAsInt("discovery.zen.minimum_master_nodes", -1);
}
// Add the not recovered as initial state block, we don't allow anything until
this.clusterService.addInitialStateBlock(STATE_NOT_RECOVERED_BLOCK);
}
@Override
protected void doStart() {
// use post applied so that the state will be visible to the background recovery thread we spawn
// in performStateRecovery
clusterService.addListener(this);
}
@Override
protected void doStop() {
clusterService.removeListener(this);
}
@Override
protected void doClose() {}
@Override
public void clusterChanged(final ClusterChangedEvent event) {
if (lifecycle.stoppedOrClosed()) {
return;
}
final ClusterState state = event.state();
if (state.nodes().isLocalNodeElectedMaster() == false) {
// not our job to recover
return;
}
if (state.blocks().hasGlobalBlock(STATE_NOT_RECOVERED_BLOCK) == false) {
// already recovered
return;
}
DiscoveryNodes nodes = state.nodes();
if (state.nodes().getMasterNodeId() == null) {
logger.debug("not recovering from gateway, no master elected yet");
} else if (recoverAfterNodes != -1
&& (nodes.getMasterAndDataNodes().size()) < recoverAfterNodes) {
logger.debug(
"not recovering from gateway, nodes_size (data+master) [{}] < recover_after_nodes [{}]",
nodes.getMasterAndDataNodes().size(),
recoverAfterNodes);
} else if (recoverAfterDataNodes != -1 && nodes.getDataNodes().size() < recoverAfterDataNodes) {
logger.debug(
"not recovering from gateway, nodes_size (data) [{}] < recover_after_data_nodes [{}]",
nodes.getDataNodes().size(),
recoverAfterDataNodes);
} else if (recoverAfterMasterNodes != -1
&& nodes.getMasterNodes().size() < recoverAfterMasterNodes) {
logger.debug(
"not recovering from gateway, nodes_size (master) [{}] < recover_after_master_nodes [{}]",
nodes.getMasterNodes().size(),
recoverAfterMasterNodes);
} else {
boolean enforceRecoverAfterTime;
String reason;
if (expectedNodes == -1 && expectedMasterNodes == -1 && expectedDataNodes == -1) {
// no expected is set, honor the setting if they are there
enforceRecoverAfterTime = true;
reason = "recover_after_time was set to [" + recoverAfterTime + "]";
} else {
// one of the expected is set, see if all of them meet the need, and ignore the timeout in
// this case
enforceRecoverAfterTime = false;
reason = "";
if (expectedNodes != -1
&& (nodes.getMasterAndDataNodes().size()
< expectedNodes)) { // does not meet the expected...
enforceRecoverAfterTime = true;
reason =
"expecting ["
+ expectedNodes
+ "] nodes, but only have ["
+ nodes.getMasterAndDataNodes().size()
+ "]";
} else if (expectedDataNodes != -1
&& (nodes.getDataNodes().size() < expectedDataNodes)) { // does not meet the expected...
enforceRecoverAfterTime = true;
reason =
"expecting ["
+ expectedDataNodes
+ "] data nodes, but only have ["
+ nodes.getDataNodes().size()
+ "]";
} else if (expectedMasterNodes != -1
&& (nodes.getMasterNodes().size()
< expectedMasterNodes)) { // does not meet the expected...
enforceRecoverAfterTime = true;
reason =
"expecting ["
+ expectedMasterNodes
+ "] master nodes, but only have ["
+ nodes.getMasterNodes().size()
+ "]";
}
}
performStateRecovery(enforceRecoverAfterTime, reason);
}
}
private void performStateRecovery(boolean enforceRecoverAfterTime, String reason) {
final Gateway.GatewayStateRecoveredListener recoveryListener = new GatewayRecoveryListener();
if (enforceRecoverAfterTime && recoverAfterTime != null) {
if (scheduledRecovery.compareAndSet(false, true)) {
logger.info("delaying initial state recovery for [{}]. {}", recoverAfterTime, reason);
threadPool.schedule(
recoverAfterTime,
ThreadPool.Names.GENERIC,
() -> {
if (recovered.compareAndSet(false, true)) {
logger.info(
"recover_after_time [{}] elapsed. performing state recovery...",
recoverAfterTime);
gateway.performStateRecovery(recoveryListener);
}
});
}
} else {
if (recovered.compareAndSet(false, true)) {
threadPool
.generic()
.execute(
new AbstractRunnable() {
@Override
public void onFailure(Exception e) {
logger.warn("Recovery failed", e);
// we reset `recovered` in the listener don't reset it here otherwise there
// might be a race
// that resets it to false while a new recover is already running?
recoveryListener.onFailure("state recovery failed: " + e.getMessage());
}
@Override
protected void doRun() throws Exception {
gateway.performStateRecovery(recoveryListener);
}
});
}
}
}
public Gateway getGateway() {
return gateway;
}
class GatewayRecoveryListener implements Gateway.GatewayStateRecoveredListener {
@Override
public void onSuccess(final ClusterState recoveredState) {
logger.trace("successful state recovery, importing cluster state...");
clusterService.submitStateUpdateTask(
"local-gateway-elected-state",
new ClusterStateUpdateTask() {
@Override
public ClusterState execute(ClusterState currentState) {
assert currentState.metaData().indices().isEmpty();
// remove the block, since we recovered from gateway
ClusterBlocks.Builder blocks =
ClusterBlocks.builder()
.blocks(currentState.blocks())
.blocks(recoveredState.blocks())
.removeGlobalBlock(STATE_NOT_RECOVERED_BLOCK);
MetaData.Builder metaDataBuilder = MetaData.builder(recoveredState.metaData());
// automatically generate a UID for the metadata if we need to
metaDataBuilder.generateClusterUuidIfNeeded();
if (MetaData.SETTING_READ_ONLY_SETTING.get(recoveredState.metaData().settings())
|| MetaData.SETTING_READ_ONLY_SETTING.get(currentState.metaData().settings())) {
blocks.addGlobalBlock(MetaData.CLUSTER_READ_ONLY_BLOCK);
}
for (IndexMetaData indexMetaData : recoveredState.metaData()) {
metaDataBuilder.put(indexMetaData, false);
blocks.addBlocks(indexMetaData);
}
// update the state to reflect the new metadata and routing
ClusterState updatedState =
ClusterState.builder(currentState)
.blocks(blocks)
.metaData(metaDataBuilder)
.build();
// initialize all index routing tables as empty
RoutingTable.Builder routingTableBuilder =
RoutingTable.builder(updatedState.routingTable());
for (ObjectCursor<IndexMetaData> cursor :
updatedState.metaData().indices().values()) {
routingTableBuilder.addAsRecovery(cursor.value);
}
// start with 0 based versions for routing table
routingTableBuilder.version(0);
// now, reroute
updatedState =
ClusterState.builder(updatedState)
.routingTable(routingTableBuilder.build())
.build();
return allocationService.reroute(updatedState, "state recovered");
}
@Override
public void onFailure(String source, Exception e) {
logger.error(
(Supplier<?>)
() -> new ParameterizedMessage("unexpected failure during [{}]", source),
e);
GatewayRecoveryListener.this.onFailure("failed to updated cluster state");
}
@Override
public void clusterStateProcessed(
String source, ClusterState oldState, ClusterState newState) {
logger.info(
"recovered [{}] indices into cluster_state",
newState.metaData().indices().size());
}
});
}
@Override
public void onFailure(String message) {
recovered.set(false);
scheduledRecovery.set(false);
// don't remove the block here, we don't want to allow anything in such a case
logger.info("metadata state not restored, reason: {}", message);
}
}
// used for testing
public TimeValue recoverAfterTime() {
return recoverAfterTime;
}
}
@Test
public void testSimpleScrollQueryThenFetchSmallSizeUnevenDistribution() throws Exception {
client()
.admin()
.indices()
.prepareCreate("test")
.setSettings(Settings.settingsBuilder().put("index.number_of_shards", 3))
.execute()
.actionGet();
client()
.admin()
.cluster()
.prepareHealth()
.setWaitForEvents(Priority.LANGUID)
.setWaitForGreenStatus()
.execute()
.actionGet();
client()
.admin()
.cluster()
.prepareHealth()
.setWaitForEvents(Priority.LANGUID)
.setWaitForGreenStatus()
.execute()
.actionGet();
for (int i = 0; i < 100; i++) {
String routing = "0";
if (i > 90) {
routing = "1";
} else if (i > 60) {
routing = "2";
}
client()
.prepareIndex("test", "type1", Integer.toString(i))
.setSource("field", i)
.setRouting(routing)
.execute()
.actionGet();
}
client().admin().indices().prepareRefresh().execute().actionGet();
SearchResponse searchResponse =
client()
.prepareSearch()
.setSearchType(SearchType.QUERY_THEN_FETCH)
.setQuery(matchAllQuery())
.setSize(3)
.setScroll(TimeValue.timeValueMinutes(2))
.addSort("field", SortOrder.ASC)
.execute()
.actionGet();
try {
long counter = 0;
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(3));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
for (int i = 0; i < 32; i++) {
searchResponse =
client()
.prepareSearchScroll(searchResponse.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(3));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
}
// and now, the last one is one
searchResponse =
client()
.prepareSearchScroll(searchResponse.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(1));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
// a the last is zero
searchResponse =
client()
.prepareSearchScroll(searchResponse.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(0));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
} finally {
clearScroll(searchResponse.getScrollId());
}
}
/** @author kimchy (shay.banon) */
public class RollingRestartStressTest {
private final ESLogger logger = Loggers.getLogger(getClass());
private int numberOfShards = 5;
private int numberOfReplicas = 1;
private int numberOfNodes = 4;
private int textTokens = 150;
private int numberOfFields = 10;
private long initialNumberOfDocs = 100000;
private int indexers = 0;
private TimeValue indexerThrottle = TimeValue.timeValueMillis(100);
private Settings settings = ImmutableSettings.Builder.EMPTY_SETTINGS;
private TimeValue period = TimeValue.timeValueMinutes(20);
private boolean clearNodeData = true;
private Node client;
private AtomicLong indexCounter = new AtomicLong();
private AtomicLong idCounter = new AtomicLong();
public RollingRestartStressTest numberOfNodes(int numberOfNodes) {
this.numberOfNodes = numberOfNodes;
return this;
}
public RollingRestartStressTest numberOfShards(int numberOfShards) {
this.numberOfShards = numberOfShards;
return this;
}
public RollingRestartStressTest numberOfReplicas(int numberOfReplicas) {
this.numberOfReplicas = numberOfReplicas;
return this;
}
public RollingRestartStressTest initialNumberOfDocs(long initialNumberOfDocs) {
this.initialNumberOfDocs = initialNumberOfDocs;
return this;
}
public RollingRestartStressTest textTokens(int textTokens) {
this.textTokens = textTokens;
return this;
}
public RollingRestartStressTest numberOfFields(int numberOfFields) {
this.numberOfFields = numberOfFields;
return this;
}
public RollingRestartStressTest indexers(int indexers) {
this.indexers = indexers;
return this;
}
public RollingRestartStressTest indexerThrottle(TimeValue indexerThrottle) {
this.indexerThrottle = indexerThrottle;
return this;
}
public RollingRestartStressTest period(TimeValue period) {
this.period = period;
return this;
}
public RollingRestartStressTest cleanNodeData(boolean clearNodeData) {
this.clearNodeData = clearNodeData;
return this;
}
public RollingRestartStressTest settings(Settings settings) {
this.settings = settings;
return this;
}
public void run() throws Exception {
Node[] nodes = new Node[numberOfNodes];
for (int i = 0; i < nodes.length; i++) {
nodes[i] = NodeBuilder.nodeBuilder().settings(settings).node();
}
client = NodeBuilder.nodeBuilder().settings(settings).client(true).node();
client
.client()
.admin()
.indices()
.prepareCreate("test")
.setSettings(
settingsBuilder()
.put("index.number_of_shards", numberOfShards)
.put("index.number_of_replicas", numberOfReplicas))
.execute()
.actionGet();
logger.info("********** [START] INDEXING INITIAL DOCS");
for (long i = 0; i < initialNumberOfDocs; i++) {
indexDoc();
}
logger.info("********** [DONE ] INDEXING INITIAL DOCS");
Indexer[] indexerThreads = new Indexer[indexers];
for (int i = 0; i < indexerThreads.length; i++) {
indexerThreads[i] = new Indexer();
}
for (int i = 0; i < indexerThreads.length; i++) {
indexerThreads[i].start();
}
long testStart = System.currentTimeMillis();
// start doing the rolling restart
int nodeIndex = 0;
while (true) {
File[] nodeData =
((InternalNode) nodes[nodeIndex])
.injector()
.getInstance(NodeEnvironment.class)
.nodeDataLocations();
nodes[nodeIndex].close();
if (clearNodeData) {
FileSystemUtils.deleteRecursively(nodeData);
}
try {
ClusterHealthResponse clusterHealth =
client
.client()
.admin()
.cluster()
.prepareHealth()
.setWaitForGreenStatus()
.setWaitForNodes(Integer.toString(numberOfNodes + 0 /* client node*/))
.setWaitForRelocatingShards(0)
.setTimeout("10m")
.execute()
.actionGet();
if (clusterHealth.timedOut()) {
logger.warn("timed out waiting for green status....");
}
} catch (Exception e) {
logger.warn("failed to execute cluster health....");
}
nodes[nodeIndex] = NodeBuilder.nodeBuilder().settings(settings).node();
Thread.sleep(1000);
try {
ClusterHealthResponse clusterHealth =
client
.client()
.admin()
.cluster()
.prepareHealth()
.setWaitForGreenStatus()
.setWaitForNodes(Integer.toString(numberOfNodes + 1 /* client node*/))
.setWaitForRelocatingShards(0)
.setTimeout("10m")
.execute()
.actionGet();
if (clusterHealth.timedOut()) {
logger.warn("timed out waiting for green status....");
}
} catch (Exception e) {
logger.warn("failed to execute cluster health....");
}
if (++nodeIndex == nodes.length) {
nodeIndex = 0;
}
if ((System.currentTimeMillis() - testStart) > period.millis()) {
logger.info("test finished");
break;
}
}
for (int i = 0; i < indexerThreads.length; i++) {
indexerThreads[i].close = true;
}
Thread.sleep(indexerThrottle.millis() + 10000);
for (int i = 0; i < indexerThreads.length; i++) {
if (!indexerThreads[i].closed) {
logger.warn("thread not closed!");
}
}
client.client().admin().indices().prepareRefresh().execute().actionGet();
// check the status
IndicesStatusResponse status =
client.client().admin().indices().prepareStatus("test").execute().actionGet();
for (IndexShardStatus shardStatus : status.index("test")) {
ShardStatus shard = shardStatus.shards()[0];
logger.info("shard [{}], docs [{}]", shard.shardId(), shard.getDocs().numDocs());
for (ShardStatus shardStatu : shardStatus) {
if (shard.docs().numDocs() != shardStatu.docs().numDocs()) {
logger.warn(
"shard doc number does not match!, got {} and {}",
shard.docs().numDocs(),
shardStatu.docs().numDocs());
}
}
}
// check the count
for (int i = 0; i < (nodes.length * 5); i++) {
CountResponse count =
client.client().prepareCount().setQuery(matchAllQuery()).execute().actionGet();
logger.info(
"indexed [{}], count [{}], [{}]",
count.count(),
indexCounter.get(),
count.count() == indexCounter.get() ? "OK" : "FAIL");
if (count.count() != indexCounter.get()) {
logger.warn("count does not match!");
}
}
// scan all the docs, verify all have the same version based on the number of replicas
SearchResponse searchResponse =
client
.client()
.prepareSearch()
.setSearchType(SearchType.SCAN)
.setQuery(matchAllQuery())
.setSize(50)
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
logger.info("Verifying versions for {} hits...", searchResponse.hits().totalHits());
while (true) {
searchResponse =
client
.client()
.prepareSearchScroll(searchResponse.scrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
if (searchResponse.failedShards() > 0) {
logger.warn("Search Failures " + Arrays.toString(searchResponse.shardFailures()));
}
for (SearchHit hit : searchResponse.hits()) {
long version = -1;
for (int i = 0; i < (numberOfReplicas + 1); i++) {
GetResponse getResponse =
client.client().prepareGet(hit.index(), hit.type(), hit.id()).execute().actionGet();
if (version == -1) {
version = getResponse.version();
} else {
if (version != getResponse.version()) {
logger.warn(
"Doc {} has different version numbers {} and {}",
hit.id(),
version,
getResponse.version());
}
}
}
}
if (searchResponse.hits().hits().length == 0) {
break;
}
}
logger.info("Done verifying versions");
client.close();
for (Node node : nodes) {
node.close();
}
}
private class Indexer extends Thread {
volatile boolean close = false;
volatile boolean closed = false;
@Override
public void run() {
while (true) {
if (close) {
closed = true;
return;
}
try {
indexDoc();
Thread.sleep(indexerThrottle.millis());
} catch (Exception e) {
logger.warn("failed to index / sleep", e);
}
}
}
}
private void indexDoc() throws Exception {
StringBuilder sb = new StringBuilder();
XContentBuilder json =
XContentFactory.jsonBuilder()
.startObject()
.field("field", "value" + ThreadLocalRandom.current().nextInt());
int fields = Math.abs(ThreadLocalRandom.current().nextInt()) % numberOfFields;
for (int i = 0; i < fields; i++) {
json.field("num_" + i, ThreadLocalRandom.current().nextDouble());
int tokens = ThreadLocalRandom.current().nextInt() % textTokens;
sb.setLength(0);
for (int j = 0; j < tokens; j++) {
sb.append(UUID.randomBase64UUID()).append(' ');
}
json.field("text_" + i, sb.toString());
}
json.endObject();
String id = Long.toString(idCounter.incrementAndGet());
client
.client()
.prepareIndex("test", "type1", id)
.setCreate(true)
.setSource(json)
.execute()
.actionGet();
indexCounter.incrementAndGet();
}
public static void main(String[] args) throws Exception {
System.setProperty("es.logger.prefix", "");
Settings settings =
settingsBuilder()
.put("index.shard.check_index", true)
.put("gateway.type", "none")
.put("path.data", "data/data1,data/data2")
.build();
RollingRestartStressTest test =
new RollingRestartStressTest()
.settings(settings)
.numberOfNodes(4)
.numberOfShards(5)
.numberOfReplicas(1)
.initialNumberOfDocs(1000)
.textTokens(150)
.numberOfFields(10)
.cleanNodeData(false)
.indexers(5)
.indexerThrottle(TimeValue.timeValueMillis(50))
.period(TimeValue.timeValueMinutes(3));
test.run();
}
}
@Test
public void testSimpleScrollQueryThenFetch() throws Exception {
client()
.admin()
.indices()
.prepareCreate("test")
.setSettings(Settings.settingsBuilder().put("index.number_of_shards", 3))
.execute()
.actionGet();
client()
.admin()
.cluster()
.prepareHealth()
.setWaitForEvents(Priority.LANGUID)
.setWaitForGreenStatus()
.execute()
.actionGet();
client()
.admin()
.cluster()
.prepareHealth()
.setWaitForEvents(Priority.LANGUID)
.setWaitForGreenStatus()
.execute()
.actionGet();
for (int i = 0; i < 100; i++) {
client()
.prepareIndex("test", "type1", Integer.toString(i))
.setSource(jsonBuilder().startObject().field("field", i).endObject())
.execute()
.actionGet();
}
client().admin().indices().prepareRefresh().execute().actionGet();
SearchResponse searchResponse =
client()
.prepareSearch()
.setQuery(matchAllQuery())
.setSize(35)
.setScroll(TimeValue.timeValueMinutes(2))
.addSort("field", SortOrder.ASC)
.execute()
.actionGet();
try {
long counter = 0;
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(35));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
searchResponse =
client()
.prepareSearchScroll(searchResponse.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(35));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
searchResponse =
client()
.prepareSearchScroll(searchResponse.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
assertThat(searchResponse.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse.getHits().hits().length, equalTo(30));
for (SearchHit hit : searchResponse.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter++));
}
} finally {
clearScroll(searchResponse.getScrollId());
}
}
@Test
public void testScrollAndUpdateIndex() throws Exception {
client()
.admin()
.indices()
.prepareCreate("test")
.setSettings(Settings.settingsBuilder().put("index.number_of_shards", 5))
.execute()
.actionGet();
client()
.admin()
.cluster()
.prepareHealth()
.setWaitForEvents(Priority.LANGUID)
.setWaitForGreenStatus()
.execute()
.actionGet();
for (int i = 0; i < 500; i++) {
client()
.prepareIndex("test", "tweet", Integer.toString(i))
.setSource(
jsonBuilder()
.startObject()
.field("user", "kimchy")
.field("postDate", System.currentTimeMillis())
.field("message", "test")
.endObject())
.execute()
.actionGet();
}
client().admin().indices().prepareRefresh().execute().actionGet();
assertThat(
client().prepareCount().setQuery(matchAllQuery()).execute().actionGet().getCount(),
equalTo(500l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "test"))
.execute()
.actionGet()
.getCount(),
equalTo(500l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "test"))
.execute()
.actionGet()
.getCount(),
equalTo(500l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "update"))
.execute()
.actionGet()
.getCount(),
equalTo(0l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "update"))
.execute()
.actionGet()
.getCount(),
equalTo(0l));
SearchResponse searchResponse =
client()
.prepareSearch()
.setQuery(queryStringQuery("user:kimchy"))
.setSize(35)
.setScroll(TimeValue.timeValueMinutes(2))
.addSort("postDate", SortOrder.ASC)
.execute()
.actionGet();
try {
do {
for (SearchHit searchHit : searchResponse.getHits().hits()) {
Map<String, Object> map = searchHit.sourceAsMap();
map.put("message", "update");
client()
.prepareIndex("test", "tweet", searchHit.id())
.setSource(map)
.execute()
.actionGet();
}
searchResponse =
client()
.prepareSearchScroll(searchResponse.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
} while (searchResponse.getHits().hits().length > 0);
client().admin().indices().prepareRefresh().execute().actionGet();
assertThat(
client().prepareCount().setQuery(matchAllQuery()).execute().actionGet().getCount(),
equalTo(500l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "test"))
.execute()
.actionGet()
.getCount(),
equalTo(0l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "test"))
.execute()
.actionGet()
.getCount(),
equalTo(0l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "update"))
.execute()
.actionGet()
.getCount(),
equalTo(500l));
assertThat(
client()
.prepareCount()
.setQuery(termQuery("message", "update"))
.execute()
.actionGet()
.getCount(),
equalTo(500l));
} finally {
clearScroll(searchResponse.getScrollId());
}
}
/** All indices has to be created before! */
public void mergeIndices(
Collection<String> indexList,
String intoIndex,
int hitsPerPage,
boolean forceRefresh,
CreateObjectsInterface<T> createObj,
FilterBuilder additionalFilter) {
if (forceRefresh) {
refresh(indexList);
refresh(intoIndex);
}
int keepTime = 100;
for (String fromIndex : indexList) {
SearchRequestBuilder srb =
client
.prepareSearch(fromIndex)
.setVersion(true)
.setQuery(QueryBuilders.matchAllQuery())
.setSize(hitsPerPage)
.setSearchType(SearchType.SCAN)
.setScroll(TimeValue.timeValueMinutes(keepTime));
if (additionalFilter != null) srb.setFilter(additionalFilter);
SearchResponse rsp = srb.execute().actionGet();
try {
long total = rsp.hits().totalHits();
int collectedResults = 0;
while (true) {
StopWatch queryWatch = new StopWatch().start();
rsp =
client
.prepareSearchScroll(rsp.scrollId())
.setScroll(TimeValue.timeValueMinutes(keepTime))
.execute()
.actionGet();
long currentResults = rsp.hits().hits().length;
if (currentResults == 0) break;
queryWatch.stop();
Collection<T> objs = createObj.collectObjects(rsp);
StopWatch updateWatch = new StopWatch().start();
int failed = bulkUpdate(objs, intoIndex, false, false).size();
// trying to enable flushing to avoid memory issues on the server side?
flush(intoIndex);
updateWatch.stop();
collectedResults += currentResults;
logger.info(
"Progress "
+ collectedResults
+ "/"
+ total
+ " fromIndex="
+ fromIndex
+ " update:"
+ updateWatch.totalTime().getSeconds()
+ " query:"
+ queryWatch.totalTime().getSeconds()
+ " failed:"
+ failed);
}
logger.info(
"Finished copying of index:"
+ fromIndex
+ ". Total:"
+ total
+ " collected:"
+ collectedResults);
} catch (Exception ex) {
// throw new RuntimeException(ex);
logger.error(
"Failed to copy data from index " + fromIndex + " into " + intoIndex + ".", ex);
}
}
if (forceRefresh) refresh(intoIndex);
}
@Test
public void testSimpleScrollQueryThenFetch_clearAllScrollIds() throws Exception {
client()
.admin()
.indices()
.prepareCreate("test")
.setSettings(Settings.settingsBuilder().put("index.number_of_shards", 3))
.execute()
.actionGet();
client()
.admin()
.cluster()
.prepareHealth()
.setWaitForEvents(Priority.LANGUID)
.setWaitForGreenStatus()
.execute()
.actionGet();
client()
.admin()
.cluster()
.prepareHealth()
.setWaitForEvents(Priority.LANGUID)
.setWaitForGreenStatus()
.execute()
.actionGet();
for (int i = 0; i < 100; i++) {
client()
.prepareIndex("test", "type1", Integer.toString(i))
.setSource(jsonBuilder().startObject().field("field", i).endObject())
.execute()
.actionGet();
}
client().admin().indices().prepareRefresh().execute().actionGet();
SearchResponse searchResponse1 =
client()
.prepareSearch()
.setQuery(matchAllQuery())
.setSize(35)
.setScroll(TimeValue.timeValueMinutes(2))
.setSearchType(SearchType.QUERY_THEN_FETCH)
.addSort("field", SortOrder.ASC)
.execute()
.actionGet();
SearchResponse searchResponse2 =
client()
.prepareSearch()
.setQuery(matchAllQuery())
.setSize(35)
.setScroll(TimeValue.timeValueMinutes(2))
.setSearchType(SearchType.QUERY_THEN_FETCH)
.addSort("field", SortOrder.ASC)
.execute()
.actionGet();
long counter1 = 0;
long counter2 = 0;
assertThat(searchResponse1.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse1.getHits().hits().length, equalTo(35));
for (SearchHit hit : searchResponse1.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter1++));
}
assertThat(searchResponse2.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse2.getHits().hits().length, equalTo(35));
for (SearchHit hit : searchResponse2.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter2++));
}
searchResponse1 =
client()
.prepareSearchScroll(searchResponse1.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
searchResponse2 =
client()
.prepareSearchScroll(searchResponse2.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
assertThat(searchResponse1.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse1.getHits().hits().length, equalTo(35));
for (SearchHit hit : searchResponse1.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter1++));
}
assertThat(searchResponse2.getHits().getTotalHits(), equalTo(100l));
assertThat(searchResponse2.getHits().hits().length, equalTo(35));
for (SearchHit hit : searchResponse2.getHits()) {
assertThat(((Number) hit.sortValues()[0]).longValue(), equalTo(counter2++));
}
ClearScrollResponse clearResponse =
client().prepareClearScroll().addScrollId("_all").execute().actionGet();
assertThat(clearResponse.isSucceeded(), is(true));
assertThat(clearResponse.getNumFreed(), greaterThan(0));
assertThat(clearResponse.status(), equalTo(RestStatus.OK));
assertThrows(
internalCluster()
.transportClient()
.prepareSearchScroll(searchResponse1.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2)),
RestStatus.NOT_FOUND);
assertThrows(
internalCluster()
.transportClient()
.prepareSearchScroll(searchResponse2.getScrollId())
.setScroll(TimeValue.timeValueMinutes(2)),
RestStatus.NOT_FOUND);
}
public class RecoverySettings extends AbstractComponent {
public static final Setting<ByteSizeValue> INDICES_RECOVERY_MAX_BYTES_PER_SEC_SETTING =
Setting.byteSizeSetting(
"indices.recovery.max_bytes_per_sec",
new ByteSizeValue(40, ByteSizeUnit.MB),
Property.Dynamic,
Property.NodeScope);
/**
* how long to wait before retrying after issues cause by cluster state syncing between nodes
* i.e., local node is not yet known on remote node, remote shard not yet started etc.
*/
public static final Setting<TimeValue> INDICES_RECOVERY_RETRY_DELAY_STATE_SYNC_SETTING =
Setting.positiveTimeSetting(
"indices.recovery.retry_delay_state_sync",
TimeValue.timeValueMillis(500),
Property.Dynamic,
Property.NodeScope);
/** how long to wait before retrying after network related issues */
public static final Setting<TimeValue> INDICES_RECOVERY_RETRY_DELAY_NETWORK_SETTING =
Setting.positiveTimeSetting(
"indices.recovery.retry_delay_network",
TimeValue.timeValueSeconds(5),
Property.Dynamic,
Property.NodeScope);
/** timeout value to use for requests made as part of the recovery process */
public static final Setting<TimeValue> INDICES_RECOVERY_INTERNAL_ACTION_TIMEOUT_SETTING =
Setting.positiveTimeSetting(
"indices.recovery.internal_action_timeout",
TimeValue.timeValueMinutes(15),
Property.Dynamic,
Property.NodeScope);
/**
* timeout value to use for requests made as part of the recovery process that are expected to
* take long time. defaults to twice `indices.recovery.internal_action_timeout`.
*/
public static final Setting<TimeValue> INDICES_RECOVERY_INTERNAL_LONG_ACTION_TIMEOUT_SETTING =
Setting.timeSetting(
"indices.recovery.internal_action_long_timeout",
(s) ->
TimeValue.timeValueMillis(
INDICES_RECOVERY_INTERNAL_ACTION_TIMEOUT_SETTING.get(s).millis() * 2),
TimeValue.timeValueSeconds(0),
Property.Dynamic,
Property.NodeScope);
/**
* recoveries that don't show any activity for more then this interval will be failed. defaults to
* `indices.recovery.internal_action_long_timeout`
*/
public static final Setting<TimeValue> INDICES_RECOVERY_ACTIVITY_TIMEOUT_SETTING =
Setting.timeSetting(
"indices.recovery.recovery_activity_timeout",
INDICES_RECOVERY_INTERNAL_LONG_ACTION_TIMEOUT_SETTING::get,
TimeValue.timeValueSeconds(0),
Property.Dynamic,
Property.NodeScope);
public static final ByteSizeValue DEFAULT_CHUNK_SIZE = new ByteSizeValue(512, ByteSizeUnit.KB);
private volatile ByteSizeValue maxBytesPerSec;
private volatile SimpleRateLimiter rateLimiter;
private volatile TimeValue retryDelayStateSync;
private volatile TimeValue retryDelayNetwork;
private volatile TimeValue activityTimeout;
private volatile TimeValue internalActionTimeout;
private volatile TimeValue internalActionLongTimeout;
private volatile ByteSizeValue chunkSize = DEFAULT_CHUNK_SIZE;
@Inject
public RecoverySettings(Settings settings, ClusterSettings clusterSettings) {
super(settings);
this.retryDelayStateSync = INDICES_RECOVERY_RETRY_DELAY_STATE_SYNC_SETTING.get(settings);
// doesn't have to be fast as nodes are reconnected every 10s by default (see
// InternalClusterService.ReconnectToNodes)
// and we want to give the master time to remove a faulty node
this.retryDelayNetwork = INDICES_RECOVERY_RETRY_DELAY_NETWORK_SETTING.get(settings);
this.internalActionTimeout = INDICES_RECOVERY_INTERNAL_ACTION_TIMEOUT_SETTING.get(settings);
this.internalActionLongTimeout =
INDICES_RECOVERY_INTERNAL_LONG_ACTION_TIMEOUT_SETTING.get(settings);
this.activityTimeout = INDICES_RECOVERY_ACTIVITY_TIMEOUT_SETTING.get(settings);
this.maxBytesPerSec = INDICES_RECOVERY_MAX_BYTES_PER_SEC_SETTING.get(settings);
if (maxBytesPerSec.getBytes() <= 0) {
rateLimiter = null;
} else {
rateLimiter = new SimpleRateLimiter(maxBytesPerSec.getMbFrac());
}
logger.debug("using max_bytes_per_sec[{}]", maxBytesPerSec);
clusterSettings.addSettingsUpdateConsumer(
INDICES_RECOVERY_MAX_BYTES_PER_SEC_SETTING, this::setMaxBytesPerSec);
clusterSettings.addSettingsUpdateConsumer(
INDICES_RECOVERY_RETRY_DELAY_STATE_SYNC_SETTING, this::setRetryDelayStateSync);
clusterSettings.addSettingsUpdateConsumer(
INDICES_RECOVERY_RETRY_DELAY_NETWORK_SETTING, this::setRetryDelayNetwork);
clusterSettings.addSettingsUpdateConsumer(
INDICES_RECOVERY_INTERNAL_ACTION_TIMEOUT_SETTING, this::setInternalActionTimeout);
clusterSettings.addSettingsUpdateConsumer(
INDICES_RECOVERY_INTERNAL_LONG_ACTION_TIMEOUT_SETTING, this::setInternalActionLongTimeout);
clusterSettings.addSettingsUpdateConsumer(
INDICES_RECOVERY_ACTIVITY_TIMEOUT_SETTING, this::setActivityTimeout);
}
public RateLimiter rateLimiter() {
return rateLimiter;
}
public TimeValue retryDelayNetwork() {
return retryDelayNetwork;
}
public TimeValue retryDelayStateSync() {
return retryDelayStateSync;
}
public TimeValue activityTimeout() {
return activityTimeout;
}
public TimeValue internalActionTimeout() {
return internalActionTimeout;
}
public TimeValue internalActionLongTimeout() {
return internalActionLongTimeout;
}
public ByteSizeValue getChunkSize() {
return chunkSize;
}
void setChunkSize(ByteSizeValue chunkSize) { // only settable for tests
if (chunkSize.bytesAsInt() <= 0) {
throw new IllegalArgumentException("chunkSize must be > 0");
}
this.chunkSize = chunkSize;
}
public void setRetryDelayStateSync(TimeValue retryDelayStateSync) {
this.retryDelayStateSync = retryDelayStateSync;
}
public void setRetryDelayNetwork(TimeValue retryDelayNetwork) {
this.retryDelayNetwork = retryDelayNetwork;
}
public void setActivityTimeout(TimeValue activityTimeout) {
this.activityTimeout = activityTimeout;
}
public void setInternalActionTimeout(TimeValue internalActionTimeout) {
this.internalActionTimeout = internalActionTimeout;
}
public void setInternalActionLongTimeout(TimeValue internalActionLongTimeout) {
this.internalActionLongTimeout = internalActionLongTimeout;
}
private void setMaxBytesPerSec(ByteSizeValue maxBytesPerSec) {
this.maxBytesPerSec = maxBytesPerSec;
if (maxBytesPerSec.getBytes() <= 0) {
rateLimiter = null;
} else if (rateLimiter != null) {
rateLimiter.setMBPerSec(maxBytesPerSec.getMbFrac());
} else {
rateLimiter = new SimpleRateLimiter(maxBytesPerSec.getMbFrac());
}
}
}
public void run() throws Exception {
Node[] nodes = new Node[numberOfNodes];
for (int i = 0; i < nodes.length; i++) {
nodes[i] = NodeBuilder.nodeBuilder().settings(settings).node();
}
client = NodeBuilder.nodeBuilder().settings(settings).client(true).node();
client
.client()
.admin()
.indices()
.prepareCreate("test")
.setSettings(
settingsBuilder()
.put("index.number_of_shards", numberOfShards)
.put("index.number_of_replicas", numberOfReplicas))
.execute()
.actionGet();
logger.info("********** [START] INDEXING INITIAL DOCS");
for (long i = 0; i < initialNumberOfDocs; i++) {
indexDoc();
}
logger.info("********** [DONE ] INDEXING INITIAL DOCS");
Indexer[] indexerThreads = new Indexer[indexers];
for (int i = 0; i < indexerThreads.length; i++) {
indexerThreads[i] = new Indexer();
}
for (int i = 0; i < indexerThreads.length; i++) {
indexerThreads[i].start();
}
long testStart = System.currentTimeMillis();
// start doing the rolling restart
int nodeIndex = 0;
while (true) {
File[] nodeData =
((InternalNode) nodes[nodeIndex])
.injector()
.getInstance(NodeEnvironment.class)
.nodeDataLocations();
nodes[nodeIndex].close();
if (clearNodeData) {
FileSystemUtils.deleteRecursively(nodeData);
}
try {
ClusterHealthResponse clusterHealth =
client
.client()
.admin()
.cluster()
.prepareHealth()
.setWaitForGreenStatus()
.setWaitForNodes(Integer.toString(numberOfNodes + 0 /* client node*/))
.setWaitForRelocatingShards(0)
.setTimeout("10m")
.execute()
.actionGet();
if (clusterHealth.timedOut()) {
logger.warn("timed out waiting for green status....");
}
} catch (Exception e) {
logger.warn("failed to execute cluster health....");
}
nodes[nodeIndex] = NodeBuilder.nodeBuilder().settings(settings).node();
Thread.sleep(1000);
try {
ClusterHealthResponse clusterHealth =
client
.client()
.admin()
.cluster()
.prepareHealth()
.setWaitForGreenStatus()
.setWaitForNodes(Integer.toString(numberOfNodes + 1 /* client node*/))
.setWaitForRelocatingShards(0)
.setTimeout("10m")
.execute()
.actionGet();
if (clusterHealth.timedOut()) {
logger.warn("timed out waiting for green status....");
}
} catch (Exception e) {
logger.warn("failed to execute cluster health....");
}
if (++nodeIndex == nodes.length) {
nodeIndex = 0;
}
if ((System.currentTimeMillis() - testStart) > period.millis()) {
logger.info("test finished");
break;
}
}
for (int i = 0; i < indexerThreads.length; i++) {
indexerThreads[i].close = true;
}
Thread.sleep(indexerThrottle.millis() + 10000);
for (int i = 0; i < indexerThreads.length; i++) {
if (!indexerThreads[i].closed) {
logger.warn("thread not closed!");
}
}
client.client().admin().indices().prepareRefresh().execute().actionGet();
// check the status
IndicesStatusResponse status =
client.client().admin().indices().prepareStatus("test").execute().actionGet();
for (IndexShardStatus shardStatus : status.index("test")) {
ShardStatus shard = shardStatus.shards()[0];
logger.info("shard [{}], docs [{}]", shard.shardId(), shard.getDocs().numDocs());
for (ShardStatus shardStatu : shardStatus) {
if (shard.docs().numDocs() != shardStatu.docs().numDocs()) {
logger.warn(
"shard doc number does not match!, got {} and {}",
shard.docs().numDocs(),
shardStatu.docs().numDocs());
}
}
}
// check the count
for (int i = 0; i < (nodes.length * 5); i++) {
CountResponse count =
client.client().prepareCount().setQuery(matchAllQuery()).execute().actionGet();
logger.info(
"indexed [{}], count [{}], [{}]",
count.count(),
indexCounter.get(),
count.count() == indexCounter.get() ? "OK" : "FAIL");
if (count.count() != indexCounter.get()) {
logger.warn("count does not match!");
}
}
// scan all the docs, verify all have the same version based on the number of replicas
SearchResponse searchResponse =
client
.client()
.prepareSearch()
.setSearchType(SearchType.SCAN)
.setQuery(matchAllQuery())
.setSize(50)
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
logger.info("Verifying versions for {} hits...", searchResponse.hits().totalHits());
while (true) {
searchResponse =
client
.client()
.prepareSearchScroll(searchResponse.scrollId())
.setScroll(TimeValue.timeValueMinutes(2))
.execute()
.actionGet();
if (searchResponse.failedShards() > 0) {
logger.warn("Search Failures " + Arrays.toString(searchResponse.shardFailures()));
}
for (SearchHit hit : searchResponse.hits()) {
long version = -1;
for (int i = 0; i < (numberOfReplicas + 1); i++) {
GetResponse getResponse =
client.client().prepareGet(hit.index(), hit.type(), hit.id()).execute().actionGet();
if (version == -1) {
version = getResponse.version();
} else {
if (version != getResponse.version()) {
logger.warn(
"Doc {} has different version numbers {} and {}",
hit.id(),
version,
getResponse.version());
}
}
}
}
if (searchResponse.hits().hits().length == 0) {
break;
}
}
logger.info("Done verifying versions");
client.close();
for (Node node : nodes) {
node.close();
}
}
@Override
public TimeValue defaultValue() {
return TimeValue.timeValueMinutes(30);
}