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.zookeeper;
20  
21  import java.io.BufferedReader;
22  import java.io.IOException;
23  import java.io.InputStreamReader;
24  import java.io.PrintWriter;
25  import java.net.InetSocketAddress;
26  import java.net.Socket;
27  import java.util.ArrayList;
28  import java.util.Arrays;
29  import java.util.HashMap;
30  import java.util.LinkedList;
31  import java.util.List;
32  import java.util.Map;
33  import java.util.Properties;
34  
35  import javax.security.auth.login.AppConfigurationEntry;
36  import javax.security.auth.login.AppConfigurationEntry.LoginModuleControlFlag;
37  
38  import com.google.protobuf.HBaseZeroCopyByteString;
39  import org.apache.commons.lang.StringUtils;
40  import org.apache.commons.logging.Log;
41  import org.apache.commons.logging.LogFactory;
42  import org.apache.hadoop.classification.InterfaceAudience;
43  import org.apache.hadoop.conf.Configuration;
44  import org.apache.hadoop.hbase.HConstants;
45  import org.apache.hadoop.hbase.ServerName;
46  import org.apache.hadoop.hbase.exceptions.DeserializationException;
47  import org.apache.hadoop.hbase.protobuf.ProtobufUtil;
48  import org.apache.hadoop.hbase.protobuf.generated.ZooKeeperProtos;
49  import org.apache.hadoop.hbase.protobuf.generated.ZooKeeperProtos.RegionStoreSequenceIds;
50  import org.apache.hadoop.hbase.util.Bytes;
51  import org.apache.hadoop.hbase.util.ExceptionUtil;
52  import org.apache.hadoop.hbase.util.Threads;
53  import org.apache.hadoop.hbase.zookeeper.ZKUtil.ZKUtilOp.CreateAndFailSilent;
54  import org.apache.hadoop.hbase.zookeeper.ZKUtil.ZKUtilOp.DeleteNodeFailSilent;
55  import org.apache.hadoop.hbase.zookeeper.ZKUtil.ZKUtilOp.SetData;
56  import org.apache.hadoop.security.SecurityUtil;
57  import org.apache.hadoop.security.authentication.util.KerberosUtil;
58  import org.apache.zookeeper.AsyncCallback;
59  import org.apache.zookeeper.CreateMode;
60  import org.apache.zookeeper.KeeperException;
61  import org.apache.zookeeper.KeeperException.NoNodeException;
62  import org.apache.zookeeper.Op;
63  import org.apache.zookeeper.Watcher;
64  import org.apache.zookeeper.ZooDefs.Ids;
65  import org.apache.zookeeper.ZooKeeper;
66  import org.apache.zookeeper.client.ZooKeeperSaslClient;
67  import org.apache.zookeeper.data.ACL;
68  import org.apache.zookeeper.data.Stat;
69  import org.apache.zookeeper.proto.CreateRequest;
70  import org.apache.zookeeper.proto.DeleteRequest;
71  import org.apache.zookeeper.proto.SetDataRequest;
72  import org.apache.zookeeper.server.ZooKeeperSaslServer;
73  
74  import com.google.protobuf.InvalidProtocolBufferException;
75  
76  /**
77   * Internal HBase utility class for ZooKeeper.
78   *
79   * <p>Contains only static methods and constants.
80   *
81   * <p>Methods all throw {@link KeeperException} if there is an unexpected
82   * zookeeper exception, so callers of these methods must handle appropriately.
83   * If ZK is required for the operation, the server will need to be aborted.
84   */
85  @InterfaceAudience.Private
86  public class ZKUtil {
87    private static final Log LOG = LogFactory.getLog(ZKUtil.class);
88  
89    // TODO: Replace this with ZooKeeper constant when ZOOKEEPER-277 is resolved.
90    public static final char ZNODE_PATH_SEPARATOR = '/';
91    private static int zkDumpConnectionTimeOut;
92  
93    /**
94     * Creates a new connection to ZooKeeper, pulling settings and ensemble config
95     * from the specified configuration object using methods from {@link ZKConfig}.
96     *
97     * Sets the connection status monitoring watcher to the specified watcher.
98     *
99     * @param conf configuration to pull ensemble and other settings from
100    * @param watcher watcher to monitor connection changes
101    * @return connection to zookeeper
102    * @throws IOException if unable to connect to zk or config problem
103    */
104   public static RecoverableZooKeeper connect(Configuration conf, Watcher watcher)
105   throws IOException {
106     Properties properties = ZKConfig.makeZKProps(conf);
107     String ensemble = ZKConfig.getZKQuorumServersString(properties);
108     return connect(conf, ensemble, watcher);
109   }
110 
111   public static RecoverableZooKeeper connect(Configuration conf, String ensemble,
112       Watcher watcher)
113   throws IOException {
114     return connect(conf, ensemble, watcher, null);
115   }
116 
117   public static RecoverableZooKeeper connect(Configuration conf, String ensemble,
118       Watcher watcher, final String identifier)
119   throws IOException {
120     if(ensemble == null) {
121       throw new IOException("Unable to determine ZooKeeper ensemble");
122     }
123     int timeout = conf.getInt(HConstants.ZK_SESSION_TIMEOUT,
124         HConstants.DEFAULT_ZK_SESSION_TIMEOUT);
125     if (LOG.isTraceEnabled()) {
126       LOG.trace(identifier + " opening connection to ZooKeeper ensemble=" + ensemble);
127     }
128     int retry = conf.getInt("zookeeper.recovery.retry", 3);
129     int retryIntervalMillis =
130       conf.getInt("zookeeper.recovery.retry.intervalmill", 1000);
131     zkDumpConnectionTimeOut = conf.getInt("zookeeper.dump.connection.timeout",
132         1000);
133     return new RecoverableZooKeeper(ensemble, timeout, watcher,
134         retry, retryIntervalMillis, identifier);
135   }
136 
137   /**
138    * Log in the current zookeeper server process using the given configuration
139    * keys for the credential file and login principal.
140    *
141    * <p><strong>This is only applicable when running on secure hbase</strong>
142    * On regular HBase (without security features), this will safely be ignored.
143    * </p>
144    *
145    * @param conf The configuration data to use
146    * @param keytabFileKey Property key used to configure the path to the credential file
147    * @param userNameKey Property key used to configure the login principal
148    * @param hostname Current hostname to use in any credentials
149    * @throws IOException underlying exception from SecurityUtil.login() call
150    */
151   public static void loginServer(Configuration conf, String keytabFileKey,
152       String userNameKey, String hostname) throws IOException {
153     login(conf, keytabFileKey, userNameKey, hostname,
154           ZooKeeperSaslServer.LOGIN_CONTEXT_NAME_KEY,
155           JaasConfiguration.SERVER_KEYTAB_KERBEROS_CONFIG_NAME);
156   }
157 
158   /**
159    * Log in the current zookeeper client using the given configuration
160    * keys for the credential file and login principal.
161    *
162    * <p><strong>This is only applicable when running on secure hbase</strong>
163    * On regular HBase (without security features), this will safely be ignored.
164    * </p>
165    *
166    * @param conf The configuration data to use
167    * @param keytabFileKey Property key used to configure the path to the credential file
168    * @param userNameKey Property key used to configure the login principal
169    * @param hostname Current hostname to use in any credentials
170    * @throws IOException underlying exception from SecurityUtil.login() call
171    */
172   public static void loginClient(Configuration conf, String keytabFileKey,
173       String userNameKey, String hostname) throws IOException {
174     login(conf, keytabFileKey, userNameKey, hostname,
175           ZooKeeperSaslClient.LOGIN_CONTEXT_NAME_KEY,
176           JaasConfiguration.CLIENT_KEYTAB_KERBEROS_CONFIG_NAME);
177   }
178 
179   /**
180    * Log in the current process using the given configuration keys for the
181    * credential file and login principal.
182    *
183    * <p><strong>This is only applicable when running on secure hbase</strong>
184    * On regular HBase (without security features), this will safely be ignored.
185    * </p>
186    *
187    * @param conf The configuration data to use
188    * @param keytabFileKey Property key used to configure the path to the credential file
189    * @param userNameKey Property key used to configure the login principal
190    * @param hostname Current hostname to use in any credentials
191    * @param loginContextProperty property name to expose the entry name
192    * @param loginContextName jaas entry name
193    * @throws IOException underlying exception from SecurityUtil.login() call
194    */
195   private static void login(Configuration conf, String keytabFileKey,
196       String userNameKey, String hostname,
197       String loginContextProperty, String loginContextName)
198       throws IOException {
199     if (!isSecureZooKeeper(conf))
200       return;
201 
202     // User has specified a jaas.conf, keep this one as the good one.
203     // HBASE_OPTS="-Djava.security.auth.login.config=jaas.conf"
204     if (System.getProperty("java.security.auth.login.config") != null)
205       return;
206 
207     // No keytab specified, no auth
208     String keytabFilename = conf.get(keytabFileKey);
209     if (keytabFilename == null) {
210       LOG.warn("no keytab specified for: " + keytabFileKey);
211       return;
212     }
213 
214     String principalConfig = conf.get(userNameKey, System.getProperty("user.name"));
215     String principalName = SecurityUtil.getServerPrincipal(principalConfig, hostname);
216 
217     // Initialize the "jaas.conf" for keyTab/principal,
218     // If keyTab is not specified use the Ticket Cache.
219     // and set the zookeeper login context name.
220     JaasConfiguration jaasConf = new JaasConfiguration(loginContextName,
221         principalName, keytabFilename);
222     javax.security.auth.login.Configuration.setConfiguration(jaasConf);
223     System.setProperty(loginContextProperty, loginContextName);
224   }
225 
226   /**
227    * A JAAS configuration that defines the login modules that we want to use for login.
228    */
229   private static class JaasConfiguration extends javax.security.auth.login.Configuration {
230     private static final String SERVER_KEYTAB_KERBEROS_CONFIG_NAME =
231       "zookeeper-server-keytab-kerberos";
232     private static final String CLIENT_KEYTAB_KERBEROS_CONFIG_NAME =
233       "zookeeper-client-keytab-kerberos";
234 
235     private static final Map<String, String> BASIC_JAAS_OPTIONS =
236       new HashMap<String,String>();
237     static {
238       String jaasEnvVar = System.getenv("HBASE_JAAS_DEBUG");
239       if (jaasEnvVar != null && "true".equalsIgnoreCase(jaasEnvVar)) {
240         BASIC_JAAS_OPTIONS.put("debug", "true");
241       }
242     }
243 
244     private static final Map<String,String> KEYTAB_KERBEROS_OPTIONS =
245       new HashMap<String,String>();
246     static {
247       KEYTAB_KERBEROS_OPTIONS.put("doNotPrompt", "true");
248       KEYTAB_KERBEROS_OPTIONS.put("storeKey", "true");
249       KEYTAB_KERBEROS_OPTIONS.put("refreshKrb5Config", "true");
250       KEYTAB_KERBEROS_OPTIONS.putAll(BASIC_JAAS_OPTIONS);
251     }
252 
253     private static final AppConfigurationEntry KEYTAB_KERBEROS_LOGIN =
254       new AppConfigurationEntry(KerberosUtil.getKrb5LoginModuleName(),
255                                 LoginModuleControlFlag.REQUIRED,
256                                 KEYTAB_KERBEROS_OPTIONS);
257 
258     private static final AppConfigurationEntry[] KEYTAB_KERBEROS_CONF =
259       new AppConfigurationEntry[]{KEYTAB_KERBEROS_LOGIN};
260 
261     private javax.security.auth.login.Configuration baseConfig;
262     private final String loginContextName;
263     private final boolean useTicketCache;
264     private final String keytabFile;
265     private final String principal;
266 
267     public JaasConfiguration(String loginContextName, String principal) {
268       this(loginContextName, principal, null, true);
269     }
270 
271     public JaasConfiguration(String loginContextName, String principal, String keytabFile) {
272       this(loginContextName, principal, keytabFile, keytabFile == null || keytabFile.length() == 0);
273     }
274 
275     private JaasConfiguration(String loginContextName, String principal,
276                              String keytabFile, boolean useTicketCache) {
277       try {
278         this.baseConfig = javax.security.auth.login.Configuration.getConfiguration();
279       } catch (SecurityException e) {
280         this.baseConfig = null;
281       }
282       this.loginContextName = loginContextName;
283       this.useTicketCache = useTicketCache;
284       this.keytabFile = keytabFile;
285       this.principal = principal;
286       LOG.info("JaasConfiguration loginContextName=" + loginContextName +
287                " principal=" + principal + " useTicketCache=" + useTicketCache +
288                " keytabFile=" + keytabFile);
289     }
290 
291     @Override
292     public AppConfigurationEntry[] getAppConfigurationEntry(String appName) {
293       if (loginContextName.equals(appName)) {
294         if (!useTicketCache) {
295           KEYTAB_KERBEROS_OPTIONS.put("keyTab", keytabFile);
296           KEYTAB_KERBEROS_OPTIONS.put("useKeyTab", "true");
297         }
298         KEYTAB_KERBEROS_OPTIONS.put("principal", principal);
299         KEYTAB_KERBEROS_OPTIONS.put("useTicketCache", useTicketCache ? "true" : "false");
300         return KEYTAB_KERBEROS_CONF;
301       }
302       if (baseConfig != null) return baseConfig.getAppConfigurationEntry(appName);
303       return(null);
304     }
305   }
306 
307   //
308   // Helper methods
309   //
310 
311   /**
312    * Join the prefix znode name with the suffix znode name to generate a proper
313    * full znode name.
314    *
315    * Assumes prefix does not end with slash and suffix does not begin with it.
316    *
317    * @param prefix beginning of znode name
318    * @param suffix ending of znode name
319    * @return result of properly joining prefix with suffix
320    */
321   public static String joinZNode(String prefix, String suffix) {
322     return prefix + ZNODE_PATH_SEPARATOR + suffix;
323   }
324 
325   /**
326    * Returns the full path of the immediate parent of the specified node.
327    * @param node path to get parent of
328    * @return parent of path, null if passed the root node or an invalid node
329    */
330   public static String getParent(String node) {
331     int idx = node.lastIndexOf(ZNODE_PATH_SEPARATOR);
332     return idx <= 0 ? null : node.substring(0, idx);
333   }
334 
335   /**
336    * Get the name of the current node from the specified fully-qualified path.
337    * @param path fully-qualified path
338    * @return name of the current node
339    */
340   public static String getNodeName(String path) {
341     return path.substring(path.lastIndexOf("/")+1);
342   }
343 
344   /**
345    * Get the key to the ZK ensemble for this configuration without
346    * adding a name at the end
347    * @param conf Configuration to use to build the key
348    * @return ensemble key without a name
349    */
350   public static String getZooKeeperClusterKey(Configuration conf) {
351     return getZooKeeperClusterKey(conf, null);
352   }
353 
354   /**
355    * Get the key to the ZK ensemble for this configuration and append
356    * a name at the end
357    * @param conf Configuration to use to build the key
358    * @param name Name that should be appended at the end if not empty or null
359    * @return ensemble key with a name (if any)
360    */
361   public static String getZooKeeperClusterKey(Configuration conf, String name) {
362     String ensemble = conf.get(HConstants.ZOOKEEPER_QUORUM.replaceAll(
363         "[\\t\\n\\x0B\\f\\r]", ""));
364     StringBuilder builder = new StringBuilder(ensemble);
365     builder.append(":");
366     builder.append(conf.get(HConstants.ZOOKEEPER_CLIENT_PORT));
367     builder.append(":");
368     builder.append(conf.get(HConstants.ZOOKEEPER_ZNODE_PARENT));
369     if (name != null && !name.isEmpty()) {
370       builder.append(",");
371       builder.append(name);
372     }
373     return builder.toString();
374   }
375 
376   /**
377    * Apply the settings in the given key to the given configuration, this is
378    * used to communicate with distant clusters
379    * @param conf configuration object to configure
380    * @param key string that contains the 3 required configuratins
381    * @throws IOException
382    */
383   public static void applyClusterKeyToConf(Configuration conf, String key)
384       throws IOException{
385     String[] parts = transformClusterKey(key);
386     conf.set(HConstants.ZOOKEEPER_QUORUM, parts[0]);
387     conf.set(HConstants.ZOOKEEPER_CLIENT_PORT, parts[1]);
388     conf.set(HConstants.ZOOKEEPER_ZNODE_PARENT, parts[2]);
389   }
390 
391   /**
392    * Separate the given key into the three configurations it should contain:
393    * hbase.zookeeper.quorum, hbase.zookeeper.client.port
394    * and zookeeper.znode.parent
395    * @param key
396    * @return the three configuration in the described order
397    * @throws IOException
398    */
399   public static String[] transformClusterKey(String key) throws IOException {
400     String[] parts = key.split(":");
401     if (parts.length != 3) {
402       throw new IOException("Cluster key passed " + key + " is invalid, the format should be:" +
403           HConstants.ZOOKEEPER_QUORUM + ":hbase.zookeeper.client.port:"
404           + HConstants.ZOOKEEPER_ZNODE_PARENT);
405     }
406     return parts;
407   }
408 
409   //
410   // Existence checks and watches
411   //
412 
413   /**
414    * Watch the specified znode for delete/create/change events.  The watcher is
415    * set whether or not the node exists.  If the node already exists, the method
416    * returns true.  If the node does not exist, the method returns false.
417    *
418    * @param zkw zk reference
419    * @param znode path of node to watch
420    * @return true if znode exists, false if does not exist or error
421    * @throws KeeperException if unexpected zookeeper exception
422    */
423   public static boolean watchAndCheckExists(ZooKeeperWatcher zkw, String znode)
424   throws KeeperException {
425     try {
426       Stat s = zkw.getRecoverableZooKeeper().exists(znode, zkw);
427       boolean exists = s != null ? true : false;
428       if (exists) {
429         LOG.debug(zkw.prefix("Set watcher on existing znode=" + znode));
430       } else {
431         LOG.debug(zkw.prefix("Set watcher on znode that does not yet exist, " + znode));
432       }
433       return exists;
434     } catch (KeeperException e) {
435       LOG.warn(zkw.prefix("Unable to set watcher on znode " + znode), e);
436       zkw.keeperException(e);
437       return false;
438     } catch (InterruptedException e) {
439       LOG.warn(zkw.prefix("Unable to set watcher on znode " + znode), e);
440       zkw.interruptedException(e);
441       return false;
442     }
443   }
444 
445   /**
446    * Watch the specified znode, but only if exists. Useful when watching
447    * for deletions. Uses .getData() (and handles NoNodeException) instead
448    * of .exists() to accomplish this, as .getData() will only set a watch if
449    * the znode exists.
450    * @param zkw zk reference
451    * @param znode path of node to watch
452    * @return true if the watch is set, false if node does not exists
453    * @throws KeeperException if unexpected zookeeper exception
454    */
455   public static boolean setWatchIfNodeExists(ZooKeeperWatcher zkw, String znode)
456       throws KeeperException {
457     try {
458       zkw.getRecoverableZooKeeper().getData(znode, true, null);
459       return true;
460     } catch (NoNodeException e) {
461       return false;
462     } catch (InterruptedException e) {
463       LOG.warn(zkw.prefix("Unable to set watcher on znode " + znode), e);
464       zkw.interruptedException(e);
465       return false;
466     }
467   }
468 
469   /**
470    * Check if the specified node exists.  Sets no watches.
471    *
472    * @param zkw zk reference
473    * @param znode path of node to watch
474    * @return version of the node if it exists, -1 if does not exist
475    * @throws KeeperException if unexpected zookeeper exception
476    */
477   public static int checkExists(ZooKeeperWatcher zkw, String znode)
478   throws KeeperException {
479     try {
480       Stat s = zkw.getRecoverableZooKeeper().exists(znode, null);
481       return s != null ? s.getVersion() : -1;
482     } catch (KeeperException e) {
483       LOG.warn(zkw.prefix("Unable to set watcher on znode (" + znode + ")"), e);
484       zkw.keeperException(e);
485       return -1;
486     } catch (InterruptedException e) {
487       LOG.warn(zkw.prefix("Unable to set watcher on znode (" + znode + ")"), e);
488       zkw.interruptedException(e);
489       return -1;
490     }
491   }
492 
493   //
494   // Znode listings
495   //
496 
497   /**
498    * Lists the children znodes of the specified znode.  Also sets a watch on
499    * the specified znode which will capture a NodeDeleted event on the specified
500    * znode as well as NodeChildrenChanged if any children of the specified znode
501    * are created or deleted.
502    *
503    * Returns null if the specified node does not exist.  Otherwise returns a
504    * list of children of the specified node.  If the node exists but it has no
505    * children, an empty list will be returned.
506    *
507    * @param zkw zk reference
508    * @param znode path of node to list and watch children of
509    * @return list of children of the specified node, an empty list if the node
510    *          exists but has no children, and null if the node does not exist
511    * @throws KeeperException if unexpected zookeeper exception
512    */
513   public static List<String> listChildrenAndWatchForNewChildren(
514       ZooKeeperWatcher zkw, String znode)
515   throws KeeperException {
516     try {
517       List<String> children = zkw.getRecoverableZooKeeper().getChildren(znode, zkw);
518       return children;
519     } catch(KeeperException.NoNodeException ke) {
520       LOG.debug(zkw.prefix("Unable to list children of znode " + znode + " " +
521           "because node does not exist (not an error)"));
522       return null;
523     } catch (KeeperException e) {
524       LOG.warn(zkw.prefix("Unable to list children of znode " + znode + " "), e);
525       zkw.keeperException(e);
526       return null;
527     } catch (InterruptedException e) {
528       LOG.warn(zkw.prefix("Unable to list children of znode " + znode + " "), e);
529       zkw.interruptedException(e);
530       return null;
531     }
532   }
533 
534   /**
535    * List all the children of the specified znode, setting a watch for children
536    * changes and also setting a watch on every individual child in order to get
537    * the NodeCreated and NodeDeleted events.
538    * @param zkw zookeeper reference
539    * @param znode node to get children of and watch
540    * @return list of znode names, null if the node doesn't exist
541    * @throws KeeperException
542    */
543   public static List<String> listChildrenAndWatchThem(ZooKeeperWatcher zkw,
544       String znode) throws KeeperException {
545     List<String> children = listChildrenAndWatchForNewChildren(zkw, znode);
546     if (children == null) {
547       return null;
548     }
549     for (String child : children) {
550       watchAndCheckExists(zkw, joinZNode(znode, child));
551     }
552     return children;
553   }
554 
555   /**
556    * Lists the children of the specified znode without setting any watches.
557    *
558    * Sets no watches at all, this method is best effort.
559    *
560    * Returns an empty list if the node has no children.  Returns null if the
561    * parent node itself does not exist.
562    *
563    * @param zkw zookeeper reference
564    * @param znode node to get children
565    * @return list of data of children of specified znode, empty if no children,
566    *         null if parent does not exist
567    * @throws KeeperException if unexpected zookeeper exception
568    */
569   public static List<String> listChildrenNoWatch(ZooKeeperWatcher zkw, String znode)
570   throws KeeperException {
571     List<String> children = null;
572     try {
573       // List the children without watching
574       children = zkw.getRecoverableZooKeeper().getChildren(znode, null);
575     } catch(KeeperException.NoNodeException nne) {
576       return null;
577     } catch(InterruptedException ie) {
578       zkw.interruptedException(ie);
579     }
580     return children;
581   }
582 
583   /**
584    * Simple class to hold a node path and node data.
585    * @deprecated Unused
586    */
587   @Deprecated
588   public static class NodeAndData {
589     private String node;
590     private byte [] data;
591     public NodeAndData(String node, byte [] data) {
592       this.node = node;
593       this.data = data;
594     }
595     public String getNode() {
596       return node;
597     }
598     public byte [] getData() {
599       return data;
600     }
601     @Override
602     public String toString() {
603       return node;
604     }
605     public boolean isEmpty() {
606       return (data.length == 0);
607     }
608   }
609 
610   /**
611    * Checks if the specified znode has any children.  Sets no watches.
612    *
613    * Returns true if the node exists and has children.  Returns false if the
614    * node does not exist or if the node does not have any children.
615    *
616    * Used during master initialization to determine if the master is a
617    * failed-over-to master or the first master during initial cluster startup.
618    * If the directory for regionserver ephemeral nodes is empty then this is
619    * a cluster startup, if not then it is not cluster startup.
620    *
621    * @param zkw zk reference
622    * @param znode path of node to check for children of
623    * @return true if node has children, false if not or node does not exist
624    * @throws KeeperException if unexpected zookeeper exception
625    */
626   public static boolean nodeHasChildren(ZooKeeperWatcher zkw, String znode)
627   throws KeeperException {
628     try {
629       return !zkw.getRecoverableZooKeeper().getChildren(znode, null).isEmpty();
630     } catch(KeeperException.NoNodeException ke) {
631       LOG.debug(zkw.prefix("Unable to list children of znode " + znode + " " +
632       "because node does not exist (not an error)"));
633       return false;
634     } catch (KeeperException e) {
635       LOG.warn(zkw.prefix("Unable to list children of znode " + znode), e);
636       zkw.keeperException(e);
637       return false;
638     } catch (InterruptedException e) {
639       LOG.warn(zkw.prefix("Unable to list children of znode " + znode), e);
640       zkw.interruptedException(e);
641       return false;
642     }
643   }
644 
645   /**
646    * Get the number of children of the specified node.
647    *
648    * If the node does not exist or has no children, returns 0.
649    *
650    * Sets no watches at all.
651    *
652    * @param zkw zk reference
653    * @param znode path of node to count children of
654    * @return number of children of specified node, 0 if none or parent does not
655    *         exist
656    * @throws KeeperException if unexpected zookeeper exception
657    */
658   public static int getNumberOfChildren(ZooKeeperWatcher zkw, String znode)
659   throws KeeperException {
660     try {
661       Stat stat = zkw.getRecoverableZooKeeper().exists(znode, null);
662       return stat == null ? 0 : stat.getNumChildren();
663     } catch(KeeperException e) {
664       LOG.warn(zkw.prefix("Unable to get children of node " + znode));
665       zkw.keeperException(e);
666     } catch(InterruptedException e) {
667       zkw.interruptedException(e);
668     }
669     return 0;
670   }
671 
672   //
673   // Data retrieval
674   //
675 
676   /**
677    * Get znode data. Does not set a watcher.
678    * @return ZNode data, null if the node does not exist or if there is an
679    *  error.
680    */
681   public static byte [] getData(ZooKeeperWatcher zkw, String znode)
682       throws KeeperException, InterruptedException {
683     try {
684       byte [] data = zkw.getRecoverableZooKeeper().getData(znode, null, null);
685       logRetrievedMsg(zkw, znode, data, false);
686       return data;
687     } catch (KeeperException.NoNodeException e) {
688       LOG.debug(zkw.prefix("Unable to get data of znode " + znode + " " +
689           "because node does not exist (not an error)"));
690       return null;
691     } catch (KeeperException e) {
692       LOG.warn(zkw.prefix("Unable to get data of znode " + znode), e);
693       zkw.keeperException(e);
694       return null;
695     }
696   }
697 
698   /**
699    * Get the data at the specified znode and set a watch.
700    *
701    * Returns the data and sets a watch if the node exists.  Returns null and no
702    * watch is set if the node does not exist or there is an exception.
703    *
704    * @param zkw zk reference
705    * @param znode path of node
706    * @return data of the specified znode, or null
707    * @throws KeeperException if unexpected zookeeper exception
708    */
709   public static byte [] getDataAndWatch(ZooKeeperWatcher zkw, String znode)
710   throws KeeperException {
711     return getDataInternal(zkw, znode, null, true);
712   }
713 
714   /**
715    * Get the data at the specified znode and set a watch.
716    *
717    * Returns the data and sets a watch if the node exists.  Returns null and no
718    * watch is set if the node does not exist or there is an exception.
719    *
720    * @param zkw zk reference
721    * @param znode path of node
722    * @param stat object to populate the version of the znode
723    * @return data of the specified znode, or null
724    * @throws KeeperException if unexpected zookeeper exception
725    */
726   public static byte[] getDataAndWatch(ZooKeeperWatcher zkw, String znode,
727       Stat stat) throws KeeperException {
728     return getDataInternal(zkw, znode, stat, true);
729   }
730 
731   private static byte[] getDataInternal(ZooKeeperWatcher zkw, String znode, Stat stat,
732       boolean watcherSet)
733       throws KeeperException {
734     try {
735       byte [] data = zkw.getRecoverableZooKeeper().getData(znode, zkw, stat);
736       logRetrievedMsg(zkw, znode, data, watcherSet);
737       return data;
738     } catch (KeeperException.NoNodeException e) {
739       // This log can get pretty annoying when we cycle on 100ms waits.
740       // Enable trace if you really want to see it.
741       LOG.trace(zkw.prefix("Unable to get data of znode " + znode + " " +
742         "because node does not exist (not an error)"));
743       return null;
744     } catch (KeeperException e) {
745       LOG.warn(zkw.prefix("Unable to get data of znode " + znode), e);
746       zkw.keeperException(e);
747       return null;
748     } catch (InterruptedException e) {
749       LOG.warn(zkw.prefix("Unable to get data of znode " + znode), e);
750       zkw.interruptedException(e);
751       return null;
752     }
753   }
754 
755   /**
756    * Get the data at the specified znode without setting a watch.
757    *
758    * Returns the data if the node exists.  Returns null if the node does not
759    * exist.
760    *
761    * Sets the stats of the node in the passed Stat object.  Pass a null stat if
762    * not interested.
763    *
764    * @param zkw zk reference
765    * @param znode path of node
766    * @param stat node status to get if node exists
767    * @return data of the specified znode, or null if node does not exist
768    * @throws KeeperException if unexpected zookeeper exception
769    */
770   public static byte [] getDataNoWatch(ZooKeeperWatcher zkw, String znode,
771       Stat stat)
772   throws KeeperException {
773     try {
774       byte [] data = zkw.getRecoverableZooKeeper().getData(znode, null, stat);
775       logRetrievedMsg(zkw, znode, data, false);
776       return data;
777     } catch (KeeperException.NoNodeException e) {
778       LOG.debug(zkw.prefix("Unable to get data of znode " + znode + " " +
779           "because node does not exist (not necessarily an error)"));
780       return null;
781     } catch (KeeperException e) {
782       LOG.warn(zkw.prefix("Unable to get data of znode " + znode), e);
783       zkw.keeperException(e);
784       return null;
785     } catch (InterruptedException e) {
786       LOG.warn(zkw.prefix("Unable to get data of znode " + znode), e);
787       zkw.interruptedException(e);
788       return null;
789     }
790   }
791 
792   /**
793    * Returns the date of child znodes of the specified znode.  Also sets a watch on
794    * the specified znode which will capture a NodeDeleted event on the specified
795    * znode as well as NodeChildrenChanged if any children of the specified znode
796    * are created or deleted.
797    *
798    * Returns null if the specified node does not exist.  Otherwise returns a
799    * list of children of the specified node.  If the node exists but it has no
800    * children, an empty list will be returned.
801    *
802    * @param zkw zk reference
803    * @param baseNode path of node to list and watch children of
804    * @return list of data of children of the specified node, an empty list if the node
805    *          exists but has no children, and null if the node does not exist
806    * @throws KeeperException if unexpected zookeeper exception
807    * @deprecated Unused
808    */
809   public static List<NodeAndData> getChildDataAndWatchForNewChildren(
810       ZooKeeperWatcher zkw, String baseNode) throws KeeperException {
811     List<String> nodes =
812       ZKUtil.listChildrenAndWatchForNewChildren(zkw, baseNode);
813     List<NodeAndData> newNodes = new ArrayList<NodeAndData>();
814     if (nodes != null) {
815       for (String node : nodes) {
816         String nodePath = ZKUtil.joinZNode(baseNode, node);
817         byte[] data = ZKUtil.getDataAndWatch(zkw, nodePath);
818         newNodes.add(new NodeAndData(nodePath, data));
819       }
820     }
821     return newNodes;
822   }
823 
824   /**
825    * Update the data of an existing node with the expected version to have the
826    * specified data.
827    *
828    * Throws an exception if there is a version mismatch or some other problem.
829    *
830    * Sets no watches under any conditions.
831    *
832    * @param zkw zk reference
833    * @param znode
834    * @param data
835    * @param expectedVersion
836    * @throws KeeperException if unexpected zookeeper exception
837    * @throws KeeperException.BadVersionException if version mismatch
838    * @deprecated Unused
839    */
840   public static void updateExistingNodeData(ZooKeeperWatcher zkw, String znode,
841       byte [] data, int expectedVersion)
842   throws KeeperException {
843     try {
844       zkw.getRecoverableZooKeeper().setData(znode, data, expectedVersion);
845     } catch(InterruptedException ie) {
846       zkw.interruptedException(ie);
847     }
848   }
849 
850   //
851   // Data setting
852   //
853 
854   /**
855    * Sets the data of the existing znode to be the specified data.  Ensures that
856    * the current data has the specified expected version.
857    *
858    * <p>If the node does not exist, a {@link NoNodeException} will be thrown.
859    *
860    * <p>If their is a version mismatch, method returns null.
861    *
862    * <p>No watches are set but setting data will trigger other watchers of this
863    * node.
864    *
865    * <p>If there is another problem, a KeeperException will be thrown.
866    *
867    * @param zkw zk reference
868    * @param znode path of node
869    * @param data data to set for node
870    * @param expectedVersion version expected when setting data
871    * @return true if data set, false if version mismatch
872    * @throws KeeperException if unexpected zookeeper exception
873    */
874   public static boolean setData(ZooKeeperWatcher zkw, String znode,
875       byte [] data, int expectedVersion)
876   throws KeeperException, KeeperException.NoNodeException {
877     try {
878       return zkw.getRecoverableZooKeeper().setData(znode, data, expectedVersion) != null;
879     } catch (InterruptedException e) {
880       zkw.interruptedException(e);
881       return false;
882     }
883   }
884 
885   /**
886    * Set data into node creating node if it doesn't yet exist.
887    * Does not set watch.
888    *
889    * @param zkw zk reference
890    * @param znode path of node
891    * @param data data to set for node
892    * @throws KeeperException
893    */
894   public static void createSetData(final ZooKeeperWatcher zkw, final String znode,
895       final byte [] data)
896   throws KeeperException {
897     if (checkExists(zkw, znode) == -1) {
898       ZKUtil.createWithParents(zkw, znode, data);
899     } else {
900       ZKUtil.setData(zkw, znode, data);
901     }
902   }
903 
904   /**
905    * Sets the data of the existing znode to be the specified data.  The node
906    * must exist but no checks are done on the existing data or version.
907    *
908    * <p>If the node does not exist, a {@link NoNodeException} will be thrown.
909    *
910    * <p>No watches are set but setting data will trigger other watchers of this
911    * node.
912    *
913    * <p>If there is another problem, a KeeperException will be thrown.
914    *
915    * @param zkw zk reference
916    * @param znode path of node
917    * @param data data to set for node
918    * @throws KeeperException if unexpected zookeeper exception
919    */
920   public static void setData(ZooKeeperWatcher zkw, String znode, byte [] data)
921   throws KeeperException, KeeperException.NoNodeException {
922     setData(zkw, (SetData)ZKUtilOp.setData(znode, data));
923   }
924 
925   private static void setData(ZooKeeperWatcher zkw, SetData setData)
926   throws KeeperException, KeeperException.NoNodeException {
927     SetDataRequest sd = (SetDataRequest)toZooKeeperOp(zkw, setData).toRequestRecord();
928     setData(zkw, sd.getPath(), sd.getData(), sd.getVersion());
929   }
930 
931   /**
932    * Returns whether or not secure authentication is enabled
933    * (whether <code>hbase.security.authentication</code> is set to
934    * <code>kerberos</code>.
935    */
936   public static boolean isSecureZooKeeper(Configuration conf) {
937     // Detection for embedded HBase client with jaas configuration
938     // defined for third party programs.
939     try {
940       javax.security.auth.login.Configuration testConfig = javax.security.auth.login.Configuration.getConfiguration();
941       if(testConfig.getAppConfigurationEntry("Client") == null) {
942         return false;
943       }
944     } catch(Exception e) {
945       // No Jaas configuration defined.
946       return false;
947     }
948 
949     // Master & RSs uses hbase.zookeeper.client.*
950     return("kerberos".equalsIgnoreCase(conf.get("hbase.security.authentication")) &&
951          conf.get("hbase.zookeeper.client.keytab.file") != null);
952   }
953 
954   private static ArrayList<ACL> createACL(ZooKeeperWatcher zkw, String node) {
955     if (isSecureZooKeeper(zkw.getConfiguration())) {
956       // Certain znodes are accessed directly by the client,
957       // so they must be readable by non-authenticated clients
958       if ((node.equals(zkw.baseZNode) == true) ||
959           (node.equals(zkw.metaServerZNode) == true) ||
960           (node.equals(zkw.getMasterAddressZNode()) == true) ||
961           (node.equals(zkw.clusterIdZNode) == true) ||
962           (node.equals(zkw.rsZNode) == true) ||
963           (node.equals(zkw.backupMasterAddressesZNode) == true) ||
964           (node.startsWith(zkw.assignmentZNode) == true) ||
965           (node.startsWith(zkw.tableZNode) == true)) {
966         return ZooKeeperWatcher.CREATOR_ALL_AND_WORLD_READABLE;
967       }
968       return Ids.CREATOR_ALL_ACL;
969     } else {
970       return Ids.OPEN_ACL_UNSAFE;
971     }
972   }
973 
974   //
975   // Node creation
976   //
977 
978   /**
979    *
980    * Set the specified znode to be an ephemeral node carrying the specified
981    * data.
982    *
983    * If the node is created successfully, a watcher is also set on the node.
984    *
985    * If the node is not created successfully because it already exists, this
986    * method will also set a watcher on the node.
987    *
988    * If there is another problem, a KeeperException will be thrown.
989    *
990    * @param zkw zk reference
991    * @param znode path of node
992    * @param data data of node
993    * @return true if node created, false if not, watch set in both cases
994    * @throws KeeperException if unexpected zookeeper exception
995    */
996   public static boolean createEphemeralNodeAndWatch(ZooKeeperWatcher zkw,
997       String znode, byte [] data)
998   throws KeeperException {
999     boolean ret = true;
1000     try {
1001       zkw.getRecoverableZooKeeper().create(znode, data, createACL(zkw, znode),
1002           CreateMode.EPHEMERAL);
1003     } catch (KeeperException.NodeExistsException nee) {
1004       ret = false;
1005     } catch (InterruptedException e) {
1006       LOG.info("Interrupted", e);
1007       Thread.currentThread().interrupt();
1008     }
1009     if(!watchAndCheckExists(zkw, znode)) {
1010       // It did exist but now it doesn't, try again
1011       return createEphemeralNodeAndWatch(zkw, znode, data);
1012     }
1013     return ret;
1014   }
1015 
1016   /**
1017    * Creates the specified znode to be a persistent node carrying the specified
1018    * data.
1019    *
1020    * Returns true if the node was successfully created, false if the node
1021    * already existed.
1022    *
1023    * If the node is created successfully, a watcher is also set on the node.
1024    *
1025    * If the node is not created successfully because it already exists, this
1026    * method will also set a watcher on the node but return false.
1027    *
1028    * If there is another problem, a KeeperException will be thrown.
1029    *
1030    * @param zkw zk reference
1031    * @param znode path of node
1032    * @param data data of node
1033    * @return true if node created, false if not, watch set in both cases
1034    * @throws KeeperException if unexpected zookeeper exception
1035    */
1036   public static boolean createNodeIfNotExistsAndWatch(
1037       ZooKeeperWatcher zkw, String znode, byte [] data)
1038   throws KeeperException {
1039     boolean ret = true;
1040     try {
1041       zkw.getRecoverableZooKeeper().create(znode, data, createACL(zkw, znode),
1042           CreateMode.PERSISTENT);
1043     } catch (KeeperException.NodeExistsException nee) {
1044       ret = false;
1045     } catch (InterruptedException e) {
1046       zkw.interruptedException(e);
1047       return false;
1048     }
1049     try {
1050       zkw.getRecoverableZooKeeper().exists(znode, zkw);
1051     } catch (InterruptedException e) {
1052       zkw.interruptedException(e);
1053       return false;
1054     }
1055     return ret;
1056   }
1057 
1058   /**
1059    * Creates the specified znode with the specified data but does not watch it.
1060    *
1061    * Returns the znode of the newly created node
1062    *
1063    * If there is another problem, a KeeperException will be thrown.
1064    *
1065    * @param zkw zk reference
1066    * @param znode path of node
1067    * @param data data of node
1068    * @param createMode specifying whether the node to be created is ephemeral and/or sequential
1069    * @return true name of the newly created znode or null
1070    * @throws KeeperException if unexpected zookeeper exception
1071    */
1072   public static String createNodeIfNotExistsNoWatch(ZooKeeperWatcher zkw, String znode,
1073       byte[] data, CreateMode createMode) throws KeeperException {
1074 
1075     String createdZNode = null;
1076     try {
1077       createdZNode = zkw.getRecoverableZooKeeper().create(znode, data,
1078           createACL(zkw, znode), createMode);
1079     } catch (KeeperException.NodeExistsException nee) {
1080       return znode;
1081     } catch (InterruptedException e) {
1082       zkw.interruptedException(e);
1083       return null;
1084     }
1085     return createdZNode;
1086   }
1087 
1088   /**
1089    * Creates the specified node with the specified data and watches it.
1090    *
1091    * <p>Throws an exception if the node already exists.
1092    *
1093    * <p>The node created is persistent and open access.
1094    *
1095    * <p>Returns the version number of the created node if successful.
1096    *
1097    * @param zkw zk reference
1098    * @param znode path of node to create
1099    * @param data data of node to create
1100    * @return version of node created
1101    * @throws KeeperException if unexpected zookeeper exception
1102    * @throws KeeperException.NodeExistsException if node already exists
1103    */
1104   public static int createAndWatch(ZooKeeperWatcher zkw,
1105       String znode, byte [] data)
1106   throws KeeperException, KeeperException.NodeExistsException {
1107     try {
1108       zkw.getRecoverableZooKeeper().create(znode, data, createACL(zkw, znode),
1109           CreateMode.PERSISTENT);
1110       Stat stat = zkw.getRecoverableZooKeeper().exists(znode, zkw);
1111       if (stat == null){
1112         // Likely a race condition. Someone deleted the znode.
1113         throw KeeperException.create(KeeperException.Code.SYSTEMERROR,
1114             "ZK.exists returned null (i.e.: znode does not exist) for znode=" + znode);
1115       }
1116      return stat.getVersion();
1117     } catch (InterruptedException e) {
1118       zkw.interruptedException(e);
1119       return -1;
1120     }
1121   }
1122 
1123   /**
1124    * Async creates the specified node with the specified data.
1125    *
1126    * <p>Throws an exception if the node already exists.
1127    *
1128    * <p>The node created is persistent and open access.
1129    *
1130    * @param zkw zk reference
1131    * @param znode path of node to create
1132    * @param data data of node to create
1133    * @param cb
1134    * @param ctx
1135    * @throws KeeperException if unexpected zookeeper exception
1136    * @throws KeeperException.NodeExistsException if node already exists
1137    */
1138   public static void asyncCreate(ZooKeeperWatcher zkw,
1139       String znode, byte [] data, final AsyncCallback.StringCallback cb,
1140       final Object ctx) {
1141     zkw.getRecoverableZooKeeper().getZooKeeper().create(znode, data,
1142         createACL(zkw, znode), CreateMode.PERSISTENT, cb, ctx);
1143   }
1144 
1145   /**
1146    * Creates the specified node, iff the node does not exist.  Does not set a
1147    * watch and fails silently if the node already exists.
1148    *
1149    * The node created is persistent and open access.
1150    *
1151    * @param zkw zk reference
1152    * @param znode path of node
1153    * @throws KeeperException if unexpected zookeeper exception
1154    */
1155   public static void createAndFailSilent(ZooKeeperWatcher zkw,
1156       String znode) throws KeeperException {
1157     createAndFailSilent(zkw, znode, new byte[0]);
1158   }
1159 
1160   /**
1161    * Creates the specified node containing specified data, iff the node does not exist.  Does
1162    * not set a watch and fails silently if the node already exists.
1163    *
1164    * The node created is persistent and open access.
1165    *
1166    * @param zkw zk reference
1167    * @param znode path of node
1168    * @param data a byte array data to store in the znode
1169    * @throws KeeperException if unexpected zookeeper exception
1170    */
1171   public static void createAndFailSilent(ZooKeeperWatcher zkw,
1172       String znode, byte[] data)
1173   throws KeeperException {
1174     createAndFailSilent(zkw,
1175         (CreateAndFailSilent)ZKUtilOp.createAndFailSilent(znode, data));
1176   }
1177   
1178   private static void createAndFailSilent(ZooKeeperWatcher zkw, CreateAndFailSilent cafs)
1179   throws KeeperException {
1180     CreateRequest create = (CreateRequest)toZooKeeperOp(zkw, cafs).toRequestRecord();
1181     String znode = create.getPath();
1182     try {
1183       RecoverableZooKeeper zk = zkw.getRecoverableZooKeeper();
1184       if (zk.exists(znode, false) == null) {
1185         zk.create(znode, create.getData(), create.getAcl(), CreateMode.fromFlag(create.getFlags()));
1186       }
1187     } catch(KeeperException.NodeExistsException nee) {
1188     } catch(KeeperException.NoAuthException nee){
1189       try {
1190         if (null == zkw.getRecoverableZooKeeper().exists(znode, false)) {
1191           // If we failed to create the file and it does not already exist.
1192           throw(nee);
1193         }
1194       } catch (InterruptedException ie) {
1195         zkw.interruptedException(ie);
1196       }
1197 
1198     } catch(InterruptedException ie) {
1199       zkw.interruptedException(ie);
1200     }
1201   }
1202 
1203   /**
1204    * Creates the specified node and all parent nodes required for it to exist.
1205    *
1206    * No watches are set and no errors are thrown if the node already exists.
1207    *
1208    * The nodes created are persistent and open access.
1209    *
1210    * @param zkw zk reference
1211    * @param znode path of node
1212    * @throws KeeperException if unexpected zookeeper exception
1213    */
1214   public static void createWithParents(ZooKeeperWatcher zkw, String znode)
1215   throws KeeperException {
1216     createWithParents(zkw, znode, new byte[0]);
1217   }
1218 
1219   /**
1220    * Creates the specified node and all parent nodes required for it to exist.  The creation of
1221    * parent znodes is not atomic with the leafe znode creation but the data is written atomically
1222    * when the leaf node is created.
1223    *
1224    * No watches are set and no errors are thrown if the node already exists.
1225    *
1226    * The nodes created are persistent and open access.
1227    *
1228    * @param zkw zk reference
1229    * @param znode path of node
1230    * @throws KeeperException if unexpected zookeeper exception
1231    */
1232   public static void createWithParents(ZooKeeperWatcher zkw, String znode, byte[] data)
1233   throws KeeperException {
1234     try {
1235       if(znode == null) {
1236         return;
1237       }
1238       zkw.getRecoverableZooKeeper().create(znode, data, createACL(zkw, znode),
1239           CreateMode.PERSISTENT);
1240     } catch(KeeperException.NodeExistsException nee) {
1241       return;
1242     } catch(KeeperException.NoNodeException nne) {
1243       createWithParents(zkw, getParent(znode));
1244       createWithParents(zkw, znode, data);
1245     } catch(InterruptedException ie) {
1246       zkw.interruptedException(ie);
1247     }
1248   }
1249 
1250   //
1251   // Deletes
1252   //
1253 
1254   /**
1255    * Delete the specified node.  Sets no watches.  Throws all exceptions.
1256    */
1257   public static void deleteNode(ZooKeeperWatcher zkw, String node)
1258   throws KeeperException {
1259     deleteNode(zkw, node, -1);
1260   }
1261 
1262   /**
1263    * Delete the specified node with the specified version.  Sets no watches.
1264    * Throws all exceptions.
1265    */
1266   public static boolean deleteNode(ZooKeeperWatcher zkw, String node,
1267       int version)
1268   throws KeeperException {
1269     try {
1270       zkw.getRecoverableZooKeeper().delete(node, version);
1271       return true;
1272     } catch(KeeperException.BadVersionException bve) {
1273       return false;
1274     } catch(InterruptedException ie) {
1275       zkw.interruptedException(ie);
1276       return false;
1277     }
1278   }
1279 
1280   /**
1281    * Deletes the specified node.  Fails silent if the node does not exist.
1282    * @param zkw
1283    * @param node
1284    * @throws KeeperException
1285    */
1286   public static void deleteNodeFailSilent(ZooKeeperWatcher zkw, String node)
1287   throws KeeperException {
1288     deleteNodeFailSilent(zkw,
1289       (DeleteNodeFailSilent)ZKUtilOp.deleteNodeFailSilent(node));
1290   }
1291 
1292   private static void deleteNodeFailSilent(ZooKeeperWatcher zkw,
1293       DeleteNodeFailSilent dnfs) throws KeeperException {
1294     DeleteRequest delete = (DeleteRequest)toZooKeeperOp(zkw, dnfs).toRequestRecord();
1295     try {
1296       zkw.getRecoverableZooKeeper().delete(delete.getPath(), delete.getVersion());
1297     } catch(KeeperException.NoNodeException nne) {
1298     } catch(InterruptedException ie) {
1299       zkw.interruptedException(ie);
1300     }
1301   }
1302 
1303 
1304   /**
1305    * Delete the specified node and all of it's children.
1306    * <p>
1307    * If the node does not exist, just returns.
1308    * <p>
1309    * Sets no watches. Throws all exceptions besides dealing with deletion of
1310    * children.
1311    */
1312   public static void deleteNodeRecursively(ZooKeeperWatcher zkw, String node)
1313   throws KeeperException {
1314     try {
1315       List<String> children = ZKUtil.listChildrenNoWatch(zkw, node);
1316       // the node is already deleted, so we just finish
1317       if (children == null) return;
1318 
1319       if(!children.isEmpty()) {
1320         for(String child : children) {
1321           deleteNodeRecursively(zkw, joinZNode(node, child));
1322         }
1323       }
1324       zkw.getRecoverableZooKeeper().delete(node, -1);
1325     } catch(InterruptedException ie) {
1326       zkw.interruptedException(ie);
1327     }
1328   }
1329 
1330   /**
1331    * Delete all the children of the specified node but not the node itself.
1332    *
1333    * Sets no watches.  Throws all exceptions besides dealing with deletion of
1334    * children.
1335    */
1336   public static void deleteChildrenRecursively(ZooKeeperWatcher zkw, String node)
1337   throws KeeperException {
1338     List<String> children = ZKUtil.listChildrenNoWatch(zkw, node);
1339     if (children == null || children.isEmpty()) return;
1340     for(String child : children) {
1341       deleteNodeRecursively(zkw, joinZNode(node, child));
1342     }
1343   }
1344 
1345   /**
1346    * Represents an action taken by ZKUtil, e.g. createAndFailSilent.
1347    * These actions are higher-level than ZKOp actions, which represent
1348    * individual actions in the ZooKeeper API, like create.
1349    */
1350   public abstract static class ZKUtilOp {
1351     private String path;
1352 
1353     private ZKUtilOp(String path) {
1354       this.path = path;
1355     }
1356 
1357     /**
1358      * @return a createAndFailSilent ZKUtilOp
1359      */
1360     public static ZKUtilOp createAndFailSilent(String path, byte[] data) {
1361       return new CreateAndFailSilent(path, data);
1362     }
1363 
1364     /**
1365      * @return a deleteNodeFailSilent ZKUtilOP
1366      */
1367     public static ZKUtilOp deleteNodeFailSilent(String path) {
1368       return new DeleteNodeFailSilent(path);
1369     }
1370 
1371     /**
1372      * @return a setData ZKUtilOp
1373      */
1374     public static ZKUtilOp setData(String path, byte [] data) {
1375       return new SetData(path, data);
1376     }
1377 
1378     /**
1379      * @return path to znode where the ZKOp will occur
1380      */
1381     public String getPath() {
1382       return path;
1383     }
1384 
1385     /**
1386      * ZKUtilOp representing createAndFailSilent in ZooKeeper
1387      * (attempt to create node, ignore error if already exists)
1388      */
1389     public static class CreateAndFailSilent extends ZKUtilOp {
1390       private byte [] data;
1391 
1392       private CreateAndFailSilent(String path, byte [] data) {
1393         super(path);
1394         this.data = data;
1395       }
1396 
1397       public byte[] getData() {
1398         return data;
1399       }
1400 
1401       @Override
1402       public boolean equals(Object o) {
1403         if (this == o) return true;
1404         if (!(o instanceof CreateAndFailSilent)) return false;
1405 
1406         CreateAndFailSilent op = (CreateAndFailSilent) o;
1407         return getPath().equals(op.getPath()) && Arrays.equals(data, op.data);
1408       }
1409 
1410       @Override
1411       public int hashCode() {
1412         int ret = 17 + getPath().hashCode() * 31;
1413         return ret * 31 + Bytes.hashCode(data);
1414       }
1415     }
1416 
1417     /**
1418      * ZKUtilOp representing deleteNodeFailSilent in ZooKeeper
1419      * (attempt to delete node, ignore error if node doesn't exist)
1420      */
1421     public static class DeleteNodeFailSilent extends ZKUtilOp {
1422       private DeleteNodeFailSilent(String path) {
1423         super(path);
1424       }
1425 
1426       @Override
1427       public boolean equals(Object o) {
1428         if (this == o) return true;
1429         if (!(o instanceof DeleteNodeFailSilent)) return false;
1430 
1431         return super.equals(o);
1432       }
1433 
1434       @Override
1435       public int hashCode() {
1436         return getPath().hashCode();
1437       }
1438     }
1439 
1440     /**
1441      * ZKUtilOp representing setData in ZooKeeper
1442      */
1443     public static class SetData extends ZKUtilOp {
1444       private byte [] data;
1445 
1446       private SetData(String path, byte [] data) {
1447         super(path);
1448         this.data = data;
1449       }
1450 
1451       public byte[] getData() {
1452         return data;
1453       }
1454 
1455       @Override
1456       public boolean equals(Object o) {
1457         if (this == o) return true;
1458         if (!(o instanceof SetData)) return false;
1459 
1460         SetData op = (SetData) o;
1461         return getPath().equals(op.getPath()) && Arrays.equals(data, op.data);
1462       }
1463 
1464       @Override
1465       public int hashCode() {
1466         int ret = getPath().hashCode();
1467         return ret * 31 + Bytes.hashCode(data);
1468       }
1469     }
1470   }
1471 
1472   /**
1473    * Convert from ZKUtilOp to ZKOp
1474    */
1475   private static Op toZooKeeperOp(ZooKeeperWatcher zkw, ZKUtilOp op)
1476   throws UnsupportedOperationException {
1477     if(op == null) return null;
1478 
1479     if (op instanceof CreateAndFailSilent) {
1480       CreateAndFailSilent cafs = (CreateAndFailSilent)op;
1481       return Op.create(cafs.getPath(), cafs.getData(), createACL(zkw, cafs.getPath()),
1482         CreateMode.PERSISTENT);
1483     } else if (op instanceof DeleteNodeFailSilent) {
1484       DeleteNodeFailSilent dnfs = (DeleteNodeFailSilent)op;
1485       return Op.delete(dnfs.getPath(), -1);
1486     } else if (op instanceof SetData) {
1487       SetData sd = (SetData)op;
1488       return Op.setData(sd.getPath(), sd.getData(), -1);
1489     } else {
1490       throw new UnsupportedOperationException("Unexpected ZKUtilOp type: "
1491         + op.getClass().getName());
1492     }
1493   }
1494 
1495   /**
1496    * If hbase.zookeeper.useMulti is true, use ZooKeeper's multi-update functionality.
1497    * Otherwise, run the list of operations sequentially.
1498    *
1499    * If all of the following are true:
1500    * - runSequentialOnMultiFailure is true
1501    * - hbase.zookeeper.useMulti is true
1502    * - on calling multi, we get a ZooKeeper exception that can be handled by a sequential call(*)
1503    * Then:
1504    * - we retry the operations one-by-one (sequentially)
1505    *
1506    * Note *: an example is receiving a NodeExistsException from a "create" call.  Without multi,
1507    * a user could call "createAndFailSilent" to ensure that a node exists if they don't care who
1508    * actually created the node (i.e. the NodeExistsException from ZooKeeper is caught).
1509    * This will cause all operations in the multi to fail, however, because
1510    * the NodeExistsException that zk.create throws will fail the multi transaction.
1511    * In this case, if the previous conditions hold, the commands are run sequentially, which should
1512    * result in the correct final state, but means that the operations will not run atomically.
1513    *
1514    * @throws KeeperException
1515    */
1516   public static void multiOrSequential(ZooKeeperWatcher zkw, List<ZKUtilOp> ops,
1517       boolean runSequentialOnMultiFailure) throws KeeperException {
1518     if (ops == null) return;
1519     boolean useMulti = zkw.getConfiguration().getBoolean(HConstants.ZOOKEEPER_USEMULTI, false);
1520 
1521     if (useMulti) {
1522       List<Op> zkOps = new LinkedList<Op>();
1523       for (ZKUtilOp op : ops) {
1524         zkOps.add(toZooKeeperOp(zkw, op));
1525       }
1526       try {
1527         zkw.getRecoverableZooKeeper().multi(zkOps);
1528       } catch (KeeperException ke) {
1529        switch (ke.code()) {
1530          case NODEEXISTS:
1531          case NONODE:
1532          case BADVERSION:
1533          case NOAUTH:
1534            // if we get an exception that could be solved by running sequentially
1535            // (and the client asked us to), then break out and run sequentially
1536            if (runSequentialOnMultiFailure) {
1537              LOG.info("On call to ZK.multi, received exception: " + ke.toString() + "."
1538                + "  Attempting to run operations sequentially because"
1539                + " runSequentialOnMultiFailure is: " + runSequentialOnMultiFailure + ".");
1540              processSequentially(zkw, ops);
1541              break;
1542            }
1543           default:
1544             throw ke;
1545         }
1546       } catch (InterruptedException ie) {
1547         zkw.interruptedException(ie);
1548       }
1549     } else {
1550       // run sequentially
1551       processSequentially(zkw, ops);
1552     }
1553 
1554   }
1555 
1556   private static void processSequentially(ZooKeeperWatcher zkw, List<ZKUtilOp> ops)
1557       throws KeeperException, NoNodeException {
1558     for (ZKUtilOp op : ops) {
1559       if (op instanceof CreateAndFailSilent) {
1560         createAndFailSilent(zkw, (CreateAndFailSilent) op);
1561       } else if (op instanceof DeleteNodeFailSilent) {
1562         deleteNodeFailSilent(zkw, (DeleteNodeFailSilent) op);
1563       } else if (op instanceof SetData) {
1564         setData(zkw, (SetData) op);
1565       } else {
1566         throw new UnsupportedOperationException("Unexpected ZKUtilOp type: "
1567             + op.getClass().getName());
1568       }
1569     }
1570   }
1571 
1572   //
1573   // ZooKeeper cluster information
1574   //
1575 
1576   /** @return String dump of everything in ZooKeeper. */
1577   public static String dump(ZooKeeperWatcher zkw) {
1578     StringBuilder sb = new StringBuilder();
1579     try {
1580       sb.append("HBase is rooted at ").append(zkw.baseZNode);
1581       sb.append("\nActive master address: ");
1582       try {
1583         sb.append(MasterAddressTracker.getMasterAddress(zkw));
1584       } catch (IOException e) {
1585         sb.append("<<FAILED LOOKUP: " + e.getMessage() + ">>");
1586       }
1587       sb.append("\nBackup master addresses:");
1588       for (String child : listChildrenNoWatch(zkw,
1589                                               zkw.backupMasterAddressesZNode)) {
1590         sb.append("\n ").append(child);
1591       }
1592       sb.append("\nRegion server holding hbase:meta: " + MetaRegionTracker.getMetaRegionLocation(zkw));
1593       sb.append("\nRegion servers:");
1594       for (String child : listChildrenNoWatch(zkw, zkw.rsZNode)) {
1595         sb.append("\n ").append(child);
1596       }
1597       try {
1598         getReplicationZnodesDump(zkw, sb);
1599       } catch (KeeperException ke) {
1600         LOG.warn("Couldn't get the replication znode dump", ke);
1601       }
1602       sb.append("\nQuorum Server Statistics:");
1603       String[] servers = zkw.getQuorum().split(",");
1604       for (String server : servers) {
1605         sb.append("\n ").append(server);
1606         try {
1607           String[] stat = getServerStats(server, ZKUtil.zkDumpConnectionTimeOut);
1608 
1609           if (stat == null) {
1610             sb.append("[Error] invalid quorum server: " + server);
1611             break;
1612           }
1613 
1614           for (String s : stat) {
1615             sb.append("\n  ").append(s);
1616           }
1617         } catch (Exception e) {
1618           sb.append("\n  ERROR: ").append(e.getMessage());
1619         }
1620       }
1621     } catch (KeeperException ke) {
1622       sb.append("\nFATAL ZooKeeper Exception!\n");
1623       sb.append("\n" + ke.getMessage());
1624     }
1625     return sb.toString();
1626   }
1627 
1628   /**
1629    * Appends replication znodes to the passed StringBuilder.
1630    * @param zkw
1631    * @param sb
1632    * @throws KeeperException
1633    */
1634   private static void getReplicationZnodesDump(ZooKeeperWatcher zkw, StringBuilder sb)
1635       throws KeeperException {
1636     String replicationZNodeName = zkw.getConfiguration().get("zookeeper.znode.replication",
1637       "replication");
1638     String replicationZnode = joinZNode(zkw.baseZNode, replicationZNodeName);
1639     if (ZKUtil.checkExists(zkw, replicationZnode) == -1) return;
1640     // do a ls -r on this znode
1641     sb.append("\n").append(replicationZnode).append(": ");
1642     List<String> children = ZKUtil.listChildrenNoWatch(zkw, replicationZnode);
1643     for (String child : children) {
1644       String znode = joinZNode(replicationZnode, child);
1645       if (child.equals(zkw.getConfiguration().get("zookeeper.znode.replication.peers", "peers"))) {
1646         appendPeersZnodes(zkw, znode, sb);
1647       } else if (child.equals(zkw.getConfiguration().
1648           get("zookeeper.znode.replication.rs", "rs"))) {
1649         appendRSZnodes(zkw, znode, sb);
1650       }
1651     }
1652   }
1653 
1654   private static void appendRSZnodes(ZooKeeperWatcher zkw, String znode, StringBuilder sb)
1655       throws KeeperException {
1656     List<String> stack = new LinkedList<String>();
1657     stack.add(znode);
1658     do {
1659       String znodeToProcess = stack.remove(stack.size() - 1);
1660       sb.append("\n").append(znodeToProcess).append(": ");
1661       byte[] data;
1662       try {
1663         data = ZKUtil.getData(zkw, znodeToProcess);
1664       } catch (InterruptedException e) {
1665         zkw.interruptedException(e);
1666         return;
1667       }
1668       if (data != null && data.length > 0) { // log position
1669         long position = 0;
1670         try {
1671           position = ZKUtil.parseHLogPositionFrom(ZKUtil.getData(zkw, znodeToProcess));
1672           sb.append(position);
1673         } catch (DeserializationException ignored) {
1674         } catch (InterruptedException e) {
1675           zkw.interruptedException(e);
1676           return;
1677         }
1678       }
1679       for (String zNodeChild : ZKUtil.listChildrenNoWatch(zkw, znodeToProcess)) {
1680         stack.add(ZKUtil.joinZNode(znodeToProcess, zNodeChild));
1681       }
1682     } while (stack.size() > 0);
1683   }
1684 
1685   private static void appendPeersZnodes(ZooKeeperWatcher zkw, String peersZnode,
1686     StringBuilder sb) throws KeeperException {
1687     int pblen = ProtobufUtil.lengthOfPBMagic();
1688     sb.append("\n").append(peersZnode).append(": ");
1689     for (String peerIdZnode : ZKUtil.listChildrenNoWatch(zkw, peersZnode)) {
1690       String znodeToProcess = ZKUtil.joinZNode(peersZnode, peerIdZnode);
1691       byte[] data;
1692       try {
1693         data = ZKUtil.getData(zkw, znodeToProcess);
1694       } catch (InterruptedException e) {
1695         zkw.interruptedException(e);
1696         return;
1697       }
1698       // parse the data of the above peer znode.
1699       try {
1700       String clusterKey = ZooKeeperProtos.ReplicationPeer.newBuilder().
1701         mergeFrom(data, pblen, data.length - pblen).getClusterkey();
1702       sb.append("\n").append(znodeToProcess).append(": ").append(clusterKey);
1703       // add the peer-state.
1704       appendPeerState(zkw, znodeToProcess, sb);
1705       } catch (InvalidProtocolBufferException ipbe) {
1706         LOG.warn("Got Exception while parsing peer: " + znodeToProcess, ipbe);
1707       }
1708     }
1709   }
1710 
1711   private static void appendPeerState(ZooKeeperWatcher zkw, String znodeToProcess,
1712       StringBuilder sb) throws KeeperException, InvalidProtocolBufferException {
1713     String peerState = zkw.getConfiguration().get("zookeeper.znode.replication.peers.state",
1714       "peer-state");
1715     int pblen = ProtobufUtil.lengthOfPBMagic();
1716     for (String child : ZKUtil.listChildrenNoWatch(zkw, znodeToProcess)) {
1717       if (!child.equals(peerState)) continue;
1718       String peerStateZnode = ZKUtil.joinZNode(znodeToProcess, child);
1719       sb.append("\n").append(peerStateZnode).append(": ");
1720       byte[] peerStateData;
1721       try {
1722         peerStateData = ZKUtil.getData(zkw, peerStateZnode);
1723         sb.append(ZooKeeperProtos.ReplicationState.newBuilder()
1724             .mergeFrom(peerStateData, pblen, peerStateData.length - pblen).getState().name());
1725       } catch (InterruptedException e) {
1726         zkw.interruptedException(e);
1727         return;
1728       }
1729     }
1730   }
1731 
1732   /**
1733    * Gets the statistics from the given server.
1734    *
1735    * @param server  The server to get the statistics from.
1736    * @param timeout  The socket timeout to use.
1737    * @return The array of response strings.
1738    * @throws IOException When the socket communication fails.
1739    */
1740   public static String[] getServerStats(String server, int timeout)
1741   throws IOException {
1742     String[] sp = server.split(":");
1743     if (sp == null || sp.length == 0) {
1744       return null;
1745     }
1746 
1747     String host = sp[0];
1748     int port = sp.length > 1 ? Integer.parseInt(sp[1])
1749         : HConstants.DEFAULT_ZOOKEPER_CLIENT_PORT;
1750 
1751     Socket socket = new Socket();
1752     InetSocketAddress sockAddr = new InetSocketAddress(host, port);
1753     socket.connect(sockAddr, timeout);
1754 
1755     socket.setSoTimeout(timeout);
1756     PrintWriter out = new PrintWriter(socket.getOutputStream(), true);
1757     BufferedReader in = new BufferedReader(new InputStreamReader(
1758       socket.getInputStream()));
1759     out.println("stat");
1760     out.flush();
1761     ArrayList<String> res = new ArrayList<String>();
1762     while (true) {
1763       String line = in.readLine();
1764       if (line != null) {
1765         res.add(line);
1766       } else {
1767         break;
1768       }
1769     }
1770     socket.close();
1771     return res.toArray(new String[res.size()]);
1772   }
1773 
1774   private static void logRetrievedMsg(final ZooKeeperWatcher zkw,
1775       final String znode, final byte [] data, final boolean watcherSet) {
1776     if (!LOG.isTraceEnabled()) return;
1777     LOG.trace(zkw.prefix("Retrieved " + ((data == null)? 0: data.length) +
1778       " byte(s) of data from znode " + znode +
1779       (watcherSet? " and set watcher; ": "; data=") +
1780       (data == null? "null": data.length == 0? "empty": (
1781           znode.startsWith(zkw.assignmentZNode)?
1782             ZKAssign.toString(data): // We should not be doing this reaching into another class
1783           znode.startsWith(zkw.metaServerZNode)?
1784             getServerNameOrEmptyString(data):
1785           znode.startsWith(zkw.backupMasterAddressesZNode)?
1786             getServerNameOrEmptyString(data):
1787           StringUtils.abbreviate(Bytes.toStringBinary(data), 32)))));
1788   }
1789 
1790   private static String getServerNameOrEmptyString(final byte [] data) {
1791     try {
1792       return ServerName.parseFrom(data).toString();
1793     } catch (DeserializationException e) {
1794       return "";
1795     }
1796   }
1797 
1798   /**
1799    * Waits for HBase installation's base (parent) znode to become available.
1800    * @throws IOException on ZK errors
1801    */
1802   public static void waitForBaseZNode(Configuration conf) throws IOException {
1803     LOG.info("Waiting until the base znode is available");
1804     String parentZNode = conf.get(HConstants.ZOOKEEPER_ZNODE_PARENT,
1805         HConstants.DEFAULT_ZOOKEEPER_ZNODE_PARENT);
1806     ZooKeeper zk = new ZooKeeper(ZKConfig.getZKQuorumServersString(conf),
1807         conf.getInt(HConstants.ZK_SESSION_TIMEOUT,
1808         HConstants.DEFAULT_ZK_SESSION_TIMEOUT), EmptyWatcher.instance);
1809 
1810     final int maxTimeMs = 10000;
1811     final int maxNumAttempts = maxTimeMs / HConstants.SOCKET_RETRY_WAIT_MS;
1812 
1813     KeeperException keeperEx = null;
1814     try {
1815       try {
1816         for (int attempt = 0; attempt < maxNumAttempts; ++attempt) {
1817           try {
1818             if (zk.exists(parentZNode, false) != null) {
1819               LOG.info("Parent znode exists: " + parentZNode);
1820               keeperEx = null;
1821               break;
1822             }
1823           } catch (KeeperException e) {
1824             keeperEx = e;
1825           }
1826           Threads.sleepWithoutInterrupt(HConstants.SOCKET_RETRY_WAIT_MS);
1827         }
1828       } finally {
1829         zk.close();
1830       }
1831     } catch (InterruptedException ex) {
1832       Thread.currentThread().interrupt();
1833     }
1834 
1835     if (keeperEx != null) {
1836       throw new IOException(keeperEx);
1837     }
1838   }
1839 
1840 
1841   public static byte[] blockUntilAvailable(
1842     final ZooKeeperWatcher zkw, final String znode, final long timeout)
1843     throws InterruptedException {
1844     if (timeout < 0) throw new IllegalArgumentException();
1845     if (zkw == null) throw new IllegalArgumentException();
1846     if (znode == null) throw new IllegalArgumentException();
1847 
1848     byte[] data = null;
1849     boolean finished = false;
1850     final long endTime = System.currentTimeMillis() + timeout;
1851     while (!finished) {
1852       try {
1853         data = ZKUtil.getData(zkw, znode);
1854       } catch(KeeperException e) {
1855         LOG.warn("Unexpected exception handling blockUntilAvailable", e);
1856       }
1857 
1858       if (data == null && (System.currentTimeMillis() +
1859         HConstants.SOCKET_RETRY_WAIT_MS < endTime)) {
1860         Thread.sleep(HConstants.SOCKET_RETRY_WAIT_MS);
1861       } else {
1862         finished = true;
1863       }
1864     }
1865 
1866     return data;
1867   }
1868 
1869 
1870   /**
1871    * Convert a {@link DeserializationException} to a more palatable {@link KeeperException}.
1872    * Used when can't let a {@link DeserializationException} out w/o changing public API.
1873    * @param e Exception to convert
1874    * @return Converted exception
1875    */
1876   public static KeeperException convert(final DeserializationException e) {
1877     KeeperException ke = new KeeperException.DataInconsistencyException();
1878     ke.initCause(e);
1879     return ke;
1880   }
1881 
1882   /**
1883    * Recursively print the current state of ZK (non-transactional)
1884    * @param root name of the root directory in zk to print
1885    * @throws KeeperException
1886    */
1887   public static void logZKTree(ZooKeeperWatcher zkw, String root) {
1888     if (!LOG.isDebugEnabled()) return;
1889     LOG.debug("Current zk system:");
1890     String prefix = "|-";
1891     LOG.debug(prefix + root);
1892     try {
1893       logZKTree(zkw, root, prefix);
1894     } catch (KeeperException e) {
1895       throw new RuntimeException(e);
1896     }
1897   }
1898 
1899   /**
1900    * Helper method to print the current state of the ZK tree.
1901    * @see #logZKTree(ZooKeeperWatcher, String)
1902    * @throws KeeperException if an unexpected exception occurs
1903    */
1904   protected static void logZKTree(ZooKeeperWatcher zkw, String root, String prefix) throws KeeperException {
1905     List<String> children = ZKUtil.listChildrenNoWatch(zkw, root);
1906     if (children == null) return;
1907     for (String child : children) {
1908       LOG.debug(prefix + child);
1909       String node = ZKUtil.joinZNode(root.equals("/") ? "" : root, child);
1910       logZKTree(zkw, node, prefix + "---");
1911     }
1912   }
1913 
1914   /**
1915    * @param position
1916    * @return Serialized protobuf of <code>position</code> with pb magic prefix prepended suitable
1917    *         for use as content of an hlog position in a replication queue.
1918    */
1919   public static byte[] positionToByteArray(final long position) {
1920     byte[] bytes = ZooKeeperProtos.ReplicationHLogPosition.newBuilder().setPosition(position)
1921         .build().toByteArray();
1922     return ProtobufUtil.prependPBMagic(bytes);
1923   }
1924 
1925   /**
1926    * @param bytes - Content of a HLog position znode.
1927    * @return long - The current HLog position.
1928    * @throws DeserializationException
1929    */
1930   public static long parseHLogPositionFrom(final byte[] bytes) throws DeserializationException {
1931     if (bytes == null) {
1932       throw new DeserializationException("Unable to parse null HLog position.");
1933     }
1934     if (ProtobufUtil.isPBMagicPrefix(bytes)) {
1935       int pblen = ProtobufUtil.lengthOfPBMagic();
1936       ZooKeeperProtos.ReplicationHLogPosition.Builder builder =
1937           ZooKeeperProtos.ReplicationHLogPosition.newBuilder();
1938       ZooKeeperProtos.ReplicationHLogPosition position;
1939       try {
1940         position = builder.mergeFrom(bytes, pblen, bytes.length - pblen).build();
1941       } catch (InvalidProtocolBufferException e) {
1942         throw new DeserializationException(e);
1943       }
1944       return position.getPosition();
1945     } else {
1946       if (bytes.length > 0) {
1947         return Bytes.toLong(bytes);
1948       }
1949       return 0;
1950     }
1951   }
1952 
1953   /**
1954    * @param regionLastFlushedSequenceId the flushed sequence id of a region which is the min of its
1955    *          store max seq ids
1956    * @param storeSequenceIds column family to sequence Id map
1957    * @return Serialized protobuf of <code>RegionSequenceIds</code> with pb magic prefix prepended
1958    *         suitable for use to filter wal edits in distributedLogReplay mode
1959    */
1960   public static byte[] regionSequenceIdsToByteArray(final Long regionLastFlushedSequenceId,
1961       final Map<byte[], Long> storeSequenceIds) {
1962     ZooKeeperProtos.RegionStoreSequenceIds.Builder regionSequenceIdsBuilder =
1963         ZooKeeperProtos.RegionStoreSequenceIds.newBuilder();
1964     ZooKeeperProtos.StoreSequenceId.Builder storeSequenceIdBuilder =
1965         ZooKeeperProtos.StoreSequenceId.newBuilder();
1966     if (storeSequenceIds != null) {
1967       for (Map.Entry<byte[], Long> e : storeSequenceIds.entrySet()){
1968         byte[] columnFamilyName = e.getKey();
1969         Long curSeqId = e.getValue();
1970         storeSequenceIdBuilder.setFamilyName(HBaseZeroCopyByteString.wrap(columnFamilyName));
1971         storeSequenceIdBuilder.setSequenceId(curSeqId);
1972         regionSequenceIdsBuilder.addStoreSequenceId(storeSequenceIdBuilder.build());
1973         storeSequenceIdBuilder.clear();
1974       }
1975     }
1976     regionSequenceIdsBuilder.setLastFlushedSequenceId(regionLastFlushedSequenceId);
1977     byte[] result = regionSequenceIdsBuilder.build().toByteArray();
1978     return ProtobufUtil.prependPBMagic(result);
1979   }
1980 
1981   /**
1982    * @param bytes Content of serialized data of RegionStoreSequenceIds
1983    * @return a RegionStoreSequenceIds object
1984    * @throws DeserializationException
1985    */
1986   public static RegionStoreSequenceIds parseRegionStoreSequenceIds(final byte[] bytes)
1987       throws DeserializationException {
1988     if (bytes == null || !ProtobufUtil.isPBMagicPrefix(bytes)) {
1989       throw new DeserializationException("Unable to parse RegionStoreSequenceIds.");
1990     }
1991     RegionStoreSequenceIds.Builder regionSequenceIdsBuilder =
1992         ZooKeeperProtos.RegionStoreSequenceIds.newBuilder();
1993     int pblen = ProtobufUtil.lengthOfPBMagic();
1994     RegionStoreSequenceIds storeIds = null;
1995     try {
1996       storeIds = regionSequenceIdsBuilder.mergeFrom(bytes, pblen, bytes.length - pblen).build();
1997     } catch (InvalidProtocolBufferException e) {
1998       throw new DeserializationException(e);
1999     }
2000     return storeIds;
2001   }
2002 }