/**
* Construct a scaling executor builder; the settings will have the specified key prefix.
*
* @param name the name of the executor
* @param core the minimum number of threads in the pool
* @param max the maximum number of threads in the pool
* @param keepAlive the time that spare threads above {@code core} threads will be kept alive
* @param prefix the prefix for the settings keys
*/
public ScalingExecutorBuilder(
final String name,
final int core,
final int max,
final TimeValue keepAlive,
final String prefix) {
super(name);
this.coreSetting =
Setting.intSetting(settingsKey(prefix, "core"), core, Setting.Property.NodeScope);
this.maxSetting =
Setting.intSetting(settingsKey(prefix, "max"), max, Setting.Property.NodeScope);
this.keepAliveSetting =
Setting.timeSetting(
settingsKey(prefix, "keep_alive"), keepAlive, Setting.Property.NodeScope);
}
/**
* Similar to the {@link ClusterRebalanceAllocationDecider} this {@link AllocationDecider} controls
* the number of currently in-progress re-balance (relocation) operations and restricts node
* allocations if the configured threshold is reached. The default number of concurrent rebalance
* operations is set to <tt>2</tt>
*
* <p>Re-balance operations can be controlled in real-time via the cluster update API using
* <tt>cluster.routing.allocation.cluster_concurrent_rebalance</tt>. Iff this setting is set to
* <tt>-1</tt> the number of concurrent re-balance operations are unlimited.
*/
public class ConcurrentRebalanceAllocationDecider extends AllocationDecider {
public static final String NAME = "concurrent_rebalance";
public static final Setting<Integer>
CLUSTER_ROUTING_ALLOCATION_CLUSTER_CONCURRENT_REBALANCE_SETTING =
Setting.intSetting(
"cluster.routing.allocation.cluster_concurrent_rebalance",
2,
-1,
Property.Dynamic,
Property.NodeScope);
private volatile int clusterConcurrentRebalance;
public ConcurrentRebalanceAllocationDecider(Settings settings, ClusterSettings clusterSettings) {
super(settings);
this.clusterConcurrentRebalance =
CLUSTER_ROUTING_ALLOCATION_CLUSTER_CONCURRENT_REBALANCE_SETTING.get(settings);
logger.debug("using [cluster_concurrent_rebalance] with [{}]", clusterConcurrentRebalance);
clusterSettings.addSettingsUpdateConsumer(
CLUSTER_ROUTING_ALLOCATION_CLUSTER_CONCURRENT_REBALANCE_SETTING,
this::setClusterConcurrentRebalance);
}
private void setClusterConcurrentRebalance(int concurrentRebalance) {
clusterConcurrentRebalance = concurrentRebalance;
}
@Override
public Decision canRebalance(ShardRouting shardRouting, RoutingAllocation allocation) {
if (clusterConcurrentRebalance == -1) {
return allocation.decision(Decision.YES, NAME, "unlimited concurrent rebalances are allowed");
}
int relocatingShards = allocation.routingNodes().getRelocatingShardCount();
if (relocatingShards >= clusterConcurrentRebalance) {
return allocation.decision(
Decision.NO,
NAME,
"too many shards are concurrently rebalancing [%d], limit: [%d]",
relocatingShards,
clusterConcurrentRebalance);
}
return allocation.decision(
Decision.YES,
NAME,
"below threshold [%d] for concurrent rebalances, current rebalance shard count [%d]",
clusterConcurrentRebalance,
relocatingShards);
}
}
public class EsExecutors {
/**
* Settings key to manually set the number of available processors. This is used to adjust thread
* pools sizes etc. per node.
*/
public static final Setting<Integer> PROCESSORS_SETTING =
Setting.intSetting(
"processors", Runtime.getRuntime().availableProcessors(), 1, Property.NodeScope);
/**
* Returns the number of available processors. Defaults to {@link Runtime#availableProcessors()}
* but can be overridden by passing a {@link Settings} instance with the key "processors" set to
* the desired value.
*
* @param settings a {@link Settings} instance from which to derive the available processors
* @return the number of available processors
*/
public static int numberOfProcessors(final Settings settings) {
return PROCESSORS_SETTING.get(settings);
}
public static PrioritizedEsThreadPoolExecutor newSinglePrioritizing(
String name, ThreadFactory threadFactory, ThreadContext contextHolder) {
return new PrioritizedEsThreadPoolExecutor(
name, 1, 1, 0L, TimeUnit.MILLISECONDS, threadFactory, contextHolder);
}
public static EsThreadPoolExecutor newScaling(
String name,
int min,
int max,
long keepAliveTime,
TimeUnit unit,
ThreadFactory threadFactory,
ThreadContext contextHolder) {
ExecutorScalingQueue<Runnable> queue = new ExecutorScalingQueue<>();
EsThreadPoolExecutor executor =
new EsThreadPoolExecutor(
name,
min,
max,
keepAliveTime,
unit,
queue,
threadFactory,
new ForceQueuePolicy(),
contextHolder);
queue.executor = executor;
return executor;
}
public static EsThreadPoolExecutor newFixed(
String name,
int size,
int queueCapacity,
ThreadFactory threadFactory,
ThreadContext contextHolder) {
BlockingQueue<Runnable> queue;
if (queueCapacity < 0) {
queue = ConcurrentCollections.newBlockingQueue();
} else {
queue =
new SizeBlockingQueue<>(
ConcurrentCollections.<Runnable>newBlockingQueue(), queueCapacity);
}
return new EsThreadPoolExecutor(
name,
size,
size,
0,
TimeUnit.MILLISECONDS,
queue,
threadFactory,
new EsAbortPolicy(),
contextHolder);
}
public static String threadName(Settings settings, String... names) {
String namePrefix =
Arrays.stream(names)
.filter(name -> name != null)
.collect(Collectors.joining(".", "[", "]"));
return threadName(settings, namePrefix);
}
public static String threadName(Settings settings, String namePrefix) {
if (Node.NODE_NAME_SETTING.exists(settings)) {
return threadName(Node.NODE_NAME_SETTING.get(settings), namePrefix);
} else {
return threadName("", namePrefix);
}
}
public static String threadName(final String nodeName, final String namePrefix) {
return "elasticsearch"
+ (nodeName.isEmpty() ? "" : "[")
+ nodeName
+ (nodeName.isEmpty() ? "" : "]")
+ "["
+ namePrefix
+ "]";
}
public static ThreadFactory daemonThreadFactory(Settings settings, String namePrefix) {
return daemonThreadFactory(threadName(settings, namePrefix));
}
public static ThreadFactory daemonThreadFactory(Settings settings, String... names) {
return daemonThreadFactory(threadName(settings, names));
}
public static ThreadFactory daemonThreadFactory(String namePrefix) {
return new EsThreadFactory(namePrefix);
}
static class EsThreadFactory implements ThreadFactory {
final ThreadGroup group;
final AtomicInteger threadNumber = new AtomicInteger(1);
final String namePrefix;
public EsThreadFactory(String namePrefix) {
this.namePrefix = namePrefix;
SecurityManager s = System.getSecurityManager();
group = (s != null) ? s.getThreadGroup() : Thread.currentThread().getThreadGroup();
}
@Override
public Thread newThread(Runnable r) {
Thread t = new Thread(group, r, namePrefix + "[T#" + threadNumber.getAndIncrement() + "]", 0);
t.setDaemon(true);
return t;
}
}
/** Cannot instantiate. */
private EsExecutors() {}
static class ExecutorScalingQueue<E> extends LinkedTransferQueue<E> {
ThreadPoolExecutor executor;
public ExecutorScalingQueue() {}
@Override
public boolean offer(E e) {
// first try to transfer to a waiting worker thread
if (!tryTransfer(e)) {
// check if there might be spare capacity in the thread
// pool executor
int left = executor.getMaximumPoolSize() - executor.getCorePoolSize();
if (left > 0) {
// reject queuing the task to force the thread pool
// executor to add a worker if it can; combined
// with ForceQueuePolicy, this causes the thread
// pool to always scale up to max pool size and we
// only queue when there is no spare capacity
return false;
} else {
return super.offer(e);
}
} else {
return true;
}
}
}
/**
* A handler for rejected tasks that adds the specified element to this queue, waiting if
* necessary for space to become available.
*/
static class ForceQueuePolicy implements XRejectedExecutionHandler {
@Override
public void rejectedExecution(Runnable r, ThreadPoolExecutor executor) {
try {
executor.getQueue().put(r);
} catch (InterruptedException e) {
// should never happen since we never wait
throw new EsRejectedExecutionException(e);
}
}
@Override
public long rejected() {
return 0;
}
}
}
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;
}
}
public final class HttpTransportSettings {
public static final Setting<Boolean> SETTING_CORS_ENABLED =
Setting.boolSetting("http.cors.enabled", false, false, Scope.CLUSTER);
public static final Setting<String> SETTING_CORS_ALLOW_ORIGIN =
new Setting<String>("http.cors.allow-origin", "", (value) -> value, false, Scope.CLUSTER);
public static final Setting<Integer> SETTING_CORS_MAX_AGE =
Setting.intSetting("http.cors.max-age", 1728000, false, Scope.CLUSTER);
public static final Setting<String> SETTING_CORS_ALLOW_METHODS =
new Setting<String>(
"http.cors.allow-methods",
"OPTIONS, HEAD, GET, POST, PUT, DELETE",
(value) -> value,
false,
Scope.CLUSTER);
public static final Setting<String> SETTING_CORS_ALLOW_HEADERS =
new Setting<String>(
"http.cors.allow-headers",
"X-Requested-With, Content-Type, Content-Length",
(value) -> value,
false,
Scope.CLUSTER);
public static final Setting<Boolean> SETTING_CORS_ALLOW_CREDENTIALS =
Setting.boolSetting("http.cors.allow-credentials", false, false, Scope.CLUSTER);
public static final Setting<Boolean> SETTING_PIPELINING =
Setting.boolSetting("http.pipelining", true, false, Scope.CLUSTER);
public static final Setting<Integer> SETTING_PIPELINING_MAX_EVENTS =
Setting.intSetting("http.pipelining.max_events", 10000, false, Scope.CLUSTER);
public static final Setting<Boolean> SETTING_HTTP_COMPRESSION =
Setting.boolSetting("http.compression", false, false, Scope.CLUSTER);
public static final Setting<Integer> SETTING_HTTP_COMPRESSION_LEVEL =
Setting.intSetting("http.compression_level", 6, false, Scope.CLUSTER);
public static final Setting<List<String>> SETTING_HTTP_HOST =
listSetting("http.host", emptyList(), s -> s, false, Scope.CLUSTER);
public static final Setting<List<String>> SETTING_HTTP_PUBLISH_HOST =
listSetting("http.publish_host", SETTING_HTTP_HOST, s -> s, false, Scope.CLUSTER);
public static final Setting<List<String>> SETTING_HTTP_BIND_HOST =
listSetting("http.bind_host", SETTING_HTTP_HOST, s -> s, false, Scope.CLUSTER);
public static final Setting<PortsRange> SETTING_HTTP_PORT =
new Setting<PortsRange>("http.port", "9200-9300", PortsRange::new, false, Scope.CLUSTER);
public static final Setting<Integer> SETTING_HTTP_PUBLISH_PORT =
Setting.intSetting("http.publish_port", 0, 0, false, Scope.CLUSTER);
public static final Setting<Boolean> SETTING_HTTP_DETAILED_ERRORS_ENABLED =
Setting.boolSetting("http.detailed_errors.enabled", true, false, Scope.CLUSTER);
public static final Setting<ByteSizeValue> SETTING_HTTP_MAX_CONTENT_LENGTH =
Setting.byteSizeSetting(
"http.max_content_length", new ByteSizeValue(100, ByteSizeUnit.MB), false, Scope.CLUSTER);
public static final Setting<ByteSizeValue> SETTING_HTTP_MAX_CHUNK_SIZE =
Setting.byteSizeSetting(
"http.max_chunk_size", new ByteSizeValue(8, ByteSizeUnit.KB), false, Scope.CLUSTER);
public static final Setting<ByteSizeValue> SETTING_HTTP_MAX_HEADER_SIZE =
Setting.byteSizeSetting(
"http.max_header_size", new ByteSizeValue(8, ByteSizeUnit.KB), false, Scope.CLUSTER);
public static final Setting<ByteSizeValue> SETTING_HTTP_MAX_INITIAL_LINE_LENGTH =
Setting.byteSizeSetting(
"http.max_initial_line_length",
new ByteSizeValue(4, ByteSizeUnit.KB),
false,
Scope.CLUSTER);
// don't reset cookies by default, since I don't think we really need to
// note, parsing cookies was fixed in netty 3.5.1 regarding stack allocation, but still,
// currently, we don't need cookies
public static final Setting<Boolean> SETTING_HTTP_RESET_COOKIES =
Setting.boolSetting("http.reset_cookies", false, false, Scope.CLUSTER);
private HttpTransportSettings() {}
}
/**
* A base class for {@link org.elasticsearch.discovery.zen.fd.MasterFaultDetection} & {@link
* org.elasticsearch.discovery.zen.fd.NodesFaultDetection}, making sure both use the same setting.
*/
public abstract class FaultDetection extends AbstractComponent {
public static final Setting<Boolean> CONNECT_ON_NETWORK_DISCONNECT_SETTING =
Setting.boolSetting(
"discovery.zen.fd.connect_on_network_disconnect", false, Property.NodeScope);
public static final Setting<TimeValue> PING_INTERVAL_SETTING =
Setting.positiveTimeSetting(
"discovery.zen.fd.ping_interval", timeValueSeconds(1), Property.NodeScope);
public static final Setting<TimeValue> PING_TIMEOUT_SETTING =
Setting.timeSetting(
"discovery.zen.fd.ping_timeout", timeValueSeconds(30), Property.NodeScope);
public static final Setting<Integer> PING_RETRIES_SETTING =
Setting.intSetting("discovery.zen.fd.ping_retries", 3, Property.NodeScope);
public static final Setting<Boolean> REGISTER_CONNECTION_LISTENER_SETTING =
Setting.boolSetting(
"discovery.zen.fd.register_connection_listener", true, Property.NodeScope);
protected final ThreadPool threadPool;
protected final ClusterName clusterName;
protected final TransportService transportService;
// used mainly for testing, should always be true
protected final boolean registerConnectionListener;
protected final FDConnectionListener connectionListener;
protected final boolean connectOnNetworkDisconnect;
protected final TimeValue pingInterval;
protected final TimeValue pingRetryTimeout;
protected final int pingRetryCount;
public FaultDetection(
Settings settings,
ThreadPool threadPool,
TransportService transportService,
ClusterName clusterName) {
super(settings);
this.threadPool = threadPool;
this.transportService = transportService;
this.clusterName = clusterName;
this.connectOnNetworkDisconnect = CONNECT_ON_NETWORK_DISCONNECT_SETTING.get(settings);
this.pingInterval = PING_INTERVAL_SETTING.get(settings);
this.pingRetryTimeout = PING_TIMEOUT_SETTING.get(settings);
this.pingRetryCount = PING_RETRIES_SETTING.get(settings);
this.registerConnectionListener = REGISTER_CONNECTION_LISTENER_SETTING.get(settings);
this.connectionListener = new FDConnectionListener();
if (registerConnectionListener) {
transportService.addConnectionListener(connectionListener);
}
}
public void close() {
transportService.removeConnectionListener(connectionListener);
}
/**
* This method will be called when the {@link org.elasticsearch.transport.TransportService} raised
* a node disconnected event
*/
abstract void handleTransportDisconnect(DiscoveryNode node);
private class FDConnectionListener implements TransportConnectionListener {
@Override
public void onNodeConnected(DiscoveryNode node) {}
@Override
public void onNodeDisconnected(DiscoveryNode node) {
handleTransportDisconnect(node);
}
}
}
public class IndexMetaData
implements Diffable<IndexMetaData>, FromXContentBuilder<IndexMetaData>, ToXContent {
public interface Custom extends Diffable<Custom>, ToXContent {
String type();
Custom fromMap(Map<String, Object> map) throws IOException;
Custom fromXContent(XContentParser parser) throws IOException;
/**
* Merges from this to another, with this being more important, i.e., if something exists in
* this and another, this will prevail.
*/
Custom mergeWith(Custom another);
}
public static Map<String, Custom> customPrototypes = new HashMap<>();
/** Register a custom index meta data factory. Make sure to call it from a static block. */
public static void registerPrototype(String type, Custom proto) {
customPrototypes.put(type, proto);
}
@Nullable
public static <T extends Custom> T lookupPrototype(String type) {
//noinspection unchecked
return (T) customPrototypes.get(type);
}
public static <T extends Custom> T lookupPrototypeSafe(String type) {
//noinspection unchecked
T proto = (T) customPrototypes.get(type);
if (proto == null) {
throw new IllegalArgumentException(
"No custom metadata prototype registered for type [" + type + "]");
}
return proto;
}
public static final ClusterBlock INDEX_READ_ONLY_BLOCK =
new ClusterBlock(
5,
"index read-only (api)",
false,
false,
RestStatus.FORBIDDEN,
EnumSet.of(ClusterBlockLevel.WRITE, ClusterBlockLevel.METADATA_WRITE));
public static final ClusterBlock INDEX_READ_BLOCK =
new ClusterBlock(
7,
"index read (api)",
false,
false,
RestStatus.FORBIDDEN,
EnumSet.of(ClusterBlockLevel.READ));
public static final ClusterBlock INDEX_WRITE_BLOCK =
new ClusterBlock(
8,
"index write (api)",
false,
false,
RestStatus.FORBIDDEN,
EnumSet.of(ClusterBlockLevel.WRITE));
public static final ClusterBlock INDEX_METADATA_BLOCK =
new ClusterBlock(
9,
"index metadata (api)",
false,
false,
RestStatus.FORBIDDEN,
EnumSet.of(ClusterBlockLevel.METADATA_WRITE, ClusterBlockLevel.METADATA_READ));
public static enum State {
OPEN((byte) 0),
CLOSE((byte) 1);
private final byte id;
State(byte id) {
this.id = id;
}
public byte id() {
return this.id;
}
public static State fromId(byte id) {
if (id == 0) {
return OPEN;
} else if (id == 1) {
return CLOSE;
}
throw new IllegalStateException("No state match for id [" + id + "]");
}
public static State fromString(String state) {
if ("open".equals(state)) {
return OPEN;
} else if ("close".equals(state)) {
return CLOSE;
}
throw new IllegalStateException("No state match for [" + state + "]");
}
}
public static final String INDEX_SETTING_PREFIX = "index.";
public static final String SETTING_NUMBER_OF_SHARDS = "index.number_of_shards";
public static final Setting<Integer> INDEX_NUMBER_OF_SHARDS_SETTING =
Setting.intSetting(SETTING_NUMBER_OF_SHARDS, 5, 1, false, Setting.Scope.INDEX);
public static final String SETTING_NUMBER_OF_REPLICAS = "index.number_of_replicas";
public static final Setting<Integer> INDEX_NUMBER_OF_REPLICAS_SETTING =
Setting.intSetting(SETTING_NUMBER_OF_REPLICAS, 1, 0, true, Setting.Scope.INDEX);
public static final String SETTING_SHADOW_REPLICAS = "index.shadow_replicas";
public static final Setting<Boolean> INDEX_SHADOW_REPLICAS_SETTING =
Setting.boolSetting(SETTING_SHADOW_REPLICAS, false, false, Setting.Scope.INDEX);
public static final String SETTING_SHARED_FILESYSTEM = "index.shared_filesystem";
public static final Setting<Boolean> INDEX_SHARED_FILESYSTEM_SETTING =
Setting.boolSetting(SETTING_SHARED_FILESYSTEM, false, false, Setting.Scope.INDEX);
public static final String SETTING_AUTO_EXPAND_REPLICAS = "index.auto_expand_replicas";
public static final Setting<AutoExpandReplicas> INDEX_AUTO_EXPAND_REPLICAS_SETTING =
AutoExpandReplicas.SETTING;
public static final String SETTING_READ_ONLY = "index.blocks.read_only";
public static final Setting<Boolean> INDEX_READ_ONLY_SETTING =
Setting.boolSetting(SETTING_READ_ONLY, false, true, Setting.Scope.INDEX);
public static final String SETTING_BLOCKS_READ = "index.blocks.read";
public static final Setting<Boolean> INDEX_BLOCKS_READ_SETTING =
Setting.boolSetting(SETTING_BLOCKS_READ, false, true, Setting.Scope.INDEX);
public static final String SETTING_BLOCKS_WRITE = "index.blocks.write";
public static final Setting<Boolean> INDEX_BLOCKS_WRITE_SETTING =
Setting.boolSetting(SETTING_BLOCKS_WRITE, false, true, Setting.Scope.INDEX);
public static final String SETTING_BLOCKS_METADATA = "index.blocks.metadata";
public static final Setting<Boolean> INDEX_BLOCKS_METADATA_SETTING =
Setting.boolSetting(SETTING_BLOCKS_METADATA, false, true, Setting.Scope.INDEX);
public static final String SETTING_VERSION_CREATED = "index.version.created";
public static final String SETTING_VERSION_CREATED_STRING = "index.version.created_string";
public static final String SETTING_VERSION_UPGRADED = "index.version.upgraded";
public static final String SETTING_VERSION_UPGRADED_STRING = "index.version.upgraded_string";
public static final String SETTING_VERSION_MINIMUM_COMPATIBLE =
"index.version.minimum_compatible";
public static final String SETTING_CREATION_DATE = "index.creation_date";
public static final String SETTING_PRIORITY = "index.priority";
public static final Setting<Integer> INDEX_PRIORITY_SETTING =
Setting.intSetting("index.priority", 1, 0, true, Setting.Scope.INDEX);
public static final String SETTING_CREATION_DATE_STRING = "index.creation_date_string";
public static final String SETTING_INDEX_UUID = "index.uuid";
public static final String SETTING_DATA_PATH = "index.data_path";
public static final Setting<String> INDEX_DATA_PATH_SETTING =
new Setting<>(SETTING_DATA_PATH, "", Function.identity(), false, Setting.Scope.INDEX);
public static final String SETTING_SHARED_FS_ALLOW_RECOVERY_ON_ANY_NODE =
"index.shared_filesystem.recover_on_any_node";
public static final Setting<Boolean> INDEX_SHARED_FS_ALLOW_RECOVERY_ON_ANY_NODE_SETTING =
Setting.boolSetting(
SETTING_SHARED_FS_ALLOW_RECOVERY_ON_ANY_NODE, false, true, Setting.Scope.INDEX);
public static final String INDEX_UUID_NA_VALUE = "_na_";
public static final Setting<Settings> INDEX_ROUTING_REQUIRE_GROUP_SETTING =
Setting.groupSetting("index.routing.allocation.require.", true, Setting.Scope.INDEX);
public static final Setting<Settings> INDEX_ROUTING_INCLUDE_GROUP_SETTING =
Setting.groupSetting("index.routing.allocation.include.", true, Setting.Scope.INDEX);
public static final Setting<Settings> INDEX_ROUTING_EXCLUDE_GROUP_SETTING =
Setting.groupSetting("index.routing.allocation.exclude.", true, Setting.Scope.INDEX);
public static final IndexMetaData PROTO =
IndexMetaData.builder("")
.settings(Settings.builder().put(IndexMetaData.SETTING_VERSION_CREATED, Version.CURRENT))
.numberOfShards(1)
.numberOfReplicas(0)
.build();
public static final String KEY_ACTIVE_ALLOCATIONS = "active_allocations";
private final int numberOfShards;
private final int numberOfReplicas;
private final Index index;
private final long version;
private final State state;
private final ImmutableOpenMap<String, AliasMetaData> aliases;
private final Settings settings;
private final ImmutableOpenMap<String, MappingMetaData> mappings;
private final ImmutableOpenMap<String, Custom> customs;
private final ImmutableOpenIntMap<Set<String>> activeAllocationIds;
private final transient int totalNumberOfShards;
private final DiscoveryNodeFilters requireFilters;
private final DiscoveryNodeFilters includeFilters;
private final DiscoveryNodeFilters excludeFilters;
private final Version indexCreatedVersion;
private final Version indexUpgradedVersion;
private final org.apache.lucene.util.Version minimumCompatibleLuceneVersion;
private IndexMetaData(
Index index,
long version,
State state,
int numberOfShards,
int numberOfReplicas,
Settings settings,
ImmutableOpenMap<String, MappingMetaData> mappings,
ImmutableOpenMap<String, AliasMetaData> aliases,
ImmutableOpenMap<String, Custom> customs,
ImmutableOpenIntMap<Set<String>> activeAllocationIds,
DiscoveryNodeFilters requireFilters,
DiscoveryNodeFilters includeFilters,
DiscoveryNodeFilters excludeFilters,
Version indexCreatedVersion,
Version indexUpgradedVersion,
org.apache.lucene.util.Version minimumCompatibleLuceneVersion) {
this.index = index;
this.version = version;
this.state = state;
this.numberOfShards = numberOfShards;
this.numberOfReplicas = numberOfReplicas;
this.totalNumberOfShards = numberOfShards * (numberOfReplicas + 1);
this.settings = settings;
this.mappings = mappings;
this.customs = customs;
this.aliases = aliases;
this.activeAllocationIds = activeAllocationIds;
this.requireFilters = requireFilters;
this.includeFilters = includeFilters;
this.excludeFilters = excludeFilters;
this.indexCreatedVersion = indexCreatedVersion;
this.indexUpgradedVersion = indexUpgradedVersion;
this.minimumCompatibleLuceneVersion = minimumCompatibleLuceneVersion;
}
public Index getIndex() {
return index;
}
public String getIndexUUID() {
return index.getUUID();
}
/**
* Test whether the current index UUID is the same as the given one. Returns true if either are
* _na_
*/
public boolean isSameUUID(String otherUUID) {
assert otherUUID != null;
assert getIndexUUID() != null;
if (INDEX_UUID_NA_VALUE.equals(otherUUID) || INDEX_UUID_NA_VALUE.equals(getIndexUUID())) {
return true;
}
return otherUUID.equals(getIndexUUID());
}
public long getVersion() {
return this.version;
}
/**
* Return the {@link Version} on which this index has been created. This information is typically
* useful for backward compatibility.
*/
public Version getCreationVersion() {
return indexCreatedVersion;
}
/**
* Return the {@link Version} on which this index has been upgraded. This information is typically
* useful for backward compatibility.
*/
public Version getUpgradedVersion() {
return indexUpgradedVersion;
}
/** Return the {@link org.apache.lucene.util.Version} of the oldest lucene segment in the index */
public org.apache.lucene.util.Version getMinimumCompatibleVersion() {
return minimumCompatibleLuceneVersion;
}
public long getCreationDate() {
return settings.getAsLong(SETTING_CREATION_DATE, -1l);
}
public State getState() {
return this.state;
}
public int getNumberOfShards() {
return numberOfShards;
}
public int getNumberOfReplicas() {
return numberOfReplicas;
}
public int getTotalNumberOfShards() {
return totalNumberOfShards;
}
public Settings getSettings() {
return settings;
}
public ImmutableOpenMap<String, AliasMetaData> getAliases() {
return this.aliases;
}
public ImmutableOpenMap<String, MappingMetaData> getMappings() {
return mappings;
}
@Nullable
public MappingMetaData mapping(String mappingType) {
return mappings.get(mappingType);
}
/**
* Sometimes, the default mapping exists and an actual mapping is not created yet (introduced), in
* this case, we want to return the default mapping in case it has some default mapping
* definitions.
*
* <p>Note, once the mapping type is introduced, the default mapping is applied on the actual
* typed MappingMetaData, setting its routing, timestamp, and so on if needed.
*/
@Nullable
public MappingMetaData mappingOrDefault(String mappingType) {
MappingMetaData mapping = mappings.get(mappingType);
if (mapping != null) {
return mapping;
}
return mappings.get(MapperService.DEFAULT_MAPPING);
}
public ImmutableOpenMap<String, Custom> getCustoms() {
return this.customs;
}
@SuppressWarnings("unchecked")
public <T extends Custom> T custom(String type) {
return (T) customs.get(type);
}
public ImmutableOpenIntMap<Set<String>> getActiveAllocationIds() {
return activeAllocationIds;
}
public Set<String> activeAllocationIds(int shardId) {
assert shardId >= 0 && shardId < numberOfShards;
return activeAllocationIds.get(shardId);
}
@Nullable
public DiscoveryNodeFilters requireFilters() {
return requireFilters;
}
@Nullable
public DiscoveryNodeFilters includeFilters() {
return includeFilters;
}
@Nullable
public DiscoveryNodeFilters excludeFilters() {
return excludeFilters;
}
@Override
public boolean equals(Object o) {
if (this == o) {
return true;
}
if (o == null || getClass() != o.getClass()) {
return false;
}
IndexMetaData that = (IndexMetaData) o;
if (!aliases.equals(that.aliases)) {
return false;
}
if (!index.equals(that.index)) {
return false;
}
if (!mappings.equals(that.mappings)) {
return false;
}
if (!settings.equals(that.settings)) {
return false;
}
if (state != that.state) {
return false;
}
if (!customs.equals(that.customs)) {
return false;
}
if (!activeAllocationIds.equals(that.activeAllocationIds)) {
return false;
}
return true;
}
@Override
public int hashCode() {
int result = index.hashCode();
result = 31 * result + state.hashCode();
result = 31 * result + aliases.hashCode();
result = 31 * result + settings.hashCode();
result = 31 * result + mappings.hashCode();
result = 31 * result + activeAllocationIds.hashCode();
return result;
}
@Override
public Diff<IndexMetaData> diff(IndexMetaData previousState) {
return new IndexMetaDataDiff(previousState, this);
}
@Override
public Diff<IndexMetaData> readDiffFrom(StreamInput in) throws IOException {
return new IndexMetaDataDiff(in);
}
@Override
public IndexMetaData fromXContent(XContentParser parser, ParseFieldMatcher parseFieldMatcher)
throws IOException {
return Builder.fromXContent(parser);
}
@Override
public XContentBuilder toXContent(XContentBuilder builder, Params params) throws IOException {
Builder.toXContent(this, builder, params);
return builder;
}
private static class IndexMetaDataDiff implements Diff<IndexMetaData> {
private final String index;
private final long version;
private final State state;
private final Settings settings;
private final Diff<ImmutableOpenMap<String, MappingMetaData>> mappings;
private final Diff<ImmutableOpenMap<String, AliasMetaData>> aliases;
private final Diff<ImmutableOpenMap<String, Custom>> customs;
private final Diff<ImmutableOpenIntMap<Set<String>>> activeAllocationIds;
public IndexMetaDataDiff(IndexMetaData before, IndexMetaData after) {
index = after.index.getName();
version = after.version;
state = after.state;
settings = after.settings;
mappings =
DiffableUtils.diff(
before.mappings, after.mappings, DiffableUtils.getStringKeySerializer());
aliases =
DiffableUtils.diff(before.aliases, after.aliases, DiffableUtils.getStringKeySerializer());
customs =
DiffableUtils.diff(before.customs, after.customs, DiffableUtils.getStringKeySerializer());
activeAllocationIds =
DiffableUtils.diff(
before.activeAllocationIds,
after.activeAllocationIds,
DiffableUtils.getVIntKeySerializer(),
DiffableUtils.StringSetValueSerializer.getInstance());
}
public IndexMetaDataDiff(StreamInput in) throws IOException {
index = in.readString();
version = in.readLong();
state = State.fromId(in.readByte());
settings = Settings.readSettingsFromStream(in);
mappings =
DiffableUtils.readImmutableOpenMapDiff(
in, DiffableUtils.getStringKeySerializer(), MappingMetaData.PROTO);
aliases =
DiffableUtils.readImmutableOpenMapDiff(
in, DiffableUtils.getStringKeySerializer(), AliasMetaData.PROTO);
customs =
DiffableUtils.readImmutableOpenMapDiff(
in,
DiffableUtils.getStringKeySerializer(),
new DiffableUtils.DiffableValueSerializer<String, Custom>() {
@Override
public Custom read(StreamInput in, String key) throws IOException {
return lookupPrototypeSafe(key).readFrom(in);
}
@Override
public Diff<Custom> readDiff(StreamInput in, String key) throws IOException {
return lookupPrototypeSafe(key).readDiffFrom(in);
}
});
activeAllocationIds =
DiffableUtils.readImmutableOpenIntMapDiff(
in,
DiffableUtils.getVIntKeySerializer(),
DiffableUtils.StringSetValueSerializer.getInstance());
}
@Override
public void writeTo(StreamOutput out) throws IOException {
out.writeString(index);
out.writeLong(version);
out.writeByte(state.id);
Settings.writeSettingsToStream(settings, out);
mappings.writeTo(out);
aliases.writeTo(out);
customs.writeTo(out);
activeAllocationIds.writeTo(out);
}
@Override
public IndexMetaData apply(IndexMetaData part) {
Builder builder = builder(index);
builder.version(version);
builder.state(state);
builder.settings(settings);
builder.mappings.putAll(mappings.apply(part.mappings));
builder.aliases.putAll(aliases.apply(part.aliases));
builder.customs.putAll(customs.apply(part.customs));
builder.activeAllocationIds.putAll(activeAllocationIds.apply(part.activeAllocationIds));
return builder.build();
}
}
@Override
public IndexMetaData readFrom(StreamInput in) throws IOException {
Builder builder = new Builder(in.readString());
builder.version(in.readLong());
builder.state(State.fromId(in.readByte()));
builder.settings(readSettingsFromStream(in));
int mappingsSize = in.readVInt();
for (int i = 0; i < mappingsSize; i++) {
MappingMetaData mappingMd = MappingMetaData.PROTO.readFrom(in);
builder.putMapping(mappingMd);
}
int aliasesSize = in.readVInt();
for (int i = 0; i < aliasesSize; i++) {
AliasMetaData aliasMd = AliasMetaData.Builder.readFrom(in);
builder.putAlias(aliasMd);
}
int customSize = in.readVInt();
for (int i = 0; i < customSize; i++) {
String type = in.readString();
Custom customIndexMetaData = lookupPrototypeSafe(type).readFrom(in);
builder.putCustom(type, customIndexMetaData);
}
int activeAllocationIdsSize = in.readVInt();
for (int i = 0; i < activeAllocationIdsSize; i++) {
int key = in.readVInt();
Set<String> allocationIds =
DiffableUtils.StringSetValueSerializer.getInstance().read(in, key);
builder.putActiveAllocationIds(key, allocationIds);
}
return builder.build();
}
@Override
public void writeTo(StreamOutput out) throws IOException {
out.writeString(index.getName()); // uuid will come as part of settings
out.writeLong(version);
out.writeByte(state.id());
writeSettingsToStream(settings, out);
out.writeVInt(mappings.size());
for (ObjectCursor<MappingMetaData> cursor : mappings.values()) {
cursor.value.writeTo(out);
}
out.writeVInt(aliases.size());
for (ObjectCursor<AliasMetaData> cursor : aliases.values()) {
cursor.value.writeTo(out);
}
out.writeVInt(customs.size());
for (ObjectObjectCursor<String, Custom> cursor : customs) {
out.writeString(cursor.key);
cursor.value.writeTo(out);
}
out.writeVInt(activeAllocationIds.size());
for (IntObjectCursor<Set<String>> cursor : activeAllocationIds) {
out.writeVInt(cursor.key);
DiffableUtils.StringSetValueSerializer.getInstance().write(cursor.value, out);
}
}
public static Builder builder(String index) {
return new Builder(index);
}
public static Builder builder(IndexMetaData indexMetaData) {
return new Builder(indexMetaData);
}
public static class Builder {
private String index;
private State state = State.OPEN;
private long version = 1;
private Settings settings = Settings.Builder.EMPTY_SETTINGS;
private final ImmutableOpenMap.Builder<String, MappingMetaData> mappings;
private final ImmutableOpenMap.Builder<String, AliasMetaData> aliases;
private final ImmutableOpenMap.Builder<String, Custom> customs;
private final ImmutableOpenIntMap.Builder<Set<String>> activeAllocationIds;
public Builder(String index) {
this.index = index;
this.mappings = ImmutableOpenMap.builder();
this.aliases = ImmutableOpenMap.builder();
this.customs = ImmutableOpenMap.builder();
this.activeAllocationIds = ImmutableOpenIntMap.builder();
}
public Builder(IndexMetaData indexMetaData) {
this.index = indexMetaData.getIndex().getName();
this.state = indexMetaData.state;
this.version = indexMetaData.version;
this.settings = indexMetaData.getSettings();
this.mappings = ImmutableOpenMap.builder(indexMetaData.mappings);
this.aliases = ImmutableOpenMap.builder(indexMetaData.aliases);
this.customs = ImmutableOpenMap.builder(indexMetaData.customs);
this.activeAllocationIds = ImmutableOpenIntMap.builder(indexMetaData.activeAllocationIds);
}
public String index() {
return index;
}
public Builder index(String index) {
this.index = index;
return this;
}
public Builder numberOfShards(int numberOfShards) {
settings =
settingsBuilder().put(settings).put(SETTING_NUMBER_OF_SHARDS, numberOfShards).build();
return this;
}
public int numberOfShards() {
return settings.getAsInt(SETTING_NUMBER_OF_SHARDS, -1);
}
public Builder numberOfReplicas(int numberOfReplicas) {
settings =
settingsBuilder().put(settings).put(SETTING_NUMBER_OF_REPLICAS, numberOfReplicas).build();
return this;
}
public int numberOfReplicas() {
return settings.getAsInt(SETTING_NUMBER_OF_REPLICAS, -1);
}
public Builder creationDate(long creationDate) {
settings = settingsBuilder().put(settings).put(SETTING_CREATION_DATE, creationDate).build();
return this;
}
public Builder settings(Settings.Builder settings) {
this.settings = settings.build();
return this;
}
public Builder settings(Settings settings) {
this.settings = settings;
return this;
}
public MappingMetaData mapping(String type) {
return mappings.get(type);
}
public Builder putMapping(String type, String source) throws IOException {
try (XContentParser parser = XContentFactory.xContent(source).createParser(source)) {
putMapping(new MappingMetaData(type, parser.mapOrdered()));
}
return this;
}
public Builder putMapping(MappingMetaData mappingMd) {
mappings.put(mappingMd.type(), mappingMd);
return this;
}
public Builder state(State state) {
this.state = state;
return this;
}
public Builder putAlias(AliasMetaData aliasMetaData) {
aliases.put(aliasMetaData.alias(), aliasMetaData);
return this;
}
public Builder putAlias(AliasMetaData.Builder aliasMetaData) {
aliases.put(aliasMetaData.alias(), aliasMetaData.build());
return this;
}
public Builder removeAlias(String alias) {
aliases.remove(alias);
return this;
}
public Builder removeAllAliases() {
aliases.clear();
return this;
}
public Builder putCustom(String type, Custom customIndexMetaData) {
this.customs.put(type, customIndexMetaData);
return this;
}
public Builder putActiveAllocationIds(int shardId, Set<String> allocationIds) {
activeAllocationIds.put(shardId, new HashSet(allocationIds));
return this;
}
public long version() {
return this.version;
}
public Builder version(long version) {
this.version = version;
return this;
}
public IndexMetaData build() {
ImmutableOpenMap.Builder<String, AliasMetaData> tmpAliases = aliases;
Settings tmpSettings = settings;
// update default mapping on the MappingMetaData
if (mappings.containsKey(MapperService.DEFAULT_MAPPING)) {
MappingMetaData defaultMapping = mappings.get(MapperService.DEFAULT_MAPPING);
for (ObjectCursor<MappingMetaData> cursor : mappings.values()) {
cursor.value.updateDefaultMapping(defaultMapping);
}
}
Integer maybeNumberOfShards = settings.getAsInt(SETTING_NUMBER_OF_SHARDS, null);
if (maybeNumberOfShards == null) {
throw new IllegalArgumentException("must specify numberOfShards for index [" + index + "]");
}
int numberOfShards = maybeNumberOfShards;
if (numberOfShards <= 0) {
throw new IllegalArgumentException(
"must specify positive number of shards for index [" + index + "]");
}
Integer maybeNumberOfReplicas = settings.getAsInt(SETTING_NUMBER_OF_REPLICAS, null);
if (maybeNumberOfReplicas == null) {
throw new IllegalArgumentException(
"must specify numberOfReplicas for index [" + index + "]");
}
int numberOfReplicas = maybeNumberOfReplicas;
if (numberOfReplicas < 0) {
throw new IllegalArgumentException(
"must specify non-negative number of shards for index [" + index + "]");
}
// fill missing slots in activeAllocationIds with empty set if needed and make all entries
// immutable
ImmutableOpenIntMap.Builder<Set<String>> filledActiveAllocationIds =
ImmutableOpenIntMap.builder();
for (int i = 0; i < numberOfShards; i++) {
if (activeAllocationIds.containsKey(i)) {
filledActiveAllocationIds.put(
i, Collections.unmodifiableSet(new HashSet<>(activeAllocationIds.get(i))));
} else {
filledActiveAllocationIds.put(i, Collections.emptySet());
}
}
final Map<String, String> requireMap =
INDEX_ROUTING_REQUIRE_GROUP_SETTING.get(settings).getAsMap();
final DiscoveryNodeFilters requireFilters;
if (requireMap.isEmpty()) {
requireFilters = null;
} else {
requireFilters = DiscoveryNodeFilters.buildFromKeyValue(AND, requireMap);
}
Map<String, String> includeMap = INDEX_ROUTING_INCLUDE_GROUP_SETTING.get(settings).getAsMap();
final DiscoveryNodeFilters includeFilters;
if (includeMap.isEmpty()) {
includeFilters = null;
} else {
includeFilters = DiscoveryNodeFilters.buildFromKeyValue(OR, includeMap);
}
Map<String, String> excludeMap = INDEX_ROUTING_EXCLUDE_GROUP_SETTING.get(settings).getAsMap();
final DiscoveryNodeFilters excludeFilters;
if (excludeMap.isEmpty()) {
excludeFilters = null;
} else {
excludeFilters = DiscoveryNodeFilters.buildFromKeyValue(OR, excludeMap);
}
Version indexCreatedVersion = Version.indexCreated(settings);
Version indexUpgradedVersion =
settings.getAsVersion(IndexMetaData.SETTING_VERSION_UPGRADED, indexCreatedVersion);
String stringLuceneVersion = settings.get(SETTING_VERSION_MINIMUM_COMPATIBLE);
final org.apache.lucene.util.Version minimumCompatibleLuceneVersion;
if (stringLuceneVersion != null) {
try {
minimumCompatibleLuceneVersion =
org.apache.lucene.util.Version.parse(stringLuceneVersion);
} catch (ParseException ex) {
throw new IllegalStateException(
"Cannot parse lucene version ["
+ stringLuceneVersion
+ "] in the ["
+ SETTING_VERSION_MINIMUM_COMPATIBLE
+ "] setting",
ex);
}
} else {
minimumCompatibleLuceneVersion = null;
}
final String uuid = settings.get(SETTING_INDEX_UUID, INDEX_UUID_NA_VALUE);
return new IndexMetaData(
new Index(index, uuid),
version,
state,
numberOfShards,
numberOfReplicas,
tmpSettings,
mappings.build(),
tmpAliases.build(),
customs.build(),
filledActiveAllocationIds.build(),
requireFilters,
includeFilters,
excludeFilters,
indexCreatedVersion,
indexUpgradedVersion,
minimumCompatibleLuceneVersion);
}
public static void toXContent(
IndexMetaData indexMetaData, XContentBuilder builder, ToXContent.Params params)
throws IOException {
builder.startObject(
indexMetaData.getIndex().getName(), XContentBuilder.FieldCaseConversion.NONE);
builder.field("version", indexMetaData.getVersion());
builder.field("state", indexMetaData.getState().toString().toLowerCase(Locale.ENGLISH));
boolean binary = params.paramAsBoolean("binary", false);
builder.startObject("settings");
for (Map.Entry<String, String> entry : indexMetaData.getSettings().getAsMap().entrySet()) {
builder.field(entry.getKey(), entry.getValue());
}
builder.endObject();
builder.startArray("mappings");
for (ObjectObjectCursor<String, MappingMetaData> cursor : indexMetaData.getMappings()) {
if (binary) {
builder.value(cursor.value.source().compressed());
} else {
byte[] data = cursor.value.source().uncompressed();
XContentParser parser = XContentFactory.xContent(data).createParser(data);
Map<String, Object> mapping = parser.mapOrdered();
parser.close();
builder.map(mapping);
}
}
builder.endArray();
for (ObjectObjectCursor<String, Custom> cursor : indexMetaData.getCustoms()) {
builder.startObject(cursor.key, XContentBuilder.FieldCaseConversion.NONE);
cursor.value.toXContent(builder, params);
builder.endObject();
}
builder.startObject("aliases");
for (ObjectCursor<AliasMetaData> cursor : indexMetaData.getAliases().values()) {
AliasMetaData.Builder.toXContent(cursor.value, builder, params);
}
builder.endObject();
builder.startObject(KEY_ACTIVE_ALLOCATIONS);
for (IntObjectCursor<Set<String>> cursor : indexMetaData.activeAllocationIds) {
builder.startArray(String.valueOf(cursor.key));
for (String allocationId : cursor.value) {
builder.value(allocationId);
}
builder.endArray();
}
builder.endObject();
builder.endObject();
}
public static IndexMetaData fromXContent(XContentParser parser) throws IOException {
if (parser.currentToken() == null) { // fresh parser? move to the first token
parser.nextToken();
}
if (parser.currentToken()
== XContentParser.Token.START_OBJECT) { // on a start object move to next token
parser.nextToken();
}
if (parser.currentToken() != XContentParser.Token.FIELD_NAME) {
throw new IllegalArgumentException(
"expected field name but got a " + parser.currentToken());
}
Builder builder = new Builder(parser.currentName());
String currentFieldName = null;
XContentParser.Token token = parser.nextToken();
if (token != XContentParser.Token.START_OBJECT) {
throw new IllegalArgumentException("expected object but got a " + token);
}
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token == XContentParser.Token.FIELD_NAME) {
currentFieldName = parser.currentName();
} else if (token == XContentParser.Token.START_OBJECT) {
if ("settings".equals(currentFieldName)) {
builder.settings(
Settings.settingsBuilder()
.put(SettingsLoader.Helper.loadNestedFromMap(parser.mapOrdered())));
} else if ("mappings".equals(currentFieldName)) {
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token == XContentParser.Token.FIELD_NAME) {
currentFieldName = parser.currentName();
} else if (token == XContentParser.Token.START_OBJECT) {
String mappingType = currentFieldName;
Map<String, Object> mappingSource =
MapBuilder.<String, Object>newMapBuilder()
.put(mappingType, parser.mapOrdered())
.map();
builder.putMapping(new MappingMetaData(mappingType, mappingSource));
} else {
throw new IllegalArgumentException("Unexpected token: " + token);
}
}
} else if ("aliases".equals(currentFieldName)) {
while (parser.nextToken() != XContentParser.Token.END_OBJECT) {
builder.putAlias(AliasMetaData.Builder.fromXContent(parser));
}
} else if (KEY_ACTIVE_ALLOCATIONS.equals(currentFieldName)) {
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token == XContentParser.Token.FIELD_NAME) {
currentFieldName = parser.currentName();
} else if (token == XContentParser.Token.START_ARRAY) {
String shardId = currentFieldName;
Set<String> allocationIds = new HashSet<>();
while ((token = parser.nextToken()) != XContentParser.Token.END_ARRAY) {
if (token == XContentParser.Token.VALUE_STRING) {
allocationIds.add(parser.text());
}
}
builder.putActiveAllocationIds(Integer.valueOf(shardId), allocationIds);
} else {
throw new IllegalArgumentException("Unexpected token: " + token);
}
}
} else if ("warmers".equals(currentFieldName)) {
// TODO: do this in 4.0:
// throw new IllegalArgumentException("Warmers are not supported anymore - are you
// upgrading from 1.x?");
// ignore: warmers have been removed in 3.0 and are
// simply ignored when upgrading from 2.x
assert Version.CURRENT.major <= 3;
parser.skipChildren();
} else {
// check if its a custom index metadata
Custom proto = lookupPrototype(currentFieldName);
if (proto == null) {
// TODO warn
parser.skipChildren();
} else {
Custom custom = proto.fromXContent(parser);
builder.putCustom(custom.type(), custom);
}
}
} else if (token == XContentParser.Token.START_ARRAY) {
if ("mappings".equals(currentFieldName)) {
while ((token = parser.nextToken()) != XContentParser.Token.END_ARRAY) {
if (token == XContentParser.Token.VALUE_EMBEDDED_OBJECT) {
builder.putMapping(
new MappingMetaData(new CompressedXContent(parser.binaryValue())));
} else {
Map<String, Object> mapping = parser.mapOrdered();
if (mapping.size() == 1) {
String mappingType = mapping.keySet().iterator().next();
builder.putMapping(new MappingMetaData(mappingType, mapping));
}
}
}
} else {
throw new IllegalArgumentException("Unexpected field for an array " + currentFieldName);
}
} else if (token.isValue()) {
if ("state".equals(currentFieldName)) {
builder.state(State.fromString(parser.text()));
} else if ("version".equals(currentFieldName)) {
builder.version(parser.longValue());
} else {
throw new IllegalArgumentException("Unexpected field [" + currentFieldName + "]");
}
} else {
throw new IllegalArgumentException("Unexpected token " + token);
}
}
return builder.build();
}
public static IndexMetaData readFrom(StreamInput in) throws IOException {
return PROTO.readFrom(in);
}
}
/**
* Returns <code>true</code> iff the given settings indicate that the index associated with these
* settings allocates it's shards on a shared filesystem. Otherwise <code>false</code>. The
* default setting for this is the returned value from {@link
* #isIndexUsingShadowReplicas(org.elasticsearch.common.settings.Settings)}.
*/
public static boolean isOnSharedFilesystem(Settings settings) {
return settings.getAsBoolean(SETTING_SHARED_FILESYSTEM, isIndexUsingShadowReplicas(settings));
}
/**
* Returns <code>true</code> iff the given settings indicate that the index associated with these
* settings uses shadow replicas. Otherwise <code>false</code>. The default setting for this is
* <code>false</code>.
*/
public static boolean isIndexUsingShadowReplicas(Settings settings) {
return settings.getAsBoolean(SETTING_SHADOW_REPLICAS, false);
}
/**
* Adds human readable version and creation date settings. This method is used to display the
* settings in a human readable format in REST API
*/
public static Settings addHumanReadableSettings(Settings settings) {
Settings.Builder builder = Settings.builder().put(settings);
Version version = settings.getAsVersion(SETTING_VERSION_CREATED, null);
if (version != null) {
builder.put(SETTING_VERSION_CREATED_STRING, version.toString());
}
Version versionUpgraded = settings.getAsVersion(SETTING_VERSION_UPGRADED, null);
if (versionUpgraded != null) {
builder.put(SETTING_VERSION_UPGRADED_STRING, versionUpgraded.toString());
}
Long creationDate = settings.getAsLong(SETTING_CREATION_DATE, null);
if (creationDate != null) {
DateTime creationDateTime = new DateTime(creationDate, DateTimeZone.UTC);
builder.put(SETTING_CREATION_DATE_STRING, creationDateTime.toString());
}
return builder.build();
}
}
/** A component that holds all data paths for a single node. */
public final class NodeEnvironment implements Closeable {
private final Logger logger;
public static class NodePath {
/* ${data.paths}/nodes/{node.id} */
public final Path path;
/* ${data.paths}/nodes/{node.id}/indices */
public final Path indicesPath;
/** Cached FileStore from path */
public final FileStore fileStore;
/**
* Cached result of Lucene's {@code IOUtils.spins} on path. This is a trilean value: null means
* we could not determine it (we are not running on Linux, or we hit an exception trying), True
* means the device possibly spins and False means it does not.
*/
public final Boolean spins;
public final int majorDeviceNumber;
public final int minorDeviceNumber;
public NodePath(Path path) throws IOException {
this.path = path;
this.indicesPath = path.resolve(INDICES_FOLDER);
this.fileStore = Environment.getFileStore(path);
if (fileStore.supportsFileAttributeView("lucene")) {
this.spins = (Boolean) fileStore.getAttribute("lucene:spins");
this.majorDeviceNumber = (int) fileStore.getAttribute("lucene:major_device_number");
this.minorDeviceNumber = (int) fileStore.getAttribute("lucene:minor_device_number");
} else {
this.spins = null;
this.majorDeviceNumber = -1;
this.minorDeviceNumber = -1;
}
}
/**
* Resolves the given shards directory against this NodePath
* ${data.paths}/nodes/{node.id}/indices/{index.uuid}/{shard.id}
*/
public Path resolve(ShardId shardId) {
return resolve(shardId.getIndex()).resolve(Integer.toString(shardId.id()));
}
/**
* Resolves index directory against this NodePath
* ${data.paths}/nodes/{node.id}/indices/{index.uuid}
*/
public Path resolve(Index index) {
return indicesPath.resolve(index.getUUID());
}
@Override
public String toString() {
return "NodePath{" + "path=" + path + ", spins=" + spins + '}';
}
}
private final NodePath[] nodePaths;
private final Path sharedDataPath;
private final Lock[] locks;
private final int nodeLockId;
private final AtomicBoolean closed = new AtomicBoolean(false);
private final Map<ShardId, InternalShardLock> shardLocks = new HashMap<>();
private final NodeMetaData nodeMetaData;
/** Maximum number of data nodes that should run in an environment. */
public static final Setting<Integer> MAX_LOCAL_STORAGE_NODES_SETTING =
Setting.intSetting("node.max_local_storage_nodes", 1, 1, Property.NodeScope);
/** If true automatically append node lock id to custom data paths. */
public static final Setting<Boolean> ADD_NODE_LOCK_ID_TO_CUSTOM_PATH =
Setting.boolSetting("node.add_lock_id_to_custom_path", true, Property.NodeScope);
/**
* Seed for determining a persisted unique uuid of this node. If the node has already a persisted
* uuid on disk, this seed will be ignored and the uuid from disk will be reused.
*/
public static final Setting<Long> NODE_ID_SEED_SETTING =
Setting.longSetting("node.id.seed", 0L, Long.MIN_VALUE, Property.NodeScope);
/** If true the [verbose] SegmentInfos.infoStream logging is sent to System.out. */
public static final Setting<Boolean> ENABLE_LUCENE_SEGMENT_INFOS_TRACE_SETTING =
Setting.boolSetting("node.enable_lucene_segment_infos_trace", false, Property.NodeScope);
public static final String NODES_FOLDER = "nodes";
public static final String INDICES_FOLDER = "indices";
public static final String NODE_LOCK_FILENAME = "node.lock";
public NodeEnvironment(Settings settings, Environment environment) throws IOException {
if (!DiscoveryNode.nodeRequiresLocalStorage(settings)) {
nodePaths = null;
sharedDataPath = null;
locks = null;
nodeLockId = -1;
nodeMetaData = new NodeMetaData(generateNodeId(settings));
logger =
Loggers.getLogger(
getClass(), Node.addNodeNameIfNeeded(settings, this.nodeMetaData.nodeId()));
return;
}
final NodePath[] nodePaths = new NodePath[environment.dataWithClusterFiles().length];
final Lock[] locks = new Lock[nodePaths.length];
boolean success = false;
// trace logger to debug issues before the default node name is derived from the node id
Logger startupTraceLogger = Loggers.getLogger(getClass(), settings);
try {
sharedDataPath = environment.sharedDataFile();
int nodeLockId = -1;
IOException lastException = null;
int maxLocalStorageNodes = MAX_LOCAL_STORAGE_NODES_SETTING.get(settings);
for (int possibleLockId = 0; possibleLockId < maxLocalStorageNodes; possibleLockId++) {
for (int dirIndex = 0; dirIndex < environment.dataFiles().length; dirIndex++) {
Path dataDirWithClusterName = environment.dataWithClusterFiles()[dirIndex];
Path dataDir = environment.dataFiles()[dirIndex];
Path dir = dataDir.resolve(NODES_FOLDER).resolve(Integer.toString(possibleLockId));
Files.createDirectories(dir);
try (Directory luceneDir = FSDirectory.open(dir, NativeFSLockFactory.INSTANCE)) {
startupTraceLogger.trace("obtaining node lock on {} ...", dir.toAbsolutePath());
try {
locks[dirIndex] = luceneDir.obtainLock(NODE_LOCK_FILENAME);
nodePaths[dirIndex] = new NodePath(dir);
nodeLockId = possibleLockId;
} catch (LockObtainFailedException ex) {
startupTraceLogger.trace("failed to obtain node lock on {}", dir.toAbsolutePath());
// release all the ones that were obtained up until now
releaseAndNullLocks(locks);
break;
}
} catch (IOException e) {
startupTraceLogger.trace(
(Supplier<?>)
() ->
new ParameterizedMessage(
"failed to obtain node lock on {}", dir.toAbsolutePath()),
e);
lastException = new IOException("failed to obtain lock on " + dir.toAbsolutePath(), e);
// release all the ones that were obtained up until now
releaseAndNullLocks(locks);
break;
}
}
if (locks[0] != null) {
// we found a lock, break
break;
}
}
if (locks[0] == null) {
final String message =
String.format(
Locale.ROOT,
"failed to obtain node locks, tried [%s] with lock id%s;"
+ " maybe these locations are not writable or multiple nodes were started without increasing [%s] (was [%d])?",
Arrays.toString(environment.dataWithClusterFiles()),
maxLocalStorageNodes == 1 ? " [0]" : "s [0--" + (maxLocalStorageNodes - 1) + "]",
MAX_LOCAL_STORAGE_NODES_SETTING.getKey(),
maxLocalStorageNodes);
throw new IllegalStateException(message, lastException);
}
this.nodeMetaData = loadOrCreateNodeMetaData(settings, startupTraceLogger, nodePaths);
this.logger =
Loggers.getLogger(
getClass(), Node.addNodeNameIfNeeded(settings, this.nodeMetaData.nodeId()));
this.nodeLockId = nodeLockId;
this.locks = locks;
this.nodePaths = nodePaths;
if (logger.isDebugEnabled()) {
logger.debug("using node location [{}], local_lock_id [{}]", nodePaths, nodeLockId);
}
maybeLogPathDetails();
maybeLogHeapDetails();
applySegmentInfosTrace(settings);
assertCanWrite();
success = true;
} finally {
if (success == false) {
IOUtils.closeWhileHandlingException(locks);
}
}
}
/** Returns true if the directory is empty */
private static boolean dirEmpty(final Path path) throws IOException {
try (DirectoryStream<Path> stream = Files.newDirectoryStream(path)) {
return stream.iterator().hasNext() == false;
}
}
private static void releaseAndNullLocks(Lock[] locks) {
for (int i = 0; i < locks.length; i++) {
if (locks[i] != null) {
IOUtils.closeWhileHandlingException(locks[i]);
}
locks[i] = null;
}
}
private void maybeLogPathDetails() throws IOException {
// We do some I/O in here, so skip this if DEBUG/INFO are not enabled:
if (logger.isDebugEnabled()) {
// Log one line per path.data:
StringBuilder sb = new StringBuilder();
for (NodePath nodePath : nodePaths) {
sb.append('\n').append(" -> ").append(nodePath.path.toAbsolutePath());
String spinsDesc;
if (nodePath.spins == null) {
spinsDesc = "unknown";
} else if (nodePath.spins) {
spinsDesc = "possibly";
} else {
spinsDesc = "no";
}
FsInfo.Path fsPath = FsProbe.getFSInfo(nodePath);
sb.append(", free_space [")
.append(fsPath.getFree())
.append("], usable_space [")
.append(fsPath.getAvailable())
.append("], total_space [")
.append(fsPath.getTotal())
.append("], spins? [")
.append(spinsDesc)
.append("], mount [")
.append(fsPath.getMount())
.append("], type [")
.append(fsPath.getType())
.append(']');
}
logger.debug("node data locations details:{}", sb);
} else if (logger.isInfoEnabled()) {
FsInfo.Path totFSPath = new FsInfo.Path();
Set<String> allTypes = new HashSet<>();
Set<String> allSpins = new HashSet<>();
Set<String> allMounts = new HashSet<>();
for (NodePath nodePath : nodePaths) {
FsInfo.Path fsPath = FsProbe.getFSInfo(nodePath);
String mount = fsPath.getMount();
if (allMounts.contains(mount) == false) {
allMounts.add(mount);
String type = fsPath.getType();
if (type != null) {
allTypes.add(type);
}
Boolean spins = fsPath.getSpins();
if (spins == null) {
allSpins.add("unknown");
} else if (spins.booleanValue()) {
allSpins.add("possibly");
} else {
allSpins.add("no");
}
totFSPath.add(fsPath);
}
}
// Just log a 1-line summary:
logger.info(
"using [{}] data paths, mounts [{}], net usable_space [{}], net total_space [{}], spins? [{}], types [{}]",
nodePaths.length,
allMounts,
totFSPath.getAvailable(),
totFSPath.getTotal(),
toString(allSpins),
toString(allTypes));
}
}
private void maybeLogHeapDetails() {
JvmInfo jvmInfo = JvmInfo.jvmInfo();
ByteSizeValue maxHeapSize = jvmInfo.getMem().getHeapMax();
String useCompressedOops = jvmInfo.useCompressedOops();
logger.info(
"heap size [{}], compressed ordinary object pointers [{}]", maxHeapSize, useCompressedOops);
}
/**
* scans the node paths and loads existing metaData file. If not found a new meta data will be
* generated and persisted into the nodePaths
*/
private static NodeMetaData loadOrCreateNodeMetaData(
Settings settings, Logger logger, NodePath... nodePaths) throws IOException {
final Path[] paths = Arrays.stream(nodePaths).map(np -> np.path).toArray(Path[]::new);
NodeMetaData metaData = NodeMetaData.FORMAT.loadLatestState(logger, paths);
if (metaData == null) {
metaData = new NodeMetaData(generateNodeId(settings));
}
// we write again to make sure all paths have the latest state file
NodeMetaData.FORMAT.write(metaData, paths);
return metaData;
}
public static String generateNodeId(Settings settings) {
Random random = Randomness.get(settings, NODE_ID_SEED_SETTING);
return UUIDs.randomBase64UUID(random);
}
@SuppressForbidden(reason = "System.out.*")
static void applySegmentInfosTrace(Settings settings) {
if (ENABLE_LUCENE_SEGMENT_INFOS_TRACE_SETTING.get(settings)) {
SegmentInfos.setInfoStream(System.out);
}
}
private static String toString(Collection<String> items) {
StringBuilder b = new StringBuilder();
for (String item : items) {
if (b.length() > 0) {
b.append(", ");
}
b.append(item);
}
return b.toString();
}
/**
* Deletes a shard data directory iff the shards locks were successfully acquired.
*
* @param shardId the id of the shard to delete to delete
* @throws IOException if an IOException occurs
*/
public void deleteShardDirectorySafe(ShardId shardId, IndexSettings indexSettings)
throws IOException, ShardLockObtainFailedException {
final Path[] paths = availableShardPaths(shardId);
logger.trace("deleting shard {} directory, paths: [{}]", shardId, paths);
try (ShardLock lock = shardLock(shardId)) {
deleteShardDirectoryUnderLock(lock, indexSettings);
}
}
/**
* Acquires, then releases, all {@code write.lock} files in the given shard paths. The
* "write.lock" file is assumed to be under the shard path's "index" directory as used by
* Elasticsearch.
*
* @throws LockObtainFailedException if any of the locks could not be acquired
*/
public static void acquireFSLockForPaths(IndexSettings indexSettings, Path... shardPaths)
throws IOException {
Lock[] locks = new Lock[shardPaths.length];
Directory[] dirs = new Directory[shardPaths.length];
try {
for (int i = 0; i < shardPaths.length; i++) {
// resolve the directory the shard actually lives in
Path p = shardPaths[i].resolve("index");
// open a directory (will be immediately closed) on the shard's location
dirs[i] =
new SimpleFSDirectory(
p, indexSettings.getValue(FsDirectoryService.INDEX_LOCK_FACTOR_SETTING));
// create a lock for the "write.lock" file
try {
locks[i] = dirs[i].obtainLock(IndexWriter.WRITE_LOCK_NAME);
} catch (IOException ex) {
throw new LockObtainFailedException(
"unable to acquire " + IndexWriter.WRITE_LOCK_NAME + " for " + p, ex);
}
}
} finally {
IOUtils.closeWhileHandlingException(locks);
IOUtils.closeWhileHandlingException(dirs);
}
}
/**
* Deletes a shard data directory. Note: this method assumes that the shard lock is acquired. This
* method will also attempt to acquire the write locks for the shard's paths before deleting the
* data, but this is best effort, as the lock is released before the deletion happens in order to
* allow the folder to be deleted
*
* @param lock the shards lock
* @throws IOException if an IOException occurs
* @throws ElasticsearchException if the write.lock is not acquirable
*/
public void deleteShardDirectoryUnderLock(ShardLock lock, IndexSettings indexSettings)
throws IOException {
final ShardId shardId = lock.getShardId();
assert isShardLocked(shardId) : "shard " + shardId + " is not locked";
final Path[] paths = availableShardPaths(shardId);
logger.trace("acquiring locks for {}, paths: [{}]", shardId, paths);
acquireFSLockForPaths(indexSettings, paths);
IOUtils.rm(paths);
if (indexSettings.hasCustomDataPath()) {
Path customLocation = resolveCustomLocation(indexSettings, shardId);
logger.trace("acquiring lock for {}, custom path: [{}]", shardId, customLocation);
acquireFSLockForPaths(indexSettings, customLocation);
logger.trace("deleting custom shard {} directory [{}]", shardId, customLocation);
IOUtils.rm(customLocation);
}
logger.trace("deleted shard {} directory, paths: [{}]", shardId, paths);
assert FileSystemUtils.exists(paths) == false;
}
private boolean isShardLocked(ShardId id) {
try {
shardLock(id, 0).close();
return false;
} catch (ShardLockObtainFailedException ex) {
return true;
}
}
/**
* Deletes an indexes data directory recursively iff all of the indexes shards locks were
* successfully acquired. If any of the indexes shard directories can't be locked non of the
* shards will be deleted
*
* @param index the index to delete
* @param lockTimeoutMS how long to wait for acquiring the indices shard locks
* @param indexSettings settings for the index being deleted
* @throws IOException if any of the shards data directories can't be locked or deleted
*/
public void deleteIndexDirectorySafe(Index index, long lockTimeoutMS, IndexSettings indexSettings)
throws IOException, ShardLockObtainFailedException {
final List<ShardLock> locks = lockAllForIndex(index, indexSettings, lockTimeoutMS);
try {
deleteIndexDirectoryUnderLock(index, indexSettings);
} finally {
IOUtils.closeWhileHandlingException(locks);
}
}
/**
* Deletes an indexes data directory recursively. Note: this method assumes that the shard lock is
* acquired
*
* @param index the index to delete
* @param indexSettings settings for the index being deleted
*/
public void deleteIndexDirectoryUnderLock(Index index, IndexSettings indexSettings)
throws IOException {
final Path[] indexPaths = indexPaths(index);
logger.trace(
"deleting index {} directory, paths({}): [{}]", index, indexPaths.length, indexPaths);
IOUtils.rm(indexPaths);
if (indexSettings.hasCustomDataPath()) {
Path customLocation = resolveIndexCustomLocation(indexSettings);
logger.trace("deleting custom index {} directory [{}]", index, customLocation);
IOUtils.rm(customLocation);
}
}
/**
* Tries to lock all local shards for the given index. If any of the shard locks can't be acquired
* a {@link ShardLockObtainFailedException} is thrown and all previously acquired locks are
* released.
*
* @param index the index to lock shards for
* @param lockTimeoutMS how long to wait for acquiring the indices shard locks
* @return the {@link ShardLock} instances for this index.
* @throws IOException if an IOException occurs.
*/
public List<ShardLock> lockAllForIndex(Index index, IndexSettings settings, long lockTimeoutMS)
throws IOException, ShardLockObtainFailedException {
final int numShards = settings.getNumberOfShards();
if (numShards <= 0) {
throw new IllegalArgumentException("settings must contain a non-null > 0 number of shards");
}
logger.trace("locking all shards for index {} - [{}]", index, numShards);
List<ShardLock> allLocks = new ArrayList<>(numShards);
boolean success = false;
long startTimeNS = System.nanoTime();
try {
for (int i = 0; i < numShards; i++) {
long timeoutLeftMS =
Math.max(0, lockTimeoutMS - TimeValue.nsecToMSec((System.nanoTime() - startTimeNS)));
allLocks.add(shardLock(new ShardId(index, i), timeoutLeftMS));
}
success = true;
} finally {
if (success == false) {
logger.trace("unable to lock all shards for index {}", index);
IOUtils.closeWhileHandlingException(allLocks);
}
}
return allLocks;
}
/**
* Tries to lock the given shards ID. A shard lock is required to perform any kind of write
* operation on a shards data directory like deleting files, creating a new index writer or
* recover from a different shard instance into it. If the shard lock can not be acquired a {@link
* ShardLockObtainFailedException} is thrown.
*
* <p>Note: this method will return immediately if the lock can't be acquired.
*
* @param id the shard ID to lock
* @return the shard lock. Call {@link ShardLock#close()} to release the lock
*/
public ShardLock shardLock(ShardId id) throws ShardLockObtainFailedException {
return shardLock(id, 0);
}
/**
* Tries to lock the given shards ID. A shard lock is required to perform any kind of write
* operation on a shards data directory like deleting files, creating a new index writer or
* recover from a different shard instance into it. If the shard lock can not be acquired a {@link
* ShardLockObtainFailedException} is thrown
*
* @param shardId the shard ID to lock
* @param lockTimeoutMS the lock timeout in milliseconds
* @return the shard lock. Call {@link ShardLock#close()} to release the lock
*/
public ShardLock shardLock(final ShardId shardId, long lockTimeoutMS)
throws ShardLockObtainFailedException {
logger.trace("acquiring node shardlock on [{}], timeout [{}]", shardId, lockTimeoutMS);
final InternalShardLock shardLock;
final boolean acquired;
synchronized (shardLocks) {
if (shardLocks.containsKey(shardId)) {
shardLock = shardLocks.get(shardId);
shardLock.incWaitCount();
acquired = false;
} else {
shardLock = new InternalShardLock(shardId);
shardLocks.put(shardId, shardLock);
acquired = true;
}
}
if (acquired == false) {
boolean success = false;
try {
shardLock.acquire(lockTimeoutMS);
success = true;
} finally {
if (success == false) {
shardLock.decWaitCount();
}
}
}
logger.trace("successfully acquired shardlock for [{}]", shardId);
return new ShardLock(shardId) { // new instance prevents double closing
@Override
protected void closeInternal() {
shardLock.release();
logger.trace("released shard lock for [{}]", shardId);
}
};
}
/** A functional interface that people can use to reference {@link #shardLock(ShardId, long)} */
@FunctionalInterface
public interface ShardLocker {
ShardLock lock(ShardId shardId, long lockTimeoutMS) throws ShardLockObtainFailedException;
}
/**
* Returns all currently lock shards.
*
* <p>Note: the shard ids return do not contain a valid Index UUID
*/
public Set<ShardId> lockedShards() {
synchronized (shardLocks) {
return unmodifiableSet(new HashSet<>(shardLocks.keySet()));
}
}
private final class InternalShardLock {
/*
* This class holds a mutex for exclusive access and timeout / wait semantics
* and a reference count to cleanup the shard lock instance form the internal data
* structure if nobody is waiting for it. the wait count is guarded by the same lock
* that is used to mutate the map holding the shard locks to ensure exclusive access
*/
private final Semaphore mutex = new Semaphore(1);
private int waitCount = 1; // guarded by shardLocks
private final ShardId shardId;
InternalShardLock(ShardId shardId) {
this.shardId = shardId;
mutex.acquireUninterruptibly();
}
protected void release() {
mutex.release();
decWaitCount();
}
void incWaitCount() {
synchronized (shardLocks) {
assert waitCount > 0 : "waitCount is " + waitCount + " but should be > 0";
waitCount++;
}
}
private void decWaitCount() {
synchronized (shardLocks) {
assert waitCount > 0 : "waitCount is " + waitCount + " but should be > 0";
--waitCount;
logger.trace("shard lock wait count for {} is now [{}]", shardId, waitCount);
if (waitCount == 0) {
logger.trace("last shard lock wait decremented, removing lock for {}", shardId);
InternalShardLock remove = shardLocks.remove(shardId);
assert remove != null : "Removed lock was null";
}
}
}
void acquire(long timeoutInMillis) throws ShardLockObtainFailedException {
try {
if (mutex.tryAcquire(timeoutInMillis, TimeUnit.MILLISECONDS) == false) {
throw new ShardLockObtainFailedException(
shardId, "obtaining shard lock timed out after " + timeoutInMillis + "ms");
}
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
throw new ShardLockObtainFailedException(
shardId, "thread interrupted while trying to obtain shard lock", e);
}
}
}
public boolean hasNodeFile() {
return nodePaths != null && locks != null;
}
/**
* Returns an array of all of the nodes data locations.
*
* @throws IllegalStateException if the node is not configured to store local locations
*/
public Path[] nodeDataPaths() {
assertEnvIsLocked();
Path[] paths = new Path[nodePaths.length];
for (int i = 0; i < paths.length; i++) {
paths[i] = nodePaths[i].path;
}
return paths;
}
/**
* returns the unique uuid describing this node. The uuid is persistent in the data folder of this
* node and remains across restarts.
*/
public String nodeId() {
// we currently only return the ID and hide the underlying nodeMetaData implementation in order
// to avoid
// confusion with other "metadata" like node settings found in elasticsearch.yml. In future
// we can encapsulate both (and more) in one NodeMetaData (or NodeSettings) object ala
// IndexSettings
return nodeMetaData.nodeId();
}
/** Returns an array of all of the {@link NodePath}s. */
public NodePath[] nodePaths() {
assertEnvIsLocked();
if (nodePaths == null || locks == null) {
throw new IllegalStateException("node is not configured to store local location");
}
return nodePaths;
}
/** Returns all index paths. */
public Path[] indexPaths(Index index) {
assertEnvIsLocked();
Path[] indexPaths = new Path[nodePaths.length];
for (int i = 0; i < nodePaths.length; i++) {
indexPaths[i] = nodePaths[i].resolve(index);
}
return indexPaths;
}
/**
* Returns all shard paths excluding custom shard path. Note: Shards are only allocated on one of
* the returned paths. The returned array may contain paths to non-existing directories.
*
* @see IndexSettings#hasCustomDataPath()
* @see #resolveCustomLocation(IndexSettings, ShardId)
*/
public Path[] availableShardPaths(ShardId shardId) {
assertEnvIsLocked();
final NodePath[] nodePaths = nodePaths();
final Path[] shardLocations = new Path[nodePaths.length];
for (int i = 0; i < nodePaths.length; i++) {
shardLocations[i] = nodePaths[i].resolve(shardId);
}
return shardLocations;
}
/** Returns all folder names in ${data.paths}/nodes/{node.id}/indices folder */
public Set<String> availableIndexFolders() throws IOException {
if (nodePaths == null || locks == null) {
throw new IllegalStateException("node is not configured to store local location");
}
assertEnvIsLocked();
Set<String> indexFolders = new HashSet<>();
for (NodePath nodePath : nodePaths) {
Path indicesLocation = nodePath.indicesPath;
if (Files.isDirectory(indicesLocation)) {
try (DirectoryStream<Path> stream = Files.newDirectoryStream(indicesLocation)) {
for (Path index : stream) {
if (Files.isDirectory(index)) {
indexFolders.add(index.getFileName().toString());
}
}
}
}
}
return indexFolders;
}
/**
* Resolves all existing paths to <code>indexFolderName</code> in
* ${data.paths}/nodes/{node.id}/indices
*/
public Path[] resolveIndexFolder(String indexFolderName) throws IOException {
if (nodePaths == null || locks == null) {
throw new IllegalStateException("node is not configured to store local location");
}
assertEnvIsLocked();
List<Path> paths = new ArrayList<>(nodePaths.length);
for (NodePath nodePath : nodePaths) {
Path indexFolder = nodePath.indicesPath.resolve(indexFolderName);
if (Files.exists(indexFolder)) {
paths.add(indexFolder);
}
}
return paths.toArray(new Path[paths.size()]);
}
/**
* Tries to find all allocated shards for the given index on the current node. NOTE: This methods
* is prone to race-conditions on the filesystem layer since it might not see directories created
* concurrently or while it's traversing.
*
* @param index the index to filter shards
* @return a set of shard IDs
* @throws IOException if an IOException occurs
*/
public Set<ShardId> findAllShardIds(final Index index) throws IOException {
assert index != null;
if (nodePaths == null || locks == null) {
throw new IllegalStateException("node is not configured to store local location");
}
assertEnvIsLocked();
final Set<ShardId> shardIds = new HashSet<>();
final String indexUniquePathId = index.getUUID();
for (final NodePath nodePath : nodePaths) {
Path location = nodePath.indicesPath;
if (Files.isDirectory(location)) {
try (DirectoryStream<Path> indexStream = Files.newDirectoryStream(location)) {
for (Path indexPath : indexStream) {
if (indexUniquePathId.equals(indexPath.getFileName().toString())) {
shardIds.addAll(findAllShardsForIndex(indexPath, index));
}
}
}
}
}
return shardIds;
}
private static Set<ShardId> findAllShardsForIndex(Path indexPath, Index index)
throws IOException {
assert indexPath.getFileName().toString().equals(index.getUUID());
Set<ShardId> shardIds = new HashSet<>();
if (Files.isDirectory(indexPath)) {
try (DirectoryStream<Path> stream = Files.newDirectoryStream(indexPath)) {
for (Path shardPath : stream) {
String fileName = shardPath.getFileName().toString();
if (Files.isDirectory(shardPath) && fileName.chars().allMatch(Character::isDigit)) {
int shardId = Integer.parseInt(fileName);
ShardId id = new ShardId(index, shardId);
shardIds.add(id);
}
}
}
}
return shardIds;
}
@Override
public void close() {
if (closed.compareAndSet(false, true) && locks != null) {
for (Lock lock : locks) {
try {
logger.trace("releasing lock [{}]", lock);
lock.close();
} catch (IOException e) {
logger.trace(
(Supplier<?>) () -> new ParameterizedMessage("failed to release lock [{}]", lock), e);
}
}
}
}
private void assertEnvIsLocked() {
if (!closed.get() && locks != null) {
for (Lock lock : locks) {
try {
lock.ensureValid();
} catch (IOException e) {
logger.warn("lock assertion failed", e);
throw new IllegalStateException("environment is not locked", e);
}
}
}
}
/**
* This method tries to write an empty file and moves it using an atomic move operation. This
* method throws an {@link IllegalStateException} if this operation is not supported by the
* filesystem. This test is executed on each of the data directories. This method cleans up all
* files even in the case of an error.
*/
public void ensureAtomicMoveSupported() throws IOException {
final NodePath[] nodePaths = nodePaths();
for (NodePath nodePath : nodePaths) {
assert Files.isDirectory(nodePath.path) : nodePath.path + " is not a directory";
final Path src = nodePath.path.resolve("__es__.tmp");
final Path target = nodePath.path.resolve("__es__.final");
try {
Files.createFile(src);
Files.move(src, target, StandardCopyOption.ATOMIC_MOVE);
} catch (AtomicMoveNotSupportedException ex) {
throw new IllegalStateException(
"atomic_move is not supported by the filesystem on path ["
+ nodePath.path
+ "] atomic_move is required for elasticsearch to work correctly.",
ex);
} finally {
try {
Files.deleteIfExists(src);
} finally {
Files.deleteIfExists(target);
}
}
}
}
/**
* Resolve the custom path for a index's shard. Uses the {@code IndexMetaData.SETTING_DATA_PATH}
* setting to determine the root path for the index.
*
* @param indexSettings settings for the index
*/
public Path resolveBaseCustomLocation(IndexSettings indexSettings) {
String customDataDir = indexSettings.customDataPath();
if (customDataDir != null) {
// This assert is because this should be caught by MetaDataCreateIndexService
assert sharedDataPath != null;
if (ADD_NODE_LOCK_ID_TO_CUSTOM_PATH.get(indexSettings.getNodeSettings())) {
return sharedDataPath.resolve(customDataDir).resolve(Integer.toString(this.nodeLockId));
} else {
return sharedDataPath.resolve(customDataDir);
}
} else {
throw new IllegalArgumentException(
"no custom " + IndexMetaData.SETTING_DATA_PATH + " setting available");
}
}
/**
* Resolve the custom path for a index's shard. Uses the {@code IndexMetaData.SETTING_DATA_PATH}
* setting to determine the root path for the index.
*
* @param indexSettings settings for the index
*/
private Path resolveIndexCustomLocation(IndexSettings indexSettings) {
return resolveBaseCustomLocation(indexSettings).resolve(indexSettings.getUUID());
}
/**
* Resolve the custom path for a index's shard. Uses the {@code IndexMetaData.SETTING_DATA_PATH}
* setting to determine the root path for the index.
*
* @param indexSettings settings for the index
* @param shardId shard to resolve the path to
*/
public Path resolveCustomLocation(IndexSettings indexSettings, final ShardId shardId) {
return resolveIndexCustomLocation(indexSettings).resolve(Integer.toString(shardId.id()));
}
/** Returns the {@code NodePath.path} for this shard. */
public static Path shardStatePathToDataPath(Path shardPath) {
int count = shardPath.getNameCount();
// Sanity check:
assert Integer.parseInt(shardPath.getName(count - 1).toString()) >= 0;
assert "indices".equals(shardPath.getName(count - 3).toString());
return shardPath.getParent().getParent().getParent();
}
/**
* This is a best effort to ensure that we actually have write permissions to write in all our
* data directories. This prevents disasters if nodes are started under the wrong username etc.
*/
private void assertCanWrite() throws IOException {
for (Path path : nodeDataPaths()) { // check node-paths are writable
tryWriteTempFile(path);
}
for (String indexFolderName : this.availableIndexFolders()) {
for (Path indexPath :
this.resolveIndexFolder(indexFolderName)) { // check index paths are writable
Path indexStatePath = indexPath.resolve(MetaDataStateFormat.STATE_DIR_NAME);
tryWriteTempFile(indexStatePath);
tryWriteTempFile(indexPath);
try (DirectoryStream<Path> stream = Files.newDirectoryStream(indexPath)) {
for (Path shardPath : stream) {
String fileName = shardPath.getFileName().toString();
if (Files.isDirectory(shardPath) && fileName.chars().allMatch(Character::isDigit)) {
Path indexDir = shardPath.resolve(ShardPath.INDEX_FOLDER_NAME);
Path statePath = shardPath.resolve(MetaDataStateFormat.STATE_DIR_NAME);
Path translogDir = shardPath.resolve(ShardPath.TRANSLOG_FOLDER_NAME);
tryWriteTempFile(indexDir);
tryWriteTempFile(translogDir);
tryWriteTempFile(statePath);
tryWriteTempFile(shardPath);
}
}
}
}
}
}
private static void tryWriteTempFile(Path path) throws IOException {
if (Files.exists(path)) {
Path resolve = path.resolve(".es_temp_file");
try {
Files.createFile(resolve);
Files.deleteIfExists(resolve);
} catch (IOException ex) {
throw new IOException(
"failed to write in data directory [" + path + "] write permission is required", ex);
}
}
}
}
public class ElectMasterService extends AbstractComponent {
public static final Setting<Integer> DISCOVERY_ZEN_MINIMUM_MASTER_NODES_SETTING =
Setting.intSetting("discovery.zen.minimum_master_nodes", -1, true, Setting.Scope.CLUSTER);
// This is the minimum version a master needs to be on, otherwise it gets ignored
// This is based on the minimum compatible version of the current version this node is on
private final Version minMasterVersion;
private final NodeComparator nodeComparator = new NodeComparator();
private volatile int minimumMasterNodes;
@Inject
public ElectMasterService(Settings settings, Version version) {
super(settings);
this.minMasterVersion = version.minimumCompatibilityVersion();
this.minimumMasterNodes = DISCOVERY_ZEN_MINIMUM_MASTER_NODES_SETTING.get(settings);
logger.debug("using minimum_master_nodes [{}]", minimumMasterNodes);
}
public void minimumMasterNodes(int minimumMasterNodes) {
this.minimumMasterNodes = minimumMasterNodes;
}
public int minimumMasterNodes() {
return minimumMasterNodes;
}
public boolean hasEnoughMasterNodes(Iterable<DiscoveryNode> nodes) {
if (minimumMasterNodes < 1) {
return true;
}
int count = 0;
for (DiscoveryNode node : nodes) {
if (node.masterNode()) {
count++;
}
}
return count >= minimumMasterNodes;
}
/**
* Returns the given nodes sorted by likelyhood of being elected as master, most likely first.
* Non-master nodes are not removed but are rather put in the end
*/
public List<DiscoveryNode> sortByMasterLikelihood(Iterable<DiscoveryNode> nodes) {
ArrayList<DiscoveryNode> sortedNodes = CollectionUtils.iterableAsArrayList(nodes);
CollectionUtil.introSort(sortedNodes, nodeComparator);
return sortedNodes;
}
/** Returns a list of the next possible masters. */
public DiscoveryNode[] nextPossibleMasters(
ObjectContainer<DiscoveryNode> nodes, int numberOfPossibleMasters) {
List<DiscoveryNode> sortedNodes =
sortedMasterNodes(Arrays.asList(nodes.toArray(DiscoveryNode.class)));
if (sortedNodes == null) {
return new DiscoveryNode[0];
}
List<DiscoveryNode> nextPossibleMasters = new ArrayList<>(numberOfPossibleMasters);
int counter = 0;
for (DiscoveryNode nextPossibleMaster : sortedNodes) {
if (++counter >= numberOfPossibleMasters) {
break;
}
nextPossibleMasters.add(nextPossibleMaster);
}
return nextPossibleMasters.toArray(new DiscoveryNode[nextPossibleMasters.size()]);
}
/**
* Elects a new master out of the possible nodes, returning it. Returns <tt>null</tt> if no master
* has been elected.
*/
public DiscoveryNode electMaster(Iterable<DiscoveryNode> nodes) {
List<DiscoveryNode> sortedNodes = sortedMasterNodes(nodes);
if (sortedNodes == null || sortedNodes.isEmpty()) {
return null;
}
DiscoveryNode masterNode = sortedNodes.get(0);
// Sanity check: maybe we don't end up here, because serialization may have failed.
if (masterNode.getVersion().before(minMasterVersion)) {
logger.warn(
"ignoring master [{}], because the version [{}] is lower than the minimum compatible version [{}]",
masterNode,
masterNode.getVersion(),
minMasterVersion);
return null;
} else {
return masterNode;
}
}
private List<DiscoveryNode> sortedMasterNodes(Iterable<DiscoveryNode> nodes) {
List<DiscoveryNode> possibleNodes = CollectionUtils.iterableAsArrayList(nodes);
if (possibleNodes.isEmpty()) {
return null;
}
// clean non master nodes
for (Iterator<DiscoveryNode> it = possibleNodes.iterator(); it.hasNext(); ) {
DiscoveryNode node = it.next();
if (!node.masterNode()) {
it.remove();
}
}
CollectionUtil.introSort(possibleNodes, nodeComparator);
return possibleNodes;
}
private static class NodeComparator implements Comparator<DiscoveryNode> {
@Override
public int compare(DiscoveryNode o1, DiscoveryNode o2) {
if (o1.masterNode() && !o2.masterNode()) {
return -1;
}
if (!o1.masterNode() && o2.masterNode()) {
return 1;
}
return o1.id().compareTo(o2.id());
}
}
}
public class IndicesQueryCache extends AbstractComponent implements QueryCache, Closeable {
public static final Setting<ByteSizeValue> INDICES_CACHE_QUERY_SIZE_SETTING =
Setting.byteSizeSetting("indices.queries.cache.size", "10%", false, Scope.CLUSTER);
public static final Setting<Integer> INDICES_CACHE_QUERY_COUNT_SETTING =
Setting.intSetting("indices.queries.cache.count", 10000, 1, false, Scope.CLUSTER);
private final LRUQueryCache cache;
private final ShardCoreKeyMap shardKeyMap = new ShardCoreKeyMap();
private final Map<ShardId, Stats> shardStats = new ConcurrentHashMap<>();
private volatile long sharedRamBytesUsed;
// This is a hack for the fact that the close listener for the
// ShardCoreKeyMap will be called before onDocIdSetEviction
// See onDocIdSetEviction for more info
private final Map<Object, StatsAndCount> stats2 = new IdentityHashMap<>();
public IndicesQueryCache(Settings settings) {
super(settings);
final ByteSizeValue size = INDICES_CACHE_QUERY_SIZE_SETTING.get(settings);
final int count = INDICES_CACHE_QUERY_COUNT_SETTING.get(settings);
logger.debug("using [node] query cache with size [{}] max filter count [{}]", size, count);
cache =
new LRUQueryCache(count, size.bytes()) {
private Stats getStats(Object coreKey) {
final ShardId shardId = shardKeyMap.getShardId(coreKey);
if (shardId == null) {
return null;
}
return shardStats.get(shardId);
}
private Stats getOrCreateStats(Object coreKey) {
final ShardId shardId = shardKeyMap.getShardId(coreKey);
Stats stats = shardStats.get(shardId);
if (stats == null) {
stats = new Stats();
shardStats.put(shardId, stats);
}
return stats;
}
// It's ok to not protect these callbacks by a lock since it is
// done in LRUQueryCache
@Override
protected void onClear() {
assert Thread.holdsLock(this);
super.onClear();
for (Stats stats : shardStats.values()) {
// don't throw away hit/miss
stats.cacheSize = 0;
stats.ramBytesUsed = 0;
}
sharedRamBytesUsed = 0;
}
@Override
protected void onQueryCache(Query filter, long ramBytesUsed) {
assert Thread.holdsLock(this);
super.onQueryCache(filter, ramBytesUsed);
sharedRamBytesUsed += ramBytesUsed;
}
@Override
protected void onQueryEviction(Query filter, long ramBytesUsed) {
assert Thread.holdsLock(this);
super.onQueryEviction(filter, ramBytesUsed);
sharedRamBytesUsed -= ramBytesUsed;
}
@Override
protected void onDocIdSetCache(Object readerCoreKey, long ramBytesUsed) {
assert Thread.holdsLock(this);
super.onDocIdSetCache(readerCoreKey, ramBytesUsed);
final Stats shardStats = getOrCreateStats(readerCoreKey);
shardStats.cacheSize += 1;
shardStats.cacheCount += 1;
shardStats.ramBytesUsed += ramBytesUsed;
StatsAndCount statsAndCount = stats2.get(readerCoreKey);
if (statsAndCount == null) {
statsAndCount = new StatsAndCount(shardStats);
stats2.put(readerCoreKey, statsAndCount);
}
statsAndCount.count += 1;
}
@Override
protected void onDocIdSetEviction(
Object readerCoreKey, int numEntries, long sumRamBytesUsed) {
assert Thread.holdsLock(this);
super.onDocIdSetEviction(readerCoreKey, numEntries, sumRamBytesUsed);
// onDocIdSetEviction might sometimes be called with a number
// of entries equal to zero if the cache for the given segment
// was already empty when the close listener was called
if (numEntries > 0) {
// We can't use ShardCoreKeyMap here because its core closed
// listener is called before the listener of the cache which
// triggers this eviction. So instead we use use stats2 that
// we only evict when nothing is cached anymore on the segment
// instead of relying on close listeners
final StatsAndCount statsAndCount = stats2.get(readerCoreKey);
final Stats shardStats = statsAndCount.stats;
shardStats.cacheSize -= numEntries;
shardStats.ramBytesUsed -= sumRamBytesUsed;
statsAndCount.count -= numEntries;
if (statsAndCount.count == 0) {
stats2.remove(readerCoreKey);
}
}
}
@Override
protected void onHit(Object readerCoreKey, Query filter) {
assert Thread.holdsLock(this);
super.onHit(readerCoreKey, filter);
final Stats shardStats = getStats(readerCoreKey);
shardStats.hitCount += 1;
}
@Override
protected void onMiss(Object readerCoreKey, Query filter) {
assert Thread.holdsLock(this);
super.onMiss(readerCoreKey, filter);
final Stats shardStats = getOrCreateStats(readerCoreKey);
shardStats.missCount += 1;
}
};
sharedRamBytesUsed = 0;
}
/** Get usage statistics for the given shard. */
public QueryCacheStats getStats(ShardId shard) {
final Map<ShardId, QueryCacheStats> stats = new HashMap<>();
for (Map.Entry<ShardId, Stats> entry : shardStats.entrySet()) {
stats.put(entry.getKey(), entry.getValue().toQueryCacheStats());
}
QueryCacheStats shardStats = new QueryCacheStats();
QueryCacheStats info = stats.get(shard);
if (info == null) {
info = new QueryCacheStats();
}
shardStats.add(info);
// We also have some shared ram usage that we try to distribute to
// proportionally to their number of cache entries of each shard
long totalSize = 0;
for (QueryCacheStats s : stats.values()) {
totalSize += s.getCacheSize();
}
final double weight =
totalSize == 0 ? 1d / stats.size() : shardStats.getCacheSize() / totalSize;
final long additionalRamBytesUsed = Math.round(weight * sharedRamBytesUsed);
shardStats.add(new QueryCacheStats(additionalRamBytesUsed, 0, 0, 0, 0));
return shardStats;
}
@Override
public Weight doCache(Weight weight, QueryCachingPolicy policy) {
while (weight instanceof CachingWeightWrapper) {
weight = ((CachingWeightWrapper) weight).in;
}
final Weight in = cache.doCache(weight, policy);
// We wrap the weight to track the readers it sees and map them with
// the shards they belong to
return new CachingWeightWrapper(in);
}
private class CachingWeightWrapper extends Weight {
private final Weight in;
protected CachingWeightWrapper(Weight in) {
super(in.getQuery());
this.in = in;
}
@Override
public void extractTerms(Set<Term> terms) {
in.extractTerms(terms);
}
@Override
public Explanation explain(LeafReaderContext context, int doc) throws IOException {
shardKeyMap.add(context.reader());
return in.explain(context, doc);
}
@Override
public float getValueForNormalization() throws IOException {
return in.getValueForNormalization();
}
@Override
public void normalize(float norm, float topLevelBoost) {
in.normalize(norm, topLevelBoost);
}
@Override
public Scorer scorer(LeafReaderContext context) throws IOException {
shardKeyMap.add(context.reader());
return in.scorer(context);
}
@Override
public BulkScorer bulkScorer(LeafReaderContext context) throws IOException {
shardKeyMap.add(context.reader());
return in.bulkScorer(context);
}
}
/** Clear all entries that belong to the given index. */
public void clearIndex(String index) {
final Set<Object> coreCacheKeys = shardKeyMap.getCoreKeysForIndex(index);
for (Object coreKey : coreCacheKeys) {
cache.clearCoreCacheKey(coreKey);
}
// This cache stores two things: filters, and doc id sets. Calling
// clear only removes the doc id sets, but if we reach the situation
// that the cache does not contain any DocIdSet anymore, then it
// probably means that the user wanted to remove everything.
if (cache.getCacheSize() == 0) {
cache.clear();
}
}
@Override
public void close() {
assert shardKeyMap.size() == 0 : shardKeyMap.size();
assert shardStats.isEmpty() : shardStats.keySet();
assert stats2.isEmpty() : stats2;
cache.clear();
}
private static class Stats implements Cloneable {
volatile long ramBytesUsed;
volatile long hitCount;
volatile long missCount;
volatile long cacheCount;
volatile long cacheSize;
QueryCacheStats toQueryCacheStats() {
return new QueryCacheStats(ramBytesUsed, hitCount, missCount, cacheCount, cacheSize);
}
}
private static class StatsAndCount {
int count;
final Stats stats;
StatsAndCount(Stats stats) {
this.stats = stats;
this.count = 0;
}
}
private boolean empty(Stats stats) {
if (stats == null) {
return true;
}
return stats.cacheSize == 0 && stats.ramBytesUsed == 0;
}
public void onClose(ShardId shardId) {
assert empty(shardStats.get(shardId));
shardStats.remove(shardId);
}
}