All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.apache.hadoop.hbase.master.ServerManager Maven / Gradle / Ivy

The newest version!
/**
 *
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you under the Apache License, Version 2.0 (the
 * "License"); you may not use this file except in compliance
 * with the License.  You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
package org.apache.hadoop.hbase.master;

import com.google.common.annotations.VisibleForTesting;
import com.google.protobuf.ByteString;
import com.google.protobuf.ServiceException;

import java.io.IOException;
import java.net.ConnectException;
import java.net.InetAddress;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentNavigableMap;
import java.util.concurrent.ConcurrentSkipListMap;
import java.util.concurrent.CopyOnWriteArrayList;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.hbase.ClockOutOfSyncException;
import org.apache.hadoop.hbase.HConstants;
import org.apache.hadoop.hbase.HRegionInfo;
import org.apache.hadoop.hbase.NotServingRegionException;
import org.apache.hadoop.hbase.RegionLoad;
import org.apache.hadoop.hbase.Server;
import org.apache.hadoop.hbase.ServerLoad;
import org.apache.hadoop.hbase.ServerName;
import org.apache.hadoop.hbase.YouAreDeadException;
import org.apache.hadoop.hbase.ZooKeeperConnectionException;
import org.apache.hadoop.hbase.classification.InterfaceAudience;
import org.apache.hadoop.hbase.client.ClusterConnection;
import org.apache.hadoop.hbase.client.ConnectionFactory;
import org.apache.hadoop.hbase.client.RetriesExhaustedException;
import org.apache.hadoop.hbase.ipc.FailedServerException;
import org.apache.hadoop.hbase.ipc.HBaseRpcController;
import org.apache.hadoop.hbase.ipc.RpcControllerFactory;
import org.apache.hadoop.hbase.master.balancer.BaseLoadBalancer;
import org.apache.hadoop.hbase.master.procedure.MasterProcedureEnv;
import org.apache.hadoop.hbase.master.procedure.ServerCrashProcedure;
import org.apache.hadoop.hbase.monitoring.MonitoredTask;
import org.apache.hadoop.hbase.procedure2.ProcedureExecutor;
import org.apache.hadoop.hbase.protobuf.ProtobufUtil;
import org.apache.hadoop.hbase.protobuf.RequestConverter;
import org.apache.hadoop.hbase.protobuf.ResponseConverter;
import org.apache.hadoop.hbase.protobuf.generated.AdminProtos.AdminService;
import org.apache.hadoop.hbase.protobuf.generated.AdminProtos.OpenRegionRequest;
import org.apache.hadoop.hbase.protobuf.generated.AdminProtos.OpenRegionResponse;
import org.apache.hadoop.hbase.protobuf.generated.AdminProtos.ServerInfo;
import org.apache.hadoop.hbase.protobuf.generated.ClusterStatusProtos.RegionStoreSequenceIds;
import org.apache.hadoop.hbase.protobuf.generated.ClusterStatusProtos.StoreSequenceId;
import org.apache.hadoop.hbase.protobuf.generated.RegionServerStatusProtos.RegionServerStartupRequest;
import org.apache.hadoop.hbase.protobuf.generated.ZooKeeperProtos.SplitLogTask.RecoveryMode;
import org.apache.hadoop.hbase.regionserver.HRegionServer;
import org.apache.hadoop.hbase.regionserver.RegionOpeningState;
import org.apache.hadoop.hbase.security.User;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.hbase.util.RetryCounter;
import org.apache.hadoop.hbase.util.RetryCounterFactory;
import org.apache.hadoop.hbase.util.Triple;
import org.apache.hadoop.hbase.zookeeper.ZKUtil;
import org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher;
import org.apache.zookeeper.KeeperException;

/**
 * The ServerManager class manages info about region servers.
 * 

* Maintains lists of online and dead servers. Processes the startups, * shutdowns, and deaths of region servers. *

* Servers are distinguished in two different ways. A given server has a * location, specified by hostname and port, and of which there can only be one * online at any given time. A server instance is specified by the location * (hostname and port) as well as the startcode (timestamp from when the server * was started). This is used to differentiate a restarted instance of a given * server from the original instance. *

* If a sever is known not to be running any more, it is called dead. The dead * server needs to be handled by a ServerShutdownHandler. If the handler is not * enabled yet, the server can't be handled right away so it is queued up. * After the handler is enabled, the server will be submitted to a handler to handle. * However, the handler may be just partially enabled. If so, * the server cannot be fully processed, and be queued up for further processing. * A server is fully processed only after the handler is fully enabled * and has completed the handling. */ @InterfaceAudience.Private @edu.umd.cs.findbugs.annotations.SuppressWarnings(value="JLM_JSR166_UTILCONCURRENT_MONITORENTER", justification="Synchronization on concurrent map is intended") public class ServerManager { public static final String WAIT_ON_REGIONSERVERS_MAXTOSTART = "hbase.master.wait.on.regionservers.maxtostart"; public static final String WAIT_ON_REGIONSERVERS_MINTOSTART = "hbase.master.wait.on.regionservers.mintostart"; public static final String WAIT_ON_REGIONSERVERS_TIMEOUT = "hbase.master.wait.on.regionservers.timeout"; public static final String WAIT_ON_REGIONSERVERS_INTERVAL = "hbase.master.wait.on.regionservers.interval"; private static final Log LOG = LogFactory.getLog(ServerManager.class); // Set if we are to shutdown the cluster. private volatile boolean clusterShutdown = false; /** * The last flushed sequence id for a region. */ private final ConcurrentNavigableMap flushedSequenceIdByRegion = new ConcurrentSkipListMap(Bytes.BYTES_COMPARATOR); /** * The last flushed sequence id for a store in a region. */ private final ConcurrentNavigableMap> storeFlushedSequenceIdsByRegion = new ConcurrentSkipListMap>(Bytes.BYTES_COMPARATOR); /** Map of registered servers to their current load */ private final ConcurrentNavigableMap onlineServers = new ConcurrentSkipListMap(); /** * Map of admin interfaces per registered regionserver; these interfaces we use to control * regionservers out on the cluster */ private final Map rsAdmins = new HashMap(); /** * List of region servers that should not get any more new * regions. */ private final ArrayList drainingServers = new ArrayList(); private final Server master; private final MasterServices services; private final ClusterConnection connection; private final DeadServer deadservers = new DeadServer(); private final long maxSkew; private final long warningSkew; private final RetryCounterFactory pingRetryCounterFactory; private final RpcControllerFactory rpcControllerFactory; /** * Set of region servers which are dead but not processed immediately. If one * server died before master enables ServerShutdownHandler, the server will be * added to this set and will be processed through calling * {@link ServerManager#processQueuedDeadServers()} by master. *

* A dead server is a server instance known to be dead, not listed in the /hbase/rs * znode any more. It may have not been submitted to ServerShutdownHandler yet * because the handler is not enabled. *

* A dead server, which has been submitted to ServerShutdownHandler while the * handler is not enabled, is queued up. *

* So this is a set of region servers known to be dead but not submitted to * ServerShutdownHander for processing yet. */ private Set queuedDeadServers = new HashSet(); /** * Set of region servers which are dead and submitted to ServerShutdownHandler to process but not * fully processed immediately. *

* If one server died before assignment manager finished the failover cleanup, the server will be * added to this set and will be processed through calling * {@link ServerManager#processQueuedDeadServers()} by assignment manager. *

* The Boolean value indicates whether log split is needed inside ServerShutdownHandler *

* ServerShutdownHandler processes a dead server submitted to the handler after the handler is * enabled. It may not be able to complete the processing because meta is not yet online or master * is currently in startup mode. In this case, the dead server will be parked in this set * temporarily. */ private Map requeuedDeadServers = new ConcurrentHashMap(); /** Listeners that are called on server events. */ private List listeners = new CopyOnWriteArrayList(); /** * Constructor. * @param master * @param services * @throws ZooKeeperConnectionException */ public ServerManager(final Server master, final MasterServices services) throws IOException { this(master, services, true); } ServerManager(final Server master, final MasterServices services, final boolean connect) throws IOException { this.master = master; this.services = services; Configuration c = master.getConfiguration(); maxSkew = c.getLong("hbase.master.maxclockskew", 30000); warningSkew = c.getLong("hbase.master.warningclockskew", 10000); this.connection = connect ? (ClusterConnection)ConnectionFactory.createConnection(c) : null; int pingMaxAttempts = Math.max(1, master.getConfiguration().getInt( "hbase.master.maximum.ping.server.attempts", 10)); int pingSleepInterval = Math.max(1, master.getConfiguration().getInt( "hbase.master.ping.server.retry.sleep.interval", 100)); this.pingRetryCounterFactory = new RetryCounterFactory(pingMaxAttempts, pingSleepInterval); this.rpcControllerFactory = this.connection == null ? null : connection.getRpcControllerFactory(); } /** * Add the listener to the notification list. * @param listener The ServerListener to register */ public void registerListener(final ServerListener listener) { this.listeners.add(listener); } /** * Remove the listener from the notification list. * @param listener The ServerListener to unregister */ public boolean unregisterListener(final ServerListener listener) { return this.listeners.remove(listener); } /** * Let the server manager know a new regionserver has come online * @param request the startup request * @param ia the InetAddress from which request is received * @return The ServerName we know this server as. * @throws IOException */ ServerName regionServerStartup(RegionServerStartupRequest request, InetAddress ia) throws IOException { // Test for case where we get a region startup message from a regionserver // that has been quickly restarted but whose znode expiration handler has // not yet run, or from a server whose fail we are currently processing. // Test its host+port combo is present in serverAddresstoServerInfo. If it // is, reject the server and trigger its expiration. The next time it comes // in, it should have been removed from serverAddressToServerInfo and queued // for processing by ProcessServerShutdown. final String hostname = request.hasUseThisHostnameInstead() ? request.getUseThisHostnameInstead() :ia.getHostName(); ServerName sn = ServerName.valueOf(hostname, request.getPort(), request.getServerStartCode()); checkClockSkew(sn, request.getServerCurrentTime()); checkIsDead(sn, "STARTUP"); if (!checkAndRecordNewServer(sn, ServerLoad.EMPTY_SERVERLOAD)) { LOG.warn("THIS SHOULD NOT HAPPEN, RegionServerStartup" + " could not record the server: " + sn); } return sn; } private ConcurrentNavigableMap getOrCreateStoreFlushedSequenceId( byte[] regionName) { ConcurrentNavigableMap storeFlushedSequenceId = storeFlushedSequenceIdsByRegion.get(regionName); if (storeFlushedSequenceId != null) { return storeFlushedSequenceId; } storeFlushedSequenceId = new ConcurrentSkipListMap(Bytes.BYTES_COMPARATOR); ConcurrentNavigableMap alreadyPut = storeFlushedSequenceIdsByRegion.putIfAbsent(regionName, storeFlushedSequenceId); return alreadyPut == null ? storeFlushedSequenceId : alreadyPut; } /** * Updates last flushed sequence Ids for the regions on server sn * @param sn * @param hsl */ private void updateLastFlushedSequenceIds(ServerName sn, ServerLoad hsl) { Map regionsLoad = hsl.getRegionsLoad(); for (Entry entry : regionsLoad.entrySet()) { byte[] encodedRegionName = Bytes.toBytes(HRegionInfo.encodeRegionName(entry.getKey())); Long existingValue = flushedSequenceIdByRegion.get(encodedRegionName); long l = entry.getValue().getCompleteSequenceId(); // Don't let smaller sequence ids override greater sequence ids. if (LOG.isTraceEnabled()) { LOG.trace(Bytes.toString(encodedRegionName) + ", existingValue=" + existingValue + ", completeSequenceId=" + l); } if (existingValue == null || (l != HConstants.NO_SEQNUM && l > existingValue)) { flushedSequenceIdByRegion.put(encodedRegionName, l); } else if (l != HConstants.NO_SEQNUM && l < existingValue) { LOG.warn("RegionServer " + sn + " indicates a last flushed sequence id (" + l + ") that is less than the previous last flushed sequence id (" + existingValue + ") for region " + Bytes.toString(entry.getKey()) + " Ignoring."); } ConcurrentNavigableMap storeFlushedSequenceId = getOrCreateStoreFlushedSequenceId(encodedRegionName); for (StoreSequenceId storeSeqId : entry.getValue().getStoreCompleteSequenceId()) { byte[] family = storeSeqId.getFamilyName().toByteArray(); existingValue = storeFlushedSequenceId.get(family); l = storeSeqId.getSequenceId(); if (LOG.isTraceEnabled()) { LOG.trace(Bytes.toString(encodedRegionName) + ", family=" + Bytes.toString(family) + ", existingValue=" + existingValue + ", completeSequenceId=" + l); } // Don't let smaller sequence ids override greater sequence ids. if (existingValue == null || (l != HConstants.NO_SEQNUM && l > existingValue.longValue())) { storeFlushedSequenceId.put(family, l); } } } } void regionServerReport(ServerName sn, ServerLoad sl) throws YouAreDeadException { checkIsDead(sn, "REPORT"); if (null == this.onlineServers.replace(sn, sl)) { // Already have this host+port combo and its just different start code? // Just let the server in. Presume master joining a running cluster. // recordNewServer is what happens at the end of reportServerStartup. // The only thing we are skipping is passing back to the regionserver // the ServerName to use. Here we presume a master has already done // that so we'll press on with whatever it gave us for ServerName. if (!checkAndRecordNewServer(sn, sl)) { LOG.info("RegionServerReport ignored, could not record the server: " + sn); return; // Not recorded, so no need to move on } } updateLastFlushedSequenceIds(sn, sl); } /** * Check is a server of same host and port already exists, * if not, or the existed one got a smaller start code, record it. * * @param sn the server to check and record * @param sl the server load on the server * @return true if the server is recorded, otherwise, false */ boolean checkAndRecordNewServer( final ServerName serverName, final ServerLoad sl) { ServerName existingServer = null; synchronized (this.onlineServers) { existingServer = findServerWithSameHostnamePortWithLock(serverName); if (existingServer != null && (existingServer.getStartcode() > serverName.getStartcode())) { LOG.info("Server serverName=" + serverName + " rejected; we already have " + existingServer.toString() + " registered with same hostname and port"); return false; } recordNewServerWithLock(serverName, sl); } // Tell our listeners that a server was added if (!this.listeners.isEmpty()) { for (ServerListener listener : this.listeners) { listener.serverAdded(serverName); } } // Note that we assume that same ts means same server, and don't expire in that case. // TODO: ts can theoretically collide due to clock shifts, so this is a bit hacky. if (existingServer != null && (existingServer.getStartcode() < serverName.getStartcode())) { LOG.info("Triggering server recovery; existingServer " + existingServer + " looks stale, new server:" + serverName); expireServer(existingServer); } return true; } /** * Checks if the clock skew between the server and the master. If the clock skew exceeds the * configured max, it will throw an exception; if it exceeds the configured warning threshold, * it will log a warning but start normally. * @param serverName Incoming servers's name * @param serverCurrentTime * @throws ClockOutOfSyncException if the skew exceeds the configured max value */ private void checkClockSkew(final ServerName serverName, final long serverCurrentTime) throws ClockOutOfSyncException { long skew = Math.abs(System.currentTimeMillis() - serverCurrentTime); if (skew > maxSkew) { String message = "Server " + serverName + " has been " + "rejected; Reported time is too far out of sync with master. " + "Time difference of " + skew + "ms > max allowed of " + maxSkew + "ms"; LOG.warn(message); throw new ClockOutOfSyncException(message); } else if (skew > warningSkew){ String message = "Reported time for server " + serverName + " is out of sync with master " + "by " + skew + "ms. (Warning threshold is " + warningSkew + "ms; " + "error threshold is " + maxSkew + "ms)"; LOG.warn(message); } } /** * If this server is on the dead list, reject it with a YouAreDeadException. * If it was dead but came back with a new start code, remove the old entry * from the dead list. * @param serverName * @param what START or REPORT * @throws org.apache.hadoop.hbase.YouAreDeadException */ private void checkIsDead(final ServerName serverName, final String what) throws YouAreDeadException { if (this.deadservers.isDeadServer(serverName)) { // host name, port and start code all match with existing one of the // dead servers. So, this server must be dead. String message = "Server " + what + " rejected; currently processing " + serverName + " as dead server"; LOG.debug(message); throw new YouAreDeadException(message); } // remove dead server with same hostname and port of newly checking in rs after master // initialization.See HBASE-5916 for more information. if ((this.services == null || ((HMaster) this.services).isInitialized()) && this.deadservers.cleanPreviousInstance(serverName)) { // This server has now become alive after we marked it as dead. // We removed it's previous entry from the dead list to reflect it. LOG.debug(what + ":" + " Server " + serverName + " came back up," + " removed it from the dead servers list"); } } /** * Assumes onlineServers is locked. * @return ServerName with matching hostname and port. */ private ServerName findServerWithSameHostnamePortWithLock( final ServerName serverName) { ServerName end = ServerName.valueOf(serverName.getHostname(), serverName.getPort(), Long.MAX_VALUE); ServerName r = onlineServers.lowerKey(end); if (r != null) { if (ServerName.isSameHostnameAndPort(r, serverName)) { return r; } } return null; } /** * Adds the onlineServers list. onlineServers should be locked. * @param serverName The remote servers name. * @param sl * @return Server load from the removed server, if any. */ @VisibleForTesting void recordNewServerWithLock(final ServerName serverName, final ServerLoad sl) { LOG.info("Registering server=" + serverName); this.onlineServers.put(serverName, sl); this.rsAdmins.remove(serverName); } public RegionStoreSequenceIds getLastFlushedSequenceId(byte[] encodedRegionName) { RegionStoreSequenceIds.Builder builder = RegionStoreSequenceIds.newBuilder(); Long seqId = flushedSequenceIdByRegion.get(encodedRegionName); builder.setLastFlushedSequenceId(seqId != null ? seqId.longValue() : HConstants.NO_SEQNUM); Map storeFlushedSequenceId = storeFlushedSequenceIdsByRegion.get(encodedRegionName); if (storeFlushedSequenceId != null) { for (Map.Entry entry : storeFlushedSequenceId.entrySet()) { builder.addStoreSequenceId(StoreSequenceId.newBuilder() .setFamilyName(ByteString.copyFrom(entry.getKey())) .setSequenceId(entry.getValue().longValue()).build()); } } return builder.build(); } /** * @param serverName * @return ServerLoad if serverName is known else null */ public ServerLoad getLoad(final ServerName serverName) { return this.onlineServers.get(serverName); } /** * Compute the average load across all region servers. * Currently, this uses a very naive computation - just uses the number of * regions being served, ignoring stats about number of requests. * @return the average load */ public double getAverageLoad() { int totalLoad = 0; int numServers = 0; for (ServerLoad sl: this.onlineServers.values()) { numServers++; totalLoad += sl.getNumberOfRegions(); } return numServers == 0 ? 0 : (double)totalLoad / (double)numServers; } /** @return the count of active regionservers */ public int countOfRegionServers() { // Presumes onlineServers is a concurrent map return this.onlineServers.size(); } /** * @return Read-only map of servers to serverinfo */ public Map getOnlineServers() { // Presumption is that iterating the returned Map is OK. synchronized (this.onlineServers) { return Collections.unmodifiableMap(this.onlineServers); } } public DeadServer getDeadServers() { return this.deadservers; } /** * Checks if any dead servers are currently in progress. * @return true if any RS are being processed as dead, false if not */ public boolean areDeadServersInProgress() { return this.deadservers.areDeadServersInProgress(); } void letRegionServersShutdown() { long previousLogTime = 0; ServerName sn = master.getServerName(); ZooKeeperWatcher zkw = master.getZooKeeper(); int onlineServersCt; while ((onlineServersCt = onlineServers.size()) > 0){ if (System.currentTimeMillis() > (previousLogTime + 1000)) { Set remainingServers = onlineServers.keySet(); synchronized (onlineServers) { if (remainingServers.size() == 1 && remainingServers.contains(sn)) { // Master will delete itself later. return; } } StringBuilder sb = new StringBuilder(); // It's ok here to not sync on onlineServers - merely logging for (ServerName key : remainingServers) { if (sb.length() > 0) { sb.append(", "); } sb.append(key); } LOG.info("Waiting on regionserver(s) to go down " + sb.toString()); previousLogTime = System.currentTimeMillis(); } try { List servers = ZKUtil.listChildrenNoWatch(zkw, zkw.rsZNode); if (servers == null || servers.size() == 0 || (servers.size() == 1 && servers.contains(sn.toString()))) { LOG.info("ZK shows there is only the master self online, exiting now"); // Master could have lost some ZK events, no need to wait more. break; } } catch (KeeperException ke) { LOG.warn("Failed to list regionservers", ke); // ZK is malfunctioning, don't hang here break; } synchronized (onlineServers) { try { if (onlineServersCt == onlineServers.size()) onlineServers.wait(100); } catch (InterruptedException ignored) { // continue } } } } private List getRegionServersInZK(final ZooKeeperWatcher zkw) throws KeeperException { return ZKUtil.listChildrenNoWatch(zkw, zkw.rsZNode); } /* * Expire the passed server. Add it to list of dead servers and queue a * shutdown processing. */ public synchronized void expireServer(final ServerName serverName) { if (serverName.equals(master.getServerName())) { if (!(master.isAborted() || master.isStopped())) { master.stop("We lost our znode?"); } return; } if (!services.isServerCrashProcessingEnabled()) { LOG.info("Master doesn't enable ServerShutdownHandler during initialization, " + "delay expiring server " + serverName); this.queuedDeadServers.add(serverName); return; } if (this.deadservers.isDeadServer(serverName)) { // TODO: Can this happen? It shouldn't be online in this case? LOG.warn("Expiration of " + serverName + " but server shutdown already in progress"); return; } moveFromOnlineToDeadServers(serverName); // If cluster is going down, yes, servers are going to be expiring; don't // process as a dead server if (this.clusterShutdown) { LOG.info("Cluster shutdown set; " + serverName + " expired; onlineServers=" + this.onlineServers.size()); if (this.onlineServers.isEmpty()) { master.stop("Cluster shutdown set; onlineServer=0"); } return; } boolean carryingMeta = services.getAssignmentManager().isCarryingMeta(serverName) == AssignmentManager.ServerHostRegion.HOSTING_REGION; ProcedureExecutor procExec = this.services.getMasterProcedureExecutor(); procExec.submitProcedure(new ServerCrashProcedure( procExec.getEnvironment(), serverName, true, carryingMeta)); LOG.debug("Added=" + serverName + " to dead servers, submitted shutdown handler to be executed meta=" + carryingMeta); // Tell our listeners that a server was removed if (!this.listeners.isEmpty()) { for (ServerListener listener : this.listeners) { listener.serverRemoved(serverName); } } } @VisibleForTesting public void moveFromOnlineToDeadServers(final ServerName sn) { synchronized (onlineServers) { if (!this.onlineServers.containsKey(sn)) { LOG.warn("Expiration of " + sn + " but server not online"); } // Remove the server from the known servers lists and update load info BUT // add to deadservers first; do this so it'll show in dead servers list if // not in online servers list. this.deadservers.add(sn); this.onlineServers.remove(sn); onlineServers.notifyAll(); } this.rsAdmins.remove(sn); } public synchronized void processDeadServer(final ServerName serverName, boolean shouldSplitWal) { // When assignment manager is cleaning up the zookeeper nodes and rebuilding the // in-memory region states, region servers could be down. Meta table can and // should be re-assigned, log splitting can be done too. However, it is better to // wait till the cleanup is done before re-assigning user regions. // // We should not wait in the server shutdown handler thread since it can clog // the handler threads and meta table could not be re-assigned in case // the corresponding server is down. So we queue them up here instead. if (!services.getAssignmentManager().isFailoverCleanupDone()) { requeuedDeadServers.put(serverName, shouldSplitWal); return; } this.deadservers.add(serverName); ProcedureExecutor procExec = this.services.getMasterProcedureExecutor(); procExec.submitProcedure(new ServerCrashProcedure( procExec.getEnvironment(), serverName, shouldSplitWal, false)); } /** * Process the servers which died during master's initialization. It will be * called after HMaster#assignMeta and AssignmentManager#joinCluster. * */ synchronized void processQueuedDeadServers() { if (!services.isServerCrashProcessingEnabled()) { LOG.info("Master hasn't enabled ServerShutdownHandler"); } Iterator serverIterator = queuedDeadServers.iterator(); while (serverIterator.hasNext()) { ServerName tmpServerName = serverIterator.next(); expireServer(tmpServerName); serverIterator.remove(); requeuedDeadServers.remove(tmpServerName); } if (!services.getAssignmentManager().isFailoverCleanupDone()) { LOG.info("AssignmentManager hasn't finished failover cleanup; waiting"); } for (Map.Entry entry : requeuedDeadServers.entrySet()) { processDeadServer(entry.getKey(), entry.getValue()); } requeuedDeadServers.clear(); } /* * Remove the server from the drain list. */ public boolean removeServerFromDrainList(final ServerName sn) { // Warn if the server (sn) is not online. ServerName is of the form: // , , if (!this.isServerOnline(sn)) { LOG.warn("Server " + sn + " is not currently online. " + "Removing from draining list anyway, as requested."); } // Remove the server from the draining servers lists. return this.drainingServers.remove(sn); } /* * Add the server to the drain list. */ public boolean addServerToDrainList(final ServerName sn) { // Warn if the server (sn) is not online. ServerName is of the form: // , , if (!this.isServerOnline(sn)) { LOG.warn("Server " + sn + " is not currently online. " + "Ignoring request to add it to draining list."); return false; } // Add the server to the draining servers lists, if it's not already in // it. if (this.drainingServers.contains(sn)) { LOG.warn("Server " + sn + " is already in the draining server list." + "Ignoring request to add it again."); return false; } LOG.info("Server " + sn + " added to draining server list."); return this.drainingServers.add(sn); } // RPC methods to region servers /** * Sends an OPEN RPC to the specified server to open the specified region. *

* Open should not fail but can if server just crashed. *

* @param server server to open a region * @param region region to open * @param versionOfOfflineNode that needs to be present in the offline node * when RS tries to change the state from OFFLINE to other states. * @param favoredNodes */ public RegionOpeningState sendRegionOpen(final ServerName server, HRegionInfo region, int versionOfOfflineNode, List favoredNodes) throws IOException { AdminService.BlockingInterface admin = getRsAdmin(server); if (admin == null) { LOG.warn("Attempting to send OPEN RPC to server " + server.toString() + " failed because no RPC connection found to this server"); return RegionOpeningState.FAILED_OPENING; } OpenRegionRequest request = RequestConverter.buildOpenRegionRequest(server, region, versionOfOfflineNode, favoredNodes, (RecoveryMode.LOG_REPLAY == this.services.getMasterFileSystem().getLogRecoveryMode())); try { OpenRegionResponse response = admin.openRegion(null, request); return ResponseConverter.getRegionOpeningState(response); } catch (ServiceException se) { checkForRSznode(server, se); throw ProtobufUtil.getRemoteException(se); } } /** * Check for an odd state, where we think an RS is up but it is not. Do it on OPEN. * This is only case where the check makes sense. * *

We are checking for instance of HBASE-9593 where a RS registered but died before it put * up its znode in zk. In this case, the RS made it into the list of online servers but it * is not actually UP. We do the check here where there is an evident problem rather * than do some crazy footwork where we'd have master check zk after a RS had reported * for duty with provisional state followed by a confirmed state; that'd be a mess. * Real fix is HBASE-17733. */ private void checkForRSznode(final ServerName serverName, final ServiceException se) { if (se.getCause() == null) return; Throwable t = se.getCause(); if (t instanceof ConnectException) { // If this, proceed to do cleanup. } else { // Look for FailedServerException if (!(t instanceof IOException)) return; if (t.getCause() == null) return; if (!(t.getCause() instanceof FailedServerException)) return; // Ok, found FailedServerException -- continue. } if (!isServerOnline(serverName)) return; // We think this server is online. Check it has a znode up. Currently, a RS // registers an ephereral znode in zk. If not present, something is up. Maybe // HBASE-9593 where RS crashed AFTER reportForDuty but BEFORE it put up an ephemeral // znode. List servers = null; try { servers = getRegionServersInZK(this.master.getZooKeeper()); } catch (KeeperException ke) { LOG.warn("Failed to list regionservers", ke); // ZK is malfunctioning, don't hang here } boolean found = false; if (servers != null) { for (String serverNameAsStr: servers) { ServerName sn = ServerName.valueOf(serverNameAsStr); if (sn.equals(serverName)) { // Found a server up in zk. found = true; break; } } } if (!found) { LOG.warn("Online server " + serverName.toString() + " has no corresponding " + "ephemeral znode (Did it die before registering in zk?); " + "calling expire to clean it up!"); expireServer(serverName); } } /** * Sends an OPEN RPC to the specified server to open the specified region. *

* Open should not fail but can if server just crashed. *

* @param server server to open a region * @param regionOpenInfos info of a list of regions to open * @return a list of region opening states */ public List sendRegionOpen(ServerName server, List>> regionOpenInfos) throws IOException { AdminService.BlockingInterface admin = getRsAdmin(server); if (admin == null) { LOG.warn("Attempting to send OPEN RPC to server " + server.toString() + " failed because no RPC connection found to this server"); return null; } OpenRegionRequest request = RequestConverter.buildOpenRegionRequest(server, regionOpenInfos, (RecoveryMode.LOG_REPLAY == this.services.getMasterFileSystem().getLogRecoveryMode())); try { OpenRegionResponse response = admin.openRegion(null, request); return ResponseConverter.getRegionOpeningStateList(response); } catch (ServiceException se) { checkForRSznode(server, se); throw ProtobufUtil.getRemoteException(se); } } private HBaseRpcController newRpcController() { return rpcControllerFactory == null ? null : rpcControllerFactory.newController(); } /** * Sends an CLOSE RPC to the specified server to close the specified region. *

* A region server could reject the close request because it either does not * have the specified region or the region is being split. * @param server server to open a region * @param region region to open * @param versionOfClosingNode * the version of znode to compare when RS transitions the znode from * CLOSING state. * @param dest - if the region is moved to another server, the destination server. null otherwise. * @return true if server acknowledged close, false if not * @throws IOException */ public boolean sendRegionClose(ServerName server, HRegionInfo region, int versionOfClosingNode, ServerName dest, boolean transitionInZK) throws IOException { if (server == null) throw new NullPointerException("Passed server is null"); AdminService.BlockingInterface admin = getRsAdmin(server); if (admin == null) { throw new IOException("Attempting to send CLOSE RPC to server " + server.toString() + " for region " + region.getRegionNameAsString() + " failed because no RPC connection found to this server"); } HBaseRpcController controller = newRpcController(); return ProtobufUtil.closeRegion(controller, admin, server, region.getRegionName(), versionOfClosingNode, dest, transitionInZK); } public boolean sendRegionClose(ServerName server, HRegionInfo region, int versionOfClosingNode) throws IOException { return sendRegionClose(server, region, versionOfClosingNode, null, true); } /** * Sends a WARMUP RPC to the specified server to warmup the specified region. *

* A region server could reject the close request because it either does not * have the specified region or the region is being split. * @param server server to warmup a region * @param region region to warmup */ public void sendRegionWarmup(ServerName server, HRegionInfo region) { if (server == null) return; try { AdminService.BlockingInterface admin = getRsAdmin(server); HBaseRpcController controller = newRpcController(); ProtobufUtil.warmupRegion(controller, admin, region); } catch (IOException e) { LOG.error("Received exception in RPC for warmup server:" + server + "region: " + region + "exception: " + e); } } /** * Contacts a region server and waits up to timeout ms * to close the region. This bypasses the active hmaster. */ public static void closeRegionSilentlyAndWait(ClusterConnection connection, ServerName server, HRegionInfo region, long timeout) throws IOException, InterruptedException { AdminService.BlockingInterface rs = connection.getAdmin(server); HBaseRpcController controller = connection.getRpcControllerFactory().newController(); try { ProtobufUtil.closeRegion(controller, rs, server, region.getRegionName(), false); } catch (IOException e) { LOG.warn("Exception when closing region: " + region.getRegionNameAsString(), e); } long expiration = timeout + System.currentTimeMillis(); while (System.currentTimeMillis() < expiration) { controller.reset(); try { HRegionInfo rsRegion = ProtobufUtil.getRegionInfo(controller, rs, region.getRegionName()); if (rsRegion == null) return; } catch (IOException ioe) { if (ioe instanceof NotServingRegionException) // no need to retry again return; LOG.warn("Exception when retrieving regioninfo from: " + region.getRegionNameAsString(), ioe); } Thread.sleep(1000); } throw new IOException("Region " + region + " failed to close within" + " timeout " + timeout); } /** * Sends an MERGE REGIONS RPC to the specified server to merge the specified * regions. *

* A region server could reject the close request because it either does not * have the specified region. * @param server server to merge regions * @param region_a region to merge * @param region_b region to merge * @param forcible true if do a compulsory merge, otherwise we will only merge * two adjacent regions * @throws IOException */ public void sendRegionsMerge(ServerName server, HRegionInfo region_a, HRegionInfo region_b, boolean forcible, User user) throws IOException { if (server == null) throw new NullPointerException("Passed server is null"); if (region_a == null || region_b == null) throw new NullPointerException("Passed region is null"); AdminService.BlockingInterface admin = getRsAdmin(server); if (admin == null) { throw new IOException("Attempting to send MERGE REGIONS RPC to server " + server.toString() + " for region " + region_a.getRegionNameAsString() + "," + region_b.getRegionNameAsString() + " failed because no RPC connection found to this server"); } HBaseRpcController controller = newRpcController(); ProtobufUtil.mergeRegions(controller, admin, region_a, region_b, forcible, user); } /** * Check if a region server is reachable and has the expected start code */ public boolean isServerReachable(ServerName server) { if (server == null) throw new NullPointerException("Passed server is null"); RetryCounter retryCounter = pingRetryCounterFactory.create(); while (retryCounter.shouldRetry()) { synchronized (this.onlineServers) { if (this.deadservers.isDeadServer(server)) { return false; } } try { HBaseRpcController controller = newRpcController(); AdminService.BlockingInterface admin = getRsAdmin(server); if (admin != null) { ServerInfo info = ProtobufUtil.getServerInfo(controller, admin); return info != null && info.hasServerName() && server.getStartcode() == info.getServerName().getStartCode(); } } catch (IOException ioe) { if (LOG.isDebugEnabled()) { LOG.debug("Couldn't reach " + server + ", try=" + retryCounter.getAttemptTimes() + " of " + retryCounter.getMaxAttempts(), ioe); } try { retryCounter.sleepUntilNextRetry(); } catch(InterruptedException ie) { Thread.currentThread().interrupt(); break; } } } return false; } /** * @param sn * @return Admin interface for the remote regionserver named sn * @throws IOException * @throws RetriesExhaustedException wrapping a ConnectException if failed */ private AdminService.BlockingInterface getRsAdmin(final ServerName sn) throws IOException { AdminService.BlockingInterface admin = this.rsAdmins.get(sn); if (admin == null) { LOG.debug("New admin connection to " + sn.toString()); if (sn.equals(master.getServerName()) && master instanceof HRegionServer) { // A master is also a region server now, see HBASE-10569 for details admin = ((HRegionServer)master).getRSRpcServices(); } else { admin = this.connection.getAdmin(sn); } this.rsAdmins.put(sn, admin); } return admin; } /** * Calculate min necessary to start. This is not an absolute. It is just * a friction that will cause us hang around a bit longer waiting on * RegionServers to check-in. */ private int getMinToStart() { // One server should be enough to get us off the ground. int requiredMinToStart = 1; if (BaseLoadBalancer.tablesOnMaster(master.getConfiguration())) { if (!BaseLoadBalancer.userTablesOnMaster(master.getConfiguration())) { // If Master is carrying regions but NOT user-space regions (the current default), // since the Master shows as a 'server', we need at least one more server to check // in before we can start up so up defaultMinToStart to 2. requiredMinToStart = 2; } } int minToStart = this.master.getConfiguration().getInt(WAIT_ON_REGIONSERVERS_MINTOSTART, -1); // Ensure we are never less than requiredMinToStart else stuff won't work. return minToStart == -1 || minToStart < requiredMinToStart? requiredMinToStart: minToStart; } /** * Wait for the region servers to report in. * We will wait until one of this condition is met: * - the master is stopped * - the 'hbase.master.wait.on.regionservers.maxtostart' number of * region servers is reached * - the 'hbase.master.wait.on.regionservers.mintostart' is reached AND * there have been no new region server in for * 'hbase.master.wait.on.regionservers.interval' time AND * the 'hbase.master.wait.on.regionservers.timeout' is reached * * @throws InterruptedException */ public void waitForRegionServers(MonitoredTask status) throws InterruptedException { final long interval = this.master.getConfiguration(). getLong(WAIT_ON_REGIONSERVERS_INTERVAL, 1500); final long timeout = this.master.getConfiguration(). getLong(WAIT_ON_REGIONSERVERS_TIMEOUT, 4500); // Min is not an absolute; just a friction making us wait longer on server checkin. int minToStart = getMinToStart(); int maxToStart = this.master.getConfiguration(). getInt(WAIT_ON_REGIONSERVERS_MAXTOSTART, Integer.MAX_VALUE); if (maxToStart < minToStart) { LOG.warn(String.format("The value of '%s' (%d) is set less than '%s' (%d), ignoring.", WAIT_ON_REGIONSERVERS_MAXTOSTART, maxToStart, WAIT_ON_REGIONSERVERS_MINTOSTART, minToStart)); maxToStart = Integer.MAX_VALUE; } long now = System.currentTimeMillis(); final long startTime = now; long slept = 0; long lastLogTime = 0; long lastCountChange = startTime; int count = countOfRegionServers(); int oldCount = 0; // This while test is a little hard to read. We try to comment it in below but in essence: // Wait if Master is not stopped and the number of regionservers that have checked-in is // less than the maxToStart. Both of these conditions will be true near universally. // Next, we will keep cycling if ANY of the following three conditions are true: // 1. The time since a regionserver registered is < interval (means servers are actively checking in). // 2. We are under the total timeout. // 3. The count of servers is < minimum. for (ServerListener listener: this.listeners) { listener.waiting(); } while (!this.master.isStopped() && count < maxToStart && ((lastCountChange + interval) > now || timeout > slept || count < minToStart)) { // Log some info at every interval time or if there is a change if (oldCount != count || lastLogTime + interval < now) { lastLogTime = now; String msg = "Waiting on RegionServer count=" + count + " to settle; waited="+ slept + "ms, expecting min=" + minToStart + " server(s), max="+ getStrForMax(maxToStart) + " server(s), " + "timeout=" + timeout + "ms, lastChange=" + (lastCountChange - now) + "ms"; LOG.info(msg); status.setStatus(msg); } // We sleep for some time final long sleepTime = 50; Thread.sleep(sleepTime); now = System.currentTimeMillis(); slept = now - startTime; oldCount = count; count = countOfRegionServers(); if (count != oldCount) { lastCountChange = now; } } LOG.info("Finished wait on RegionServer count=" + count + "; waited=" + slept + "ms," + " expected min=" + minToStart + " server(s), max=" + getStrForMax(maxToStart) + " server(s),"+ " master is "+ (this.master.isStopped() ? "stopped.": "running") ); } private String getStrForMax(final int max) { return max == Integer.MAX_VALUE? "NO_LIMIT": Integer.toString(max); } /** * @return A copy of the internal list of online servers. */ public List getOnlineServersList() { // TODO: optimize the load balancer call so we don't need to make a new list // TODO: FIX. THIS IS POPULAR CALL. return new ArrayList(this.onlineServers.keySet()); } /** * @return A copy of the internal list of draining servers. */ public List getDrainingServersList() { return new ArrayList(this.drainingServers); } /** * @return A copy of the internal set of deadNotExpired servers. */ Set getDeadNotExpiredServers() { return new HashSet(this.queuedDeadServers); } /** * During startup, if we figure it is not a failover, i.e. there is * no more WAL files to split, we won't try to recover these dead servers. * So we just remove them from the queue. Use caution in calling this. */ void removeRequeuedDeadServers() { requeuedDeadServers.clear(); } /** * @return A copy of the internal map of requeuedDeadServers servers and their corresponding * splitlog need flag. */ Map getRequeuedDeadServers() { return Collections.unmodifiableMap(this.requeuedDeadServers); } public boolean isServerOnline(ServerName serverName) { return serverName != null && onlineServers.containsKey(serverName); } /** * Check whether a server is online based on hostname and port * @return true if finding a server with matching hostname and port. */ public boolean isServerWithSameHostnamePortOnline(final ServerName serverName) { return findServerWithSameHostnamePortWithLock(serverName) != null; } /** * Check if a server is known to be dead. A server can be online, * or known to be dead, or unknown to this manager (i.e, not online, * not known to be dead either. it is simply not tracked by the * master any more, for example, a very old previous instance). */ public synchronized boolean isServerDead(ServerName serverName) { return serverName == null || deadservers.isDeadServer(serverName) || queuedDeadServers.contains(serverName) || requeuedDeadServers.containsKey(serverName); } public void shutdownCluster() { this.clusterShutdown = true; this.master.stop("Cluster shutdown requested"); } public boolean isClusterShutdown() { return this.clusterShutdown; } /** * Stop the ServerManager. Currently closes the connection to the master. */ public void stop() { if (connection != null) { try { connection.close(); } catch (IOException e) { LOG.error("Attempt to close connection to master failed", e); } } } /** * Creates a list of possible destinations for a region. It contains the online servers, but not * the draining or dying servers. * @param serversToExclude can be null if there is no server to exclude */ public List createDestinationServersList(final List serversToExclude){ final List destServers = getOnlineServersList(); if (serversToExclude != null){ destServers.removeAll(serversToExclude); } // Loop through the draining server list and remove them from the server list final List drainingServersCopy = getDrainingServersList(); if (!drainingServersCopy.isEmpty()) { for (final ServerName server: drainingServersCopy) { destServers.remove(server); } } // Remove the deadNotExpired servers from the server list. removeDeadNotExpiredServers(destServers); return destServers; } /** * Calls {@link #createDestinationServersList} without server to exclude. */ public List createDestinationServersList(){ return createDestinationServersList(null); } /** * Loop through the deadNotExpired server list and remove them from the * servers. * This function should be used carefully outside of this class. You should use a high level * method such as {@link #createDestinationServersList()} instead of managing you own list. */ void removeDeadNotExpiredServers(List servers) { Set deadNotExpiredServersCopy = this.getDeadNotExpiredServers(); if (!deadNotExpiredServersCopy.isEmpty()) { for (ServerName server : deadNotExpiredServersCopy) { LOG.debug("Removing dead but not expired server: " + server + " from eligible server pool."); servers.remove(server); } } } /** * To clear any dead server with same host name and port of any online server */ void clearDeadServersWithSameHostNameAndPortOfOnlineServer() { for (ServerName serverName : getOnlineServersList()) { deadservers.cleanAllPreviousInstances(serverName); } } /** * Called by delete table and similar to notify the ServerManager that a region was removed. */ public void removeRegion(final HRegionInfo regionInfo) { final byte[] encodedName = regionInfo.getEncodedNameAsBytes(); storeFlushedSequenceIdsByRegion.remove(encodedName); flushedSequenceIdByRegion.remove(encodedName); } @VisibleForTesting public boolean isRegionInServerManagerStates(final HRegionInfo hri) { final byte[] encodedName = hri.getEncodedNameAsBytes(); return (storeFlushedSequenceIdsByRegion.containsKey(encodedName) || flushedSequenceIdByRegion.containsKey(encodedName)); } /** * Called by delete table and similar to notify the ServerManager that a region was removed. */ public void removeRegions(final List regions) { for (HRegionInfo hri: regions) { removeRegion(hri); } } }





© 2015 - 2024 Weber Informatics LLC | Privacy Policy