View Javadoc

1   /*
2    * Copyright The Apache Software Foundation
3    *
4    * Licensed to the Apache Software Foundation (ASF) under one or more
5    * contributor license agreements. See the NOTICE file distributed with this
6    * work for additional information regarding copyright ownership. The ASF
7    * licenses this file to you under the Apache License, Version 2.0 (the
8    * "License"); you may not use this file except in compliance with the License.
9    * 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, WITHOUT
15   * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
16   * License for the specific language governing permissions and limitations
17   * under the License.
18   */
19  package org.apache.hadoop.hbase.coprocessor.example;
20  
21  import java.io.IOException;
22  import java.util.Collections;
23  import java.util.List;
24  import java.util.NavigableSet;
25  
26  import org.apache.commons.logging.Log;
27  import org.apache.commons.logging.LogFactory;
28  import org.apache.hadoop.hbase.CoprocessorEnvironment;
29  import org.apache.hadoop.hbase.HConstants;
30  import org.apache.hadoop.hbase.client.IsolationLevel;
31  import org.apache.hadoop.hbase.client.Scan;
32  import org.apache.hadoop.hbase.coprocessor.BaseRegionObserver;
33  import org.apache.hadoop.hbase.coprocessor.ObserverContext;
34  import org.apache.hadoop.hbase.coprocessor.RegionCoprocessorEnvironment;
35  import org.apache.hadoop.hbase.coprocessor.RegionObserver;
36  import org.apache.hadoop.hbase.regionserver.HStore;
37  import org.apache.hadoop.hbase.regionserver.Store;
38  import org.apache.hadoop.hbase.regionserver.ScanInfo;
39  import org.apache.hadoop.hbase.regionserver.InternalScanner;
40  import org.apache.hadoop.hbase.regionserver.KeyValueScanner;
41  import org.apache.hadoop.hbase.regionserver.ScanType;
42  import org.apache.hadoop.hbase.regionserver.Store;
43  import org.apache.hadoop.hbase.regionserver.StoreScanner;
44  import org.apache.hadoop.hbase.util.Bytes;
45  import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
46  import org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher;
47  import org.apache.zookeeper.KeeperException;
48  import org.apache.zookeeper.WatchedEvent;
49  import org.apache.zookeeper.Watcher;
50  import org.apache.zookeeper.ZooKeeper;
51  
52  /**
53   * This is an example showing how a RegionObserver could configured
54   * via ZooKeeper in order to control a Region compaction, flush, and scan policy.
55   *
56   * This also demonstrated the use of shared {@link RegionObserver} state.
57   * See {@link RegionCoprocessorEnvironment#getSharedData()}.
58   *
59   * This would be useful for an incremental backup tool, which would indicate the last
60   * time of a successful backup via ZK and instruct HBase to not delete data that was
61   * inserted since (based on wall clock time). 
62   *
63   * This implements org.apache.zookeeper.Watcher directly instead of using
64   * {@link ZooKeeperWatcher}, because RegionObservers come and go and currently
65   * listeners registered with ZooKeeperWatcher cannot be removed.
66   */
67  public class ZooKeeperScanPolicyObserver extends BaseRegionObserver {
68    public static final String node = "/backup/example/lastbackup";
69    public static final String zkkey = "ZK";
70    private static final Log LOG = LogFactory.getLog(ZooKeeperScanPolicyObserver.class);
71  
72    /**
73     * Internal watcher that keep "data" up to date asynchronously.
74     */
75    private static class ZKWatcher implements Watcher {
76      private byte[] data = null;
77      private ZooKeeper zk;
78      private volatile boolean needSetup = true;
79      private volatile long lastSetupTry = 0;
80  
81      public ZKWatcher(ZooKeeper zk) {
82        this.zk = zk;
83        // trigger the listening
84        getData();
85      }
86  
87      /**
88       * Get the maintained data. In case of any ZK exceptions this will retry
89       * establishing the connection (but not more than twice/minute).
90       *
91       * getData is on the critical path, so make sure it is fast unless there is
92       * a problem (network partion, ZK ensemble down, etc)
93       * Make sure at most one (unlucky) thread retries and other threads don't pile up
94       * while that threads tries to recreate the connection.
95       *
96       * @return the last know version of the data
97       */
98      @edu.umd.cs.findbugs.annotations.SuppressWarnings(value="REC_CATCH_EXCEPTION")
99      public byte[] getData() {
100       // try at most twice/minute
101       if (needSetup && EnvironmentEdgeManager.currentTime() > lastSetupTry + 30000) {
102         synchronized (this) {
103           // make sure only one thread tries to reconnect
104           if (needSetup) {
105             needSetup = false;
106           } else {
107             return data;
108           }
109         }
110         // do this without the lock held to avoid threads piling up on this lock,
111         // as it can take a while
112         try {
113           LOG.debug("Connecting to ZK");
114           // record this attempt
115           lastSetupTry = EnvironmentEdgeManager.currentTime();
116           if (zk.exists(node, false) != null) {
117             data = zk.getData(node, this, null);
118             LOG.debug("Read synchronously: "+(data == null ? "null" : Bytes.toLong(data)));
119           } else {
120             zk.exists(node, this);
121           }
122         } catch (Exception x) {
123           // try again if this fails
124           needSetup = true;
125         }
126       }
127       return data;
128     }
129 
130     @Override
131     public void process(WatchedEvent event) {
132       switch(event.getType()) {
133       case NodeDataChanged:
134       case NodeCreated:
135       try {
136         // get data and re-watch
137         data = zk.getData(node, this, null);
138         LOG.debug("Read asynchronously: "+(data == null ? "null" : Bytes.toLong(data)));
139       } catch (InterruptedException ix) {
140       } catch (KeeperException kx) {
141         needSetup = true;
142       }
143       break;
144 
145       case NodeDeleted:
146       try {
147         // just re-watch
148         zk.exists(node, this);
149         data = null;
150       } catch (InterruptedException ix) {
151       } catch (KeeperException kx) {
152         needSetup = true;
153       }
154       break;
155 
156       default:
157         // ignore
158       }
159     }
160   }
161 
162   @Override
163   public void start(CoprocessorEnvironment e) throws IOException {
164     RegionCoprocessorEnvironment re = (RegionCoprocessorEnvironment) e;
165     if (!re.getSharedData().containsKey(zkkey)) {
166       // there is a short race here
167       // in the worst case we create a watcher that will be notified once
168       re.getSharedData().putIfAbsent(
169           zkkey,
170           new ZKWatcher(re.getRegionServerServices().getZooKeeper()
171               .getRecoverableZooKeeper().getZooKeeper()));
172     }
173   }
174 
175   @Override
176   public void stop(CoprocessorEnvironment e) throws IOException {
177     // nothing to do here
178   }
179 
180   protected ScanInfo getScanInfo(Store store, RegionCoprocessorEnvironment e) {
181     byte[] data = ((ZKWatcher)e.getSharedData().get(zkkey)).getData();
182     if (data == null) {
183       return null;
184     }
185     ScanInfo oldSI = store.getScanInfo();
186     if (oldSI.getTtl() == Long.MAX_VALUE) {
187       return null;
188     }
189     long ttl = Math.max(EnvironmentEdgeManager.currentTime() -
190         Bytes.toLong(data), oldSI.getTtl());
191     return new ScanInfo(store.getFamily(), ttl,
192         oldSI.getTimeToPurgeDeletes(), oldSI.getComparator());
193   }
194 
195   @Override
196   public InternalScanner preFlushScannerOpen(final ObserverContext<RegionCoprocessorEnvironment> c,
197       Store store, KeyValueScanner memstoreScanner, InternalScanner s) throws IOException {
198     ScanInfo scanInfo = getScanInfo(store, c.getEnvironment());
199     if (scanInfo == null) {
200       // take default action
201       return null;
202     }
203     Scan scan = new Scan();
204     scan.setMaxVersions(scanInfo.getMaxVersions());
205     return new StoreScanner(store, scanInfo, scan, Collections.singletonList(memstoreScanner),
206         ScanType.COMPACT_RETAIN_DELETES, store.getSmallestReadPoint(), HConstants.OLDEST_TIMESTAMP);
207   }
208 
209   @Override
210   public InternalScanner preCompactScannerOpen(
211       final ObserverContext<RegionCoprocessorEnvironment> c,
212       Store store, List<? extends KeyValueScanner> scanners, ScanType scanType, long earliestPutTs,
213       InternalScanner s) throws IOException {
214     ScanInfo scanInfo = getScanInfo(store, c.getEnvironment());
215     if (scanInfo == null) {
216       // take default action
217       return null;
218     }
219     Scan scan = new Scan();
220     scan.setMaxVersions(scanInfo.getMaxVersions());
221     return new StoreScanner(store, scanInfo, scan, scanners, scanType,
222         store.getSmallestReadPoint(), earliestPutTs);
223   }
224 
225   @Override
226   public KeyValueScanner preStoreScannerOpen(final ObserverContext<RegionCoprocessorEnvironment> c,
227       final Store store, final Scan scan, final NavigableSet<byte[]> targetCols,
228       final KeyValueScanner s) throws IOException {
229     ScanInfo scanInfo = getScanInfo(store, c.getEnvironment());
230     if (scanInfo == null) {
231       // take default action
232       return null;
233     }
234     return new StoreScanner(store, scanInfo, scan, targetCols,
235       ((HStore)store).getHRegion().getReadpoint(IsolationLevel.READ_COMMITTED));
236   }
237 }