001/**
002 *
003 * Licensed to the Apache Software Foundation (ASF) under one
004 * or more contributor license agreements.  See the NOTICE file
005 * distributed with this work for additional information
006 * regarding copyright ownership.  The ASF licenses this file
007 * to you under the Apache License, Version 2.0 (the
008 * "License"); you may not use this file except in compliance
009 * with the License.  You may obtain a copy of the License at
010 *
011 *     http://www.apache.org/licenses/LICENSE-2.0
012 *
013 * Unless required by applicable law or agreed to in writing, software
014 * distributed under the License is distributed on an "AS IS" BASIS,
015 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
016 * See the License for the specific language governing permissions and
017 * limitations under the License.
018 */
019package org.apache.hadoop.hbase.master;
020
021import java.util.ArrayList;
022import java.util.Collections;
023import java.util.Comparator;
024import java.util.Date;
025import java.util.HashMap;
026import java.util.HashSet;
027import java.util.Iterator;
028import java.util.List;
029import java.util.Map;
030import java.util.Set;
031import org.apache.hadoop.hbase.ServerName;
032import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
033import org.apache.hadoop.hbase.util.Pair;
034import org.apache.yetus.audience.InterfaceAudience;
035import org.slf4j.Logger;
036import org.slf4j.LoggerFactory;
037
038import org.apache.hbase.thirdparty.com.google.common.base.Preconditions;
039
040
041/**
042 * Class to hold dead servers list and utility querying dead server list.
043 * On znode expiration, servers are added here.
044 */
045@InterfaceAudience.Private
046public class DeadServer {
047  private static final Logger LOG = LoggerFactory.getLogger(DeadServer.class);
048
049  /**
050   * Set of known dead servers.  On znode expiration, servers are added here.
051   * This is needed in case of a network partitioning where the server's lease
052   * expires, but the server is still running. After the network is healed,
053   * and it's server logs are recovered, it will be told to call server startup
054   * because by then, its regions have probably been reassigned.
055   */
056  private final Map<ServerName, Long> deadServers = new HashMap<>();
057
058  /**
059   * Set of dead servers currently being processed
060   */
061  private final Set<ServerName> processingServers = new HashSet<ServerName>();
062
063  /**
064   * Handles restart of a server. The new server instance has a different start code.
065   * The new start code should be greater than the old one. We don't check that here.
066   *
067   * @param newServerName Servername as either <code>host:port</code> or
068   *                      <code>host,port,startcode</code>.
069   * @return true if this server was dead before and coming back alive again
070   */
071  public synchronized boolean cleanPreviousInstance(final ServerName newServerName) {
072    Iterator<ServerName> it = deadServers.keySet().iterator();
073    while (it.hasNext()) {
074      ServerName sn = it.next();
075      if (ServerName.isSameAddress(sn, newServerName)) {
076        // remove from deadServers
077        it.remove();
078        // remove from processingServers
079        boolean removed = processingServers.remove(sn);
080        if (removed) {
081          LOG.debug("Removed {}, processing={}, numProcessing={}", sn, removed,
082              processingServers.size());
083        }
084        return true;
085      }
086    }
087
088    return false;
089  }
090
091  /**
092   * @param serverName server name.
093   * @return true if this server is on the dead servers list false otherwise
094   */
095  public synchronized boolean isDeadServer(final ServerName serverName) {
096    return deadServers.containsKey(serverName);
097  }
098
099  /**
100   * @param serverName server name.
101   * @return true if this server is on the processing servers list false otherwise
102   */
103  public synchronized boolean isProcessingServer(final ServerName serverName) {
104    return processingServers.contains(serverName);
105  }
106
107  /**
108   * Checks if there are currently any dead servers being processed by the
109   * master.  Returns true if at least one region server is currently being
110   * processed as dead.
111   *
112   * @return true if any RS are being processed as dead
113   */
114  public synchronized boolean areDeadServersInProgress() {
115    return !processingServers.isEmpty();
116  }
117
118  public synchronized Set<ServerName> copyServerNames() {
119    Set<ServerName> clone = new HashSet<>(deadServers.size());
120    clone.addAll(deadServers.keySet());
121    return clone;
122  }
123
124  /**
125   * Adds the server to the dead server list if it's not there already.
126   */
127  public synchronized void add(ServerName sn) {
128    if (!deadServers.containsKey(sn)){
129      deadServers.put(sn, EnvironmentEdgeManager.currentTime());
130    }
131    boolean added = processingServers.add(sn);
132    if (LOG.isDebugEnabled() && added) {
133      LOG.debug("Added " + sn + "; numProcessing=" + processingServers.size());
134    }
135  }
136
137  /**
138   * Notify that we started processing this dead server.
139   * @param sn ServerName for the dead server.
140   */
141  public synchronized void notifyServer(ServerName sn) {
142    boolean added = processingServers.add(sn);
143    if (LOG.isDebugEnabled()) {
144      if (added) {
145        LOG.debug("Added " + sn + "; numProcessing=" + processingServers.size());
146      }
147      LOG.debug("Started processing " + sn + "; numProcessing=" + processingServers.size());
148    }
149  }
150
151  /**
152   * Complete processing for this dead server.
153   * @param sn ServerName for the dead server.
154   */
155  public synchronized void finish(ServerName sn) {
156    boolean removed = processingServers.remove(sn);
157    if (LOG.isDebugEnabled()) {
158      LOG.debug("Finished processing " + sn + "; numProcessing=" + processingServers.size());
159      if (removed) {
160        LOG.debug("Removed " + sn + " ; numProcessing=" + processingServers.size());
161      }
162    }
163  }
164
165  public synchronized int size() {
166    return deadServers.size();
167  }
168
169  public synchronized boolean isEmpty() {
170    return deadServers.isEmpty();
171  }
172
173  public synchronized void cleanAllPreviousInstances(final ServerName newServerName) {
174    Iterator<ServerName> it = deadServers.keySet().iterator();
175    while (it.hasNext()) {
176      ServerName sn = it.next();
177      if (ServerName.isSameAddress(sn, newServerName)) {
178        // remove from deadServers
179        it.remove();
180        // remove from processingServers
181        boolean removed = processingServers.remove(sn);
182        if (removed) {
183          LOG.debug("Removed " + sn + " ; numProcessing=" + processingServers.size());
184        }
185      }
186    }
187  }
188
189  @Override
190  public synchronized String toString() {
191    // Display unified set of servers from both maps
192    Set<ServerName> servers = new HashSet<ServerName>();
193    servers.addAll(deadServers.keySet());
194    servers.addAll(processingServers);
195    StringBuilder sb = new StringBuilder();
196    for (ServerName sn : servers) {
197      if (sb.length() > 0) {
198        sb.append(", ");
199      }
200      sb.append(sn.toString());
201      // Star entries that are being processed
202      if (processingServers.contains(sn)) {
203        sb.append("*");
204      }
205    }
206    return sb.toString();
207  }
208
209  /**
210   * Extract all the servers dead since a given time, and sort them.
211   * @param ts the time, 0 for all
212   * @return a sorted array list, by death time, lowest values first.
213   */
214  public synchronized List<Pair<ServerName, Long>> copyDeadServersSince(long ts){
215    List<Pair<ServerName, Long>> res =  new ArrayList<>(size());
216
217    for (Map.Entry<ServerName, Long> entry:deadServers.entrySet()){
218      if (entry.getValue() >= ts){
219        res.add(new Pair<>(entry.getKey(), entry.getValue()));
220      }
221    }
222
223    Collections.sort(res, ServerNameDeathDateComparator);
224    return res;
225  }
226  
227  /**
228   * Get the time when a server died
229   * @param deadServerName the dead server name
230   * @return the date when the server died 
231   */
232  public synchronized Date getTimeOfDeath(final ServerName deadServerName){
233    Long time = deadServers.get(deadServerName);
234    return time == null ? null : new Date(time);
235  }
236
237  private static Comparator<Pair<ServerName, Long>> ServerNameDeathDateComparator =
238      new Comparator<Pair<ServerName, Long>>(){
239
240    @Override
241    public int compare(Pair<ServerName, Long> o1, Pair<ServerName, Long> o2) {
242      return o1.getSecond().compareTo(o2.getSecond());
243    }
244  };
245
246  /**
247   * remove the specified dead server
248   * @param deadServerName the dead server name
249   * @return true if this server was removed
250   */
251
252  public synchronized boolean removeDeadServer(final ServerName deadServerName) {
253    Preconditions.checkState(!processingServers.contains(deadServerName),
254      "Asked to remove server still in processingServers set " + deadServerName +
255          " (numProcessing=" + processingServers.size() + ")");
256    if (deadServers.remove(deadServerName) == null) {
257      return false;
258    }
259    return true;
260  }
261}