View Javadoc

1   /**
2    *
3    * Licensed to the Apache Software Foundation (ASF) under one
4    * or more contributor license agreements.  See the NOTICE file
5    * distributed with this work for additional information
6    * regarding copyright ownership.  The ASF licenses this file
7    * to you under the Apache License, Version 2.0 (the
8    * "License"); you may not use this file except in compliance
9    * with the License.  You may obtain a copy of the License at
10   *
11   *     http://www.apache.org/licenses/LICENSE-2.0
12   *
13   * Unless required by applicable law or agreed to in writing, software
14   * distributed under the License is distributed on an "AS IS" BASIS,
15   * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
16   * See the License for the specific language governing permissions and
17   * limitations under the License.
18   */
19  package org.apache.hadoop.hbase.master;
20  
21  import java.io.File;
22  import java.io.IOException;
23  import java.util.List;
24  
25  import org.apache.commons.cli.CommandLine;
26  import org.apache.commons.cli.GnuParser;
27  import org.apache.commons.cli.Options;
28  import org.apache.commons.cli.ParseException;
29  import org.apache.commons.logging.Log;
30  import org.apache.commons.logging.LogFactory;
31  import org.apache.hadoop.hbase.classification.InterfaceAudience;
32  import org.apache.hadoop.conf.Configuration;
33  import org.apache.hadoop.hbase.CoordinatedStateManager;
34  import org.apache.hadoop.hbase.CoordinatedStateManagerFactory;
35  import org.apache.hadoop.hbase.HConstants;
36  import org.apache.hadoop.hbase.LocalHBaseCluster;
37  import org.apache.hadoop.hbase.MasterNotRunningException;
38  import org.apache.hadoop.hbase.ZNodeClearer;
39  import org.apache.hadoop.hbase.ZooKeeperConnectionException;
40  import org.apache.hadoop.hbase.client.Admin;
41  import org.apache.hadoop.hbase.client.HBaseAdmin;
42  import org.apache.hadoop.hbase.regionserver.HRegionServer;
43  import org.apache.hadoop.hbase.util.JVMClusterUtil;
44  import org.apache.hadoop.hbase.util.ServerCommandLine;
45  import org.apache.hadoop.hbase.zookeeper.MiniZooKeeperCluster;
46  import org.apache.hadoop.hbase.zookeeper.ZKUtil;
47  import org.apache.hadoop.metrics2.lib.DefaultMetricsSystem;
48  import org.apache.zookeeper.KeeperException;
49  
50  @InterfaceAudience.Private
51  public class HMasterCommandLine extends ServerCommandLine {
52    private static final Log LOG = LogFactory.getLog(HMasterCommandLine.class);
53  
54    private static final String USAGE =
55      "Usage: Master [opts] start|stop|clear\n" +
56      " start  Start Master. If local mode, start Master and RegionServer in same JVM\n" +
57      " stop   Start cluster shutdown; Master signals RegionServer shutdown\n" +
58      " clear  Delete the master znode in ZooKeeper after a master crashes\n "+
59      " where [opts] are:\n" +
60      "   --minRegionServers=<servers>   Minimum RegionServers needed to host user tables.\n" +
61      "   --localRegionServers=<servers> " +
62        "RegionServers to start in master process when in standalone mode.\n" +
63      "   --masters=<servers>            Masters to start in this process.\n" +
64      "   --backup                       Master should start in backup mode";
65  
66    private final Class<? extends HMaster> masterClass;
67  
68    public HMasterCommandLine(Class<? extends HMaster> masterClass) {
69      this.masterClass = masterClass;
70    }
71  
72    protected String getUsage() {
73      return USAGE;
74    }
75  
76  
77    public int run(String args[]) throws Exception {
78      Options opt = new Options();
79      opt.addOption("localRegionServers", true,
80        "RegionServers to start in master process when running standalone");
81      opt.addOption("masters", true, "Masters to start in this process");
82      opt.addOption("minRegionServers", true, "Minimum RegionServers needed to host user tables");
83      opt.addOption("backup", false, "Do not try to become HMaster until the primary fails");
84  
85      CommandLine cmd;
86      try {
87        cmd = new GnuParser().parse(opt, args);
88      } catch (ParseException e) {
89        LOG.error("Could not parse: ", e);
90        usage(null);
91        return 1;
92      }
93  
94  
95      if (cmd.hasOption("minRegionServers")) {
96        String val = cmd.getOptionValue("minRegionServers");
97        getConf().setInt("hbase.regions.server.count.min",
98                    Integer.valueOf(val));
99        LOG.debug("minRegionServers set to " + val);
100     }
101 
102     // minRegionServers used to be minServers.  Support it too.
103     if (cmd.hasOption("minServers")) {
104       String val = cmd.getOptionValue("minServers");
105       getConf().setInt("hbase.regions.server.count.min", Integer.parseInt(val));
106       LOG.debug("minServers set to " + val);
107     }
108 
109     // check if we are the backup master - override the conf if so
110     if (cmd.hasOption("backup")) {
111       getConf().setBoolean(HConstants.MASTER_TYPE_BACKUP, true);
112     }
113 
114     // How many regionservers to startup in this process (we run regionservers in same process as
115     // master when we are in local/standalone mode. Useful testing)
116     if (cmd.hasOption("localRegionServers")) {
117       String val = cmd.getOptionValue("localRegionServers");
118       getConf().setInt("hbase.regionservers", Integer.valueOf(val));
119       LOG.debug("localRegionServers set to " + val);
120     }
121     // How many masters to startup inside this process; useful testing
122     if (cmd.hasOption("masters")) {
123       String val = cmd.getOptionValue("masters");
124       getConf().setInt("hbase.masters", Integer.valueOf(val));
125       LOG.debug("masters set to " + val);
126     }
127 
128     @SuppressWarnings("unchecked")
129     List<String> remainingArgs = cmd.getArgList();
130     if (remainingArgs.size() != 1) {
131       usage(null);
132       return 1;
133     }
134 
135     String command = remainingArgs.get(0);
136 
137     if ("start".equals(command)) {
138       return startMaster();
139     } else if ("stop".equals(command)) {
140       return stopMaster();
141     } else if ("clear".equals(command)) {
142       return (ZNodeClearer.clear(getConf()) ? 0 : 1);
143     } else {
144       usage("Invalid command: " + command);
145       return 1;
146     }
147   }
148 
149   private int startMaster() {
150     Configuration conf = getConf();
151     try {
152       // If 'local', defer to LocalHBaseCluster instance.  Starts master
153       // and regionserver both in the one JVM.
154       if (LocalHBaseCluster.isLocal(conf)) {
155         DefaultMetricsSystem.setMiniClusterMode(true);
156         final MiniZooKeeperCluster zooKeeperCluster = new MiniZooKeeperCluster(conf);
157         File zkDataPath = new File(conf.get(HConstants.ZOOKEEPER_DATA_DIR));
158 
159         // find out the default client port
160         int zkClientPort = 0;
161 
162         // If the zookeeper client port is specified in server quorum, use it.
163         String zkserver = conf.get(HConstants.ZOOKEEPER_QUORUM);
164         if (zkserver != null) {
165           String[] zkservers = zkserver.split(",");
166 
167           if (zkservers.length > 1) {
168             // In local mode deployment, we have the master + a region server and zookeeper server
169             // started in the same process. Therefore, we only support one zookeeper server.
170             String errorMsg = "Could not start ZK with " + zkservers.length +
171                 " ZK servers in local mode deployment. Aborting as clients (e.g. shell) will not "
172                 + "be able to find this ZK quorum.";
173               System.err.println(errorMsg);
174               throw new IOException(errorMsg);
175           }
176 
177           String[] parts = zkservers[0].split(":");
178 
179           if (parts.length == 2) {
180             // the second part is the client port
181             zkClientPort = Integer.parseInt(parts [1]);
182           }
183         }
184         // If the client port could not be find in server quorum conf, try another conf
185         if (zkClientPort == 0) {
186           zkClientPort = conf.getInt(HConstants.ZOOKEEPER_CLIENT_PORT, 0);
187           // The client port has to be set by now; if not, throw exception.
188           if (zkClientPort == 0) {
189             throw new IOException("No config value for " + HConstants.ZOOKEEPER_CLIENT_PORT);
190           }
191         }
192         zooKeeperCluster.setDefaultClientPort(zkClientPort);
193         // set the ZK tick time if specified
194         int zkTickTime = conf.getInt(HConstants.ZOOKEEPER_TICK_TIME, 0);
195         if (zkTickTime > 0) {
196           zooKeeperCluster.setTickTime(zkTickTime);
197         }
198 
199         // login the zookeeper server principal (if using security)
200         ZKUtil.loginServer(conf, HConstants.ZK_SERVER_KEYTAB_FILE,
201           HConstants.ZK_SERVER_KERBEROS_PRINCIPAL, null);
202         int localZKClusterSessionTimeout =
203           conf.getInt(HConstants.ZK_SESSION_TIMEOUT + ".localHBaseCluster", 10*1000);
204         conf.setInt(HConstants.ZK_SESSION_TIMEOUT, localZKClusterSessionTimeout);
205         LOG.info("Starting a zookeeper cluster");
206         int clientPort = zooKeeperCluster.startup(zkDataPath);
207         if (clientPort != zkClientPort) {
208           String errorMsg = "Could not start ZK at requested port of " +
209             zkClientPort + ".  ZK was started at port: " + clientPort +
210             ".  Aborting as clients (e.g. shell) will not be able to find " +
211             "this ZK quorum.";
212           System.err.println(errorMsg);
213           throw new IOException(errorMsg);
214         }
215         conf.set(HConstants.ZOOKEEPER_CLIENT_PORT, Integer.toString(clientPort));
216 
217         // Need to have the zk cluster shutdown when master is shutdown.
218         // Run a subclass that does the zk cluster shutdown on its way out.
219         int mastersCount = conf.getInt("hbase.masters", 1);
220         int regionServersCount = conf.getInt("hbase.regionservers", 1);
221         LOG.info("Starting up instance of localHBaseCluster; master=" + mastersCount +
222           ", regionserversCount=" + regionServersCount);
223         LocalHBaseCluster cluster = new LocalHBaseCluster(conf, mastersCount, regionServersCount,
224           LocalHMaster.class, HRegionServer.class);
225         ((LocalHMaster)cluster.getMaster(0)).setZKCluster(zooKeeperCluster);
226         cluster.startup();
227         waitOnMasterThreads(cluster);
228       } else {
229         logProcessInfo(getConf());
230         CoordinatedStateManager csm =
231           CoordinatedStateManagerFactory.getCoordinatedStateManager(conf);
232         HMaster master = HMaster.constructMaster(masterClass, conf, csm);
233         if (master.isStopped()) {
234           LOG.info("Won't bring the Master up as a shutdown is requested");
235           return 1;
236         }
237         master.start();
238         master.join();
239         if(master.isAborted())
240           throw new RuntimeException("HMaster Aborted");
241       }
242     } catch (Throwable t) {
243       LOG.error("Master exiting", t);
244       return 1;
245     }
246     return 0;
247   }
248 
249   @SuppressWarnings("resource")
250   private int stopMaster() {
251     Admin adm = null;
252     try {
253       Configuration conf = getConf();
254       // Don't try more than once
255       conf.setInt(HConstants.HBASE_CLIENT_RETRIES_NUMBER, 1);
256       adm = new HBaseAdmin(getConf());
257     } catch (MasterNotRunningException e) {
258       LOG.error("Master not running");
259       return 1;
260     } catch (ZooKeeperConnectionException e) {
261       LOG.error("ZooKeeper not available");
262       return 1;
263     } catch (IOException e) {
264       LOG.error("Got IOException: " +e.getMessage(), e);
265       return 1;
266     }
267     try {
268       adm.shutdown();
269     } catch (Throwable t) {
270       LOG.error("Failed to stop master", t);
271       return 1;
272     }
273     return 0;
274   }
275 
276   private void waitOnMasterThreads(LocalHBaseCluster cluster) throws InterruptedException{
277     List<JVMClusterUtil.MasterThread> masters = cluster.getMasters();
278     List<JVMClusterUtil.RegionServerThread> regionservers = cluster.getRegionServers();
279 
280     if (masters != null) {
281       for (JVMClusterUtil.MasterThread t : masters) {
282         t.join();
283         if(t.getMaster().isAborted()) {
284           closeAllRegionServerThreads(regionservers);
285           throw new RuntimeException("HMaster Aborted");
286         }
287       }
288     }
289   }
290 
291   private static void closeAllRegionServerThreads(
292       List<JVMClusterUtil.RegionServerThread> regionservers) {
293     for(JVMClusterUtil.RegionServerThread t : regionservers){
294       t.getRegionServer().stop("HMaster Aborted; Bringing down regions servers");
295     }
296   }
297 
298   /*
299    * Version of master that will shutdown the passed zk cluster on its way out.
300    */
301   public static class LocalHMaster extends HMaster {
302     private MiniZooKeeperCluster zkcluster = null;
303 
304     public LocalHMaster(Configuration conf, CoordinatedStateManager csm)
305     throws IOException, KeeperException, InterruptedException {
306       super(conf, csm);
307     }
308 
309     @Override
310     public void run() {
311       super.run();
312       if (this.zkcluster != null) {
313         try {
314           this.zkcluster.shutdown();
315         } catch (IOException e) {
316           e.printStackTrace();
317         }
318       }
319     }
320 
321     void setZKCluster(final MiniZooKeeperCluster zkcluster) {
322       this.zkcluster = zkcluster;
323     }
324   }
325 }