ZooKeeperWatcher源代码

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.zookeeper;
20 
21  import java.io.Closeable;
22  import java.io.IOException;
23  import java.util.ArrayList;
24  import java.util.List;
25  import java.util.concurrent.CopyOnWriteArrayList;
26  import java.util.concurrent.CountDownLatch;
27 
28  import org.apache.commons.logging.Log;
29  import org.apache.commons.logging.LogFactory;
30  import org.apache.hadoop.classification.InterfaceAudience;
31  import org.apache.hadoop.conf.Configuration;
32  import org.apache.hadoop.hbase.Abortable;
33  import org.apache.hadoop.hbase.HConstants;
34  import org.apache.hadoop.hbase.ZooKeeperConnectionException;
35  import org.apache.hadoop.hbase.util.Threads;
36  import org.apache.zookeeper.KeeperException;
37  import org.apache.zookeeper.WatchedEvent;
38  import org.apache.zookeeper.Watcher;
39  import org.apache.zookeeper.ZooDefs;
40  import org.apache.zookeeper.data.ACL;
41 
42  /**
43  * Acts as the single ZooKeeper Watcher.  One instance of this is instantiated
44  * for each Master, RegionServer, and client process.
45  *
46  * <p>This is the only class that implements {@link Watcher}.  Other internal
47  * classes which need to be notified of ZooKeeper events must register with
48  * the local instance of this watcher via {@link #registerListener}.
49  *
50  * <p>This class also holds and manages the connection to ZooKeeper.  Code to
51  * deal with connection related events and exceptions are handled here.
52  */
53  @InterfaceAudience.Private
54  public class ZooKeeperWatcher implements Watcher, Abortable, Closeable {
55    private static final Log LOG = LogFactory.getLog(ZooKeeperWatcher. class);
56 
57    // Identifier for this watcher (for logging only).  It is made of the prefix
58    // passed on construction and the zookeeper sessionid.
59    private String identifier;
60 
61    // zookeeper quorum
62    private String quorum;
63 
64    // zookeeper connection
65    private RecoverableZooKeeper recoverableZooKeeper;
66 
67    // abortable in case of zk failure
68    protected Abortable abortable;
69    // Used if abortable is null
70    private boolean aborted = false;
71 
72    // listeners to be notified
73    private final List<ZooKeeperListener> listeners =
74      new CopyOnWriteArrayList<ZooKeeperListener>();
75 
76    // Used by ZKUtil:waitForZKConnectionIfAuthenticating to wait for SASL
77    // negotiation to complete
78    public CountDownLatch saslLatch = new CountDownLatch(1);
79 
80    // node names
81 
82    // base znode for this cluster
83    public String baseZNode;
84    // znode containing location of server hosting meta region
85    public String metaServerZNode;
86    // znode containing ephemeral nodes of the regionservers
87    public String rsZNode;
88    // znode containing ephemeral nodes of the draining regionservers
89    public String drainingZNode;
90    // znode of currently active master
91    private String masterAddressZNode;
92    // znode of this master in backup master directory, if not the active master
93    public String backupMasterAddressesZNode;
94    // znode containing the current cluster state
95    public String clusterStateZNode;
96    // znode used for region transitioning and assignment
97    public String assignmentZNode;
98    // znode used for table disabling/enabling
99    public String tableZNode;
100   // znode containing the unique cluster ID
101   public String clusterIdZNode;
102   // znode used for log splitting work assignment
103   public String splitLogZNode;
104   // znode containing the state of the load balancer
105   public String balancerZNode;
106   // znode containing the lock for the tables
107   public String tableLockZNode;
108   // znode containing the state of recovering regions
109   public String recoveringRegionsZNode;
110   // znode containing namespace descriptors
111   public static String namespaceZNode = "namespace";
112
113
114   // Certain ZooKeeper nodes need to be world-readable
115   public static final ArrayList<ACL> CREATOR_ALL_AND_WORLD_READABLE =
116     new ArrayList<ACL>() { {
117       add( new ACL(ZooDefs.Perms.READ,ZooDefs.Ids.ANYONE_ID_UNSAFE));
118       add( new ACL(ZooDefs.Perms.ALL,ZooDefs.Ids.AUTH_IDS));
119     }};
120
121   private final Configuration conf;
122
123   private final Exception constructorCaller;
124
125   /**
126    * Instantiate a ZooKeeper connection and watcher.
127    * @param identifier string that is passed to RecoverableZookeeper to be used as
128    * identifier for this instance. Use null for default.
129    * @throws IOException
130    * @throws ZooKeeperConnectionException
131    */
132   public ZooKeeperWatcher(Configuration conf, String identifier,
133       Abortable abortable) throws ZooKeeperConnectionException, IOException {
134     this(conf, identifier, abortable, false);
135   }
136
137   /**
138    * Instantiate a ZooKeeper connection and watcher.
139    * @param conf
140    * @param identifier string that is passed to RecoverableZookeeper to be used as identifier for
141    *          this instance. Use null for default.
142    * @param abortable Can be null if there is on error there is no host to abort: e.g. client
143    *          context.
144    * @param canCreateBaseZNode
145    * @throws IOException
146    * @throws ZooKeeperConnectionException
147    */
148   public ZooKeeperWatcher(Configuration conf, String identifier,
149       Abortable abortable, boolean canCreateBaseZNode)
150   throws IOException, ZooKeeperConnectionException {
151     this.conf = conf;
152     // Capture a stack trace now.  Will print it out later if problem so we can
153     // distingush amongst the myriad ZKWs.
154     try {
155       throw new Exception( "ZKW CONSTRUCTOR STACK TRACE FOR DEBUGGING");
156     } catch (Exception e) {
157       this.constructorCaller = e;
158     }
159     this.quorum = ZKConfig.getZKQuorumServersString(conf);
160     // Identifier will get the sessionid appended later below down when we
161     // handle the syncconnect event.
162     this.identifier = identifier;
163     this.abortable = abortable;
164     setNodeNames(conf);
165     this.recoverableZooKeeper = ZKUtil.connect(conf, quorum, this, identifier);
166     if (canCreateBaseZNode) {
167       createBaseZNodes();
168     }
169   }
170
171   private void createBaseZNodes() throws ZooKeeperConnectionException {
172     try {
173       // Create all the necessary "directories" of znodes
174       ZKUtil.createWithParents( this, baseZNode);
175       ZKUtil.createAndFailSilent( this, assignmentZNode);
176       ZKUtil.createAndFailSilent( this, rsZNode);
177       ZKUtil.createAndFailSilent( this, drainingZNode);
178       ZKUtil.createAndFailSilent( this, tableZNode);
179       ZKUtil.createAndFailSilent( this, splitLogZNode);
180       ZKUtil.createAndFailSilent( this, backupMasterAddressesZNode);
181       ZKUtil.createAndFailSilent( this, tableLockZNode);
182       ZKUtil.createAndFailSilent( this, recoveringRegionsZNode);
183     } catch (KeeperException e) {
184       throw new ZooKeeperConnectionException(
185           prefix( "Unexpected KeeperException creating base node"), e);
186     }
187   }
188
189   @Override
190   public String toString() {
191     return this.identifier + ", quorum=" + quorum + ", baseZNode=" + baseZNode;
192   }
193
194   /**
195    * Adds this instance's identifier as a prefix to the passed <code>str</code>
196    * @param str String to amend.
197    * @return A new string with this instance's identifier as prefix: e.g.
198    * if passed 'hello world', the returned string could be
199    */
200   public String prefix( final String str) {
201     return this.toString() + " " + str;
202   }
203
204   /**
205    * Set the local variable node names using the specified configuration.
206    */
207   private void setNodeNames(Configuration conf) {
208     baseZNode = conf.get(HConstants.ZOOKEEPER_ZNODE_PARENT,
209         HConstants.DEFAULT_ZOOKEEPER_ZNODE_PARENT);
210     metaServerZNode = ZKUtil.joinZNode(baseZNode,
211         conf.get( "zookeeper.znode.metaserver", "meta-region-server"));
212     rsZNode = ZKUtil.joinZNode(baseZNode,
213         conf.get( "zookeeper.znode.rs", "rs"));
214     drainingZNode = ZKUtil.joinZNode(baseZNode,
215         conf.get( "zookeeper.znode.draining.rs", "draining"));
216     masterAddressZNode = ZKUtil.joinZNode(baseZNode,
217         conf.get( "zookeeper.znode.master", "master"));
218     backupMasterAddressesZNode = ZKUtil.joinZNode(baseZNode,
219         conf.get( "zookeeper.znode.backup.masters", "backup-masters"));
220     clusterStateZNode = ZKUtil.joinZNode(baseZNode,
221         conf.get( "zookeeper.znode.state", "running"));
222     assignmentZNode = ZKUtil.joinZNode(baseZNode,
223         conf.get( "zookeeper.znode.unassigned", "region-in-transition"));
224     tableZNode = ZKUtil.joinZNode(baseZNode,
225         conf.get( "zookeeper.znode.tableEnableDisable", "table"));
226     clusterIdZNode = ZKUtil.joinZNode(baseZNode,
227         conf.get( "zookeeper.znode.clusterId", "hbaseid"));
228     splitLogZNode = ZKUtil.joinZNode(baseZNode,
229         conf.get( "zookeeper.znode.splitlog", HConstants.SPLIT_LOGDIR_NAME));
230     balancerZNode = ZKUtil.joinZNode(baseZNode,
231         conf.get( "zookeeper.znode.balancer", "balancer"));
232     tableLockZNode = ZKUtil.joinZNode(baseZNode,
233         conf.get( "zookeeper.znode.tableLock", "table-lock"));
234     recoveringRegionsZNode = ZKUtil.joinZNode(baseZNode,
235         conf.get( "zookeeper.znode.recovering.regions", "recovering-regions"));
236     namespaceZNode = ZKUtil.joinZNode(baseZNode,
237         conf.get( "zookeeper.znode.namespace", "namespace"));
238   }
239
240   /**
241    * Register the specified listener to receive ZooKeeper events.
242    * @param listener
243    */
244   public void registerListener( ZooKeeperListener listener) {
245     listeners.add(listener);
246   }
247
248   /**
249    * Register the specified listener to receive ZooKeeper events and add it as
250    * the first in the list of current listeners.
251    * @param listener
252    */
253   public void registerListenerFirst( ZooKeeperListener listener) {
254     listeners.add(0, listener);
255   }
256
257   public void unregisterListener( ZooKeeperListener listener) {
258     listeners.remove(listener);
259   }
260
261   /**
262    * Clean all existing listeners
263    */
264   public void unregisterAllListeners() {
265     listeners.clear();
266   }
267
268   /**
269    * Get a copy of current registered listeners
270    */
271   public List<ZooKeeperListener> getListeners() {
272     return new ArrayList<ZooKeeperListener>(listeners);
273   }
274
275   /**
276    * @return The number of currently registered listeners
277    */
278   public int getNumberOfListeners() {
279     return listeners.size();
280   }
281
282   /**
283    * Get the connection to ZooKeeper.
284    * @return connection reference to zookeeper
285    */
286   public RecoverableZooKeeper getRecoverableZooKeeper() {
287     return recoverableZooKeeper;
288   }
289
290   public void reconnectAfterExpiration() throws IOException, InterruptedException {
291     recoverableZooKeeper.reconnectAfterExpiration();
292   }
293
294   /**
295    * Get the quorum address of this instance.
296    * @return quorum string of this zookeeper connection instance
297    */
298   public String getQuorum() {
299     return quorum;
300   }
301
302   /**
303    * Method called from ZooKeeper for events and connection status.
304    * <p>
305    * Valid events are passed along to listeners.  Connection status changes
306    * are dealt with locally.
307    */
308   @Override
309   public void process(WatchedEvent event) {
310     LOG.debug(prefix( "Received ZooKeeper Event, " +
311         "type=" + event.getType() + ", " +
312         "state=" + event.getState() + ", " +
313         "path=" + event.getPath()));
314
315     switch(event.getType()) {
316
317       // If event type is NONE, this is a connection status change
318       case None: {
319         connectionEvent(event);
320         break;
321       }
322
323       // Otherwise pass along to the listeners
324
325       case NodeCreated: {
326         for( ZooKeeperListener listener : listeners) {
327           listener.nodeCreated(event.getPath());
328         }
329         break;
330       }
331
332       case NodeDeleted: {
333         for( ZooKeeperListener listener : listeners) {
334           listener.nodeDeleted(event.getPath());
335         }
336         break;
337       }
338
339       case NodeDataChanged: {
340         for( ZooKeeperListener listener : listeners) {
341           listener.nodeDataChanged(event.getPath());
342         }
343         break;
344       }
345
346       case NodeChildrenChanged: {
347         for( ZooKeeperListener listener : listeners) {
348           listener.nodeChildrenChanged(event.getPath());
349         }
350         break;
351       }
352     }
353   }
354
355   // Connection management
356
357   /**
358    * Called when there is a connection-related event via the Watcher callback.
359    * <p>
360    * If Disconnected or Expired, this should shutdown the cluster. But, since
361    * we send a KeeperException.SessionExpiredException along with the abort
362    * call, it's possible for the Abortable to catch it and try to create a new
363    * session with ZooKeeper. This is what the client does in HCM.
364    * <p>
365    * @param event
366    */
367   private void connectionEvent(WatchedEvent event) {
368     switch(event.getState()) {
369       case SyncConnected:
370         // Now, this callback can be invoked before the this.zookeeper is set.
371         // Wait a little while.
372         long finished = System.currentTimeMillis() +
373           this.conf.getLong( "hbase.zookeeper.watcher.sync.connected.wait", 2000);
374         while (System.currentTimeMillis() < finished) {
375           Threads.sleep(1);
376           if ( this.recoverableZooKeeper != null) break;
377         }
378         if ( this.recoverableZooKeeper == null) {
379           LOG.error( "ZK is null on connection event -- see stack trace " +
380             "for the stack trace when constructor was called on this zkw",
381             this.constructorCaller);
382           throw new NullPointerException( "ZK is null");
383         }
384         this.identifier = this.identifier + "-0x" +
385           Long.toHexString( this.recoverableZooKeeper.getSessionId());
386         // Update our identifier.  Otherwise ignore.
387         LOG.debug( this.identifier + " connected");
388         break;
389
390       // Abort the server if Disconnected or Expired
391       case Disconnected:
392         LOG.debug(prefix( "Received Disconnected from ZooKeeper, ignoring"));
393         break;
394
395       case Expired:
396         String msg = prefix( this.identifier + " received expired from " +
397           "ZooKeeper, aborting");
398         // TODO: One thought is to add call to ZooKeeperListener so say,
399         // ZooKeeperNodeTracker can zero out its data values.
400         if ( this.abortable != null) {
401           this.abortable.abort(msg, new KeeperException.SessionExpiredException());
402         }
403         break;
404
405       case ConnectedReadOnly:
406       case SaslAuthenticated:
407         break;
408
409       default:
410         throw new IllegalStateException( "Received event is not valid: " + event.getState());
411     }
412   }
413
414   /**
415    * Forces a synchronization of this ZooKeeper client connection.
416    * <p>
417    * Executing this method before running other methods will ensure that the
418    * subsequent operations are up-to-date and consistent as of the time that
419    * the sync is complete.
420    * <p>
421    * This is used for compareAndSwap type operations where we need to read the
422    * data of an existing node and delete or transition that node, utilizing the
423    * previously read version and data.  We want to ensure that the version read
424    * is up-to-date from when we begin the operation.
425    */
426   public void sync(String path) {
427     this.recoverableZooKeeper.sync(path, null, null);
428   }
429
430   /**
431    * Handles KeeperExceptions in client calls.
432    * <p>
433    * This may be temporary but for now this gives one place to deal with these.
434    * <p>
435    * TODO: Currently this method rethrows the exception to let the caller handle
436    * <p>
437    * @param ke
438    * @throws KeeperException
439    */
440   public void keeperException(KeeperException ke)
441   throws KeeperException {
442     LOG.error(prefix( "Received unexpected KeeperException, re-throwing exception"), ke);
443     throw ke;
444   }
445
446   /**
447    * Handles InterruptedExceptions in client calls.
448    * <p>
449    * This may be temporary but for now this gives one place to deal with these.
450    * <p>
451    * TODO: Currently, this method does nothing.
452    *       Is this ever expected to happen?  Do we abort or can we let it run?
453    *       Maybe this should be logged as WARN?  It shouldn't happen?
454    * <p>
455    * @param ie
456    */
457   public void interruptedException(InterruptedException ie) {
458     LOG.debug(prefix( "Received InterruptedException, doing nothing here"), ie);
459     // At least preserver interrupt.
460     Thread.currentThread().interrupt();
461     // no-op
462   }
463
464   /**
465    * Close the connection to ZooKeeper.
466    *
467    * @throws InterruptedException
468    */
469   public void close() {
470     try {
471       if (recoverableZooKeeper != null) {
472         recoverableZooKeeper.close();
473       }
474     } catch (InterruptedException e) {
475       Thread.currentThread().interrupt();
476     }
477   }
478
479   public Configuration getConfiguration() {
480     return conf;
481   }
482
483   @Override
484   public void abort(String why, Throwable e) {
485     if ( this.abortable != null) this.abortable.abort(why, e);
486     else this.aborted = true;
487   }
488
489   @Override
490   public boolean isAborted() {
491     return this.abortable == null? this.aborted: this.abortable.isAborted();
492   }
493
494   /**
495    * @return Path to the currently active master.
496    */
497   public String getMasterAddressZNode() {
498     return this.masterAddressZNode;
499   }
500
501 }

---

This page was automatically generated by Maven