Source code

001/**
002 * Licensed to the Apache Software Foundation (ASF) under one
003 * or more contributor license agreements.  See the NOTICE file
004 * distributed with this work for additional information
005 * regarding copyright ownership.  The ASF licenses this file
006 * to you under the Apache License, Version 2.0 (the
007 * "License"); you may not use this file except in compliance
008 * with the License.  You may obtain a copy of the License at
009 *
010 *     http://www.apache.org/licenses/LICENSE-2.0
011 *
012 * Unless required by applicable law or agreed to in writing, software
013 * distributed under the License is distributed on an "AS IS" BASIS,
014 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
015 * See the License for the specific language governing permissions and
016 * limitations under the License.
017 */
018package org.apache.hadoop.hbase;
019
020import static org.junit.Assert.assertEquals;
021import static org.junit.Assert.assertFalse;
022import static org.junit.Assert.assertTrue;
023
024import java.util.List;
025import java.util.Map;
026import org.apache.hadoop.conf.Configuration;
027import org.apache.hadoop.hbase.client.Admin;
028import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;
029import org.apache.hadoop.hbase.client.Put;
030import org.apache.hadoop.hbase.client.RegionInfo;
031import org.apache.hadoop.hbase.client.ResultScanner;
032import org.apache.hadoop.hbase.client.Scan;
033import org.apache.hadoop.hbase.client.Table;
034import org.apache.hadoop.hbase.client.TableDescriptor;
035import org.apache.hadoop.hbase.client.TableDescriptorBuilder;
036import org.apache.hadoop.hbase.coordination.ZkSplitLogWorkerCoordination;
037import org.apache.hadoop.hbase.master.HMaster;
038import org.apache.hadoop.hbase.master.LoadBalancer;
039import org.apache.hadoop.hbase.master.balancer.SimpleLoadBalancer;
040import org.apache.hadoop.hbase.testclassification.LargeTests;
041import org.apache.hadoop.hbase.testclassification.MiscTests;
042import org.apache.hadoop.hbase.util.Bytes;
043import org.apache.hadoop.hbase.util.FSUtils;
044import org.apache.hadoop.hbase.zookeeper.ZKUtil;
045import org.apache.hadoop.hbase.zookeeper.ZKWatcher;
046import org.apache.zookeeper.KeeperException;
047import org.junit.After;
048import org.junit.AfterClass;
049import org.junit.Before;
050import org.junit.BeforeClass;
051import org.junit.ClassRule;
052import org.junit.Rule;
053import org.junit.Test;
054import org.junit.experimental.categories.Category;
055import org.junit.rules.TestName;
056import org.slf4j.Logger;
057import org.slf4j.LoggerFactory;
058
059@Category({MiscTests.class, LargeTests.class})
060public class TestZooKeeper {
061
062  @ClassRule
063  public static final HBaseClassTestRule CLASS_RULE =
064      HBaseClassTestRule.forClass(TestZooKeeper.class);
065
066  private static final Logger LOG = LoggerFactory.getLogger(TestZooKeeper.class);
067
068  private final static HBaseTestingUtility TEST_UTIL = new HBaseTestingUtility();
069
070  @Rule
071  public TestName name = new TestName();
072
073  @BeforeClass
074  public static void setUpBeforeClass() throws Exception {
075    // Test we can first start the ZK cluster by itself
076    Configuration conf = TEST_UTIL.getConfiguration();
077    TEST_UTIL.startMiniDFSCluster(2);
078    TEST_UTIL.startMiniZKCluster();
079    conf.setInt(HConstants.ZK_SESSION_TIMEOUT, 1000);
080    conf.setClass(HConstants.HBASE_MASTER_LOADBALANCER_CLASS, MockLoadBalancer.class,
081        LoadBalancer.class);
082  }
083
084  @AfterClass
085  public static void tearDownAfterClass() throws Exception {
086    TEST_UTIL.shutdownMiniCluster();
087  }
088
089  @Before
090  public void setUp() throws Exception {
091    TEST_UTIL.startMiniHBaseCluster(2, 2);
092  }
093
094  @After
095  public void after() throws Exception {
096    try {
097      TEST_UTIL.getHBaseCluster().waitForActiveAndReadyMaster(10000);
098      // Some regionserver could fail to delete its znode.
099      // So shutdown could hang. Let's kill them all instead.
100      TEST_UTIL.getHBaseCluster().killAll();
101
102      // Still need to clean things up
103      TEST_UTIL.shutdownMiniHBaseCluster();
104    } finally {
105      TEST_UTIL.getTestFileSystem().delete(FSUtils.getRootDir(TEST_UTIL.getConfiguration()), true);
106      ZKUtil.deleteNodeRecursively(TEST_UTIL.getZooKeeperWatcher(), "/hbase");
107    }
108  }
109
110  @Test
111  public void testRegionServerSessionExpired() throws Exception {
112    LOG.info("Starting " + name.getMethodName());
113    TEST_UTIL.expireRegionServerSession(0);
114    testSanity(name.getMethodName());
115  }
116
117  @Test
118  public void testMasterSessionExpired() throws Exception {
119    LOG.info("Starting " + name.getMethodName());
120    TEST_UTIL.expireMasterSession();
121    testSanity(name.getMethodName());
122  }
123
124  /**
125   * Master recovery when the znode already exists. Internally, this
126   *  test differs from {@link #testMasterSessionExpired} because here
127   *  the master znode will exist in ZK.
128   */
129  @Test
130  public void testMasterZKSessionRecoveryFailure() throws Exception {
131    LOG.info("Starting " + name.getMethodName());
132    MiniHBaseCluster cluster = TEST_UTIL.getHBaseCluster();
133    HMaster m = cluster.getMaster();
134    m.abort("Test recovery from zk session expired",
135        new KeeperException.SessionExpiredException());
136    assertTrue(m.isStopped()); // Master doesn't recover any more
137    testSanity(name.getMethodName());
138  }
139
140  /**
141   * Make sure we can use the cluster
142   */
143  private void testSanity(final String testName) throws Exception {
144    String tableName = testName + "_" + System.currentTimeMillis();
145    TableDescriptor desc = TableDescriptorBuilder.newBuilder(TableName.valueOf(tableName))
146        .setColumnFamily(ColumnFamilyDescriptorBuilder.of("fam")).build();
147    LOG.info("Creating table " + tableName);
148    Admin admin = TEST_UTIL.getAdmin();
149    try {
150      admin.createTable(desc);
151    } finally {
152      admin.close();
153    }
154
155    Table table = TEST_UTIL.getConnection().getTable(desc.getTableName());
156    Put put = new Put(Bytes.toBytes("testrow"));
157    put.addColumn(Bytes.toBytes("fam"), Bytes.toBytes("col"), Bytes.toBytes("testdata"));
158    LOG.info("Putting table " + tableName);
159    table.put(put);
160    table.close();
161  }
162
163  /**
164   * Tests that the master does not call retainAssignment after recovery from expired zookeeper
165   * session. Without the HBASE-6046 fix master always tries to assign all the user regions by
166   * calling retainAssignment.
167   */
168  @Test
169  public void testRegionAssignmentAfterMasterRecoveryDueToZKExpiry() throws Exception {
170    MiniHBaseCluster cluster = TEST_UTIL.getHBaseCluster();
171    cluster.startRegionServer();
172    cluster.waitForActiveAndReadyMaster(10000);
173    HMaster m = cluster.getMaster();
174    final ZKWatcher zkw = m.getZooKeeper();
175    // now the cluster is up. So assign some regions.
176    try (Admin admin = TEST_UTIL.getAdmin()) {
177      byte[][] SPLIT_KEYS = new byte[][] { Bytes.toBytes("a"), Bytes.toBytes("b"),
178          Bytes.toBytes("c"), Bytes.toBytes("d"), Bytes.toBytes("e"), Bytes.toBytes("f"),
179          Bytes.toBytes("g"), Bytes.toBytes("h"), Bytes.toBytes("i"), Bytes.toBytes("j") };
180      TableDescriptor htd =
181          TableDescriptorBuilder.newBuilder(TableName.valueOf(name.getMethodName()))
182              .setColumnFamily(ColumnFamilyDescriptorBuilder.of(HConstants.CATALOG_FAMILY)).build();
183      admin.createTable(htd, SPLIT_KEYS);
184      TEST_UTIL.waitUntilNoRegionsInTransition(60000);
185      m.getZooKeeper().close();
186      MockLoadBalancer.retainAssignCalled = false;
187      final int expectedNumOfListeners = countPermanentListeners(zkw);
188      m.abort("Test recovery from zk session expired",
189          new KeeperException.SessionExpiredException());
190      assertTrue(m.isStopped()); // Master doesn't recover any more
191      // The recovered master should not call retainAssignment, as it is not a
192      // clean startup.
193      assertFalse("Retain assignment should not be called", MockLoadBalancer.retainAssignCalled);
194      // number of listeners should be same as the value before master aborted
195      // wait for new master is initialized
196      cluster.waitForActiveAndReadyMaster(120000);
197      final HMaster newMaster = cluster.getMasterThread().getMaster();
198      assertEquals(expectedNumOfListeners, countPermanentListeners(newMaster.getZooKeeper()));
199    }
200  }
201
202  /**
203   * Count listeners in zkw excluding listeners, that belongs to workers or other
204   * temporary processes.
205   */
206  private int countPermanentListeners(ZKWatcher watcher) {
207    return countListeners(watcher, ZkSplitLogWorkerCoordination.class);
208  }
209
210  /**
211   * Count listeners in zkw excluding provided classes
212   */
213  private int countListeners(ZKWatcher watcher, Class<?>... exclude) {
214    int cnt = 0;
215    for (Object o : watcher.getListeners()) {
216      boolean skip = false;
217      for (Class<?> aClass : exclude) {
218        if (aClass.isAssignableFrom(o.getClass())) {
219          skip = true;
220          break;
221        }
222      }
223      if (!skip) {
224        cnt += 1;
225      }
226    }
227    return cnt;
228  }
229
230  /**
231   * Tests whether the logs are split when master recovers from a expired zookeeper session and an
232   * RS goes down.
233   */
234  @Test
235  public void testLogSplittingAfterMasterRecoveryDueToZKExpiry() throws Exception {
236    MiniHBaseCluster cluster = TEST_UTIL.getHBaseCluster();
237    cluster.startRegionServer();
238    TableName tableName = TableName.valueOf(name.getMethodName());
239    byte[] family = Bytes.toBytes("col");
240    try (Admin admin = TEST_UTIL.getAdmin()) {
241      byte[][] SPLIT_KEYS = new byte[][] { Bytes.toBytes("1"), Bytes.toBytes("2"),
242        Bytes.toBytes("3"), Bytes.toBytes("4"), Bytes.toBytes("5") };
243      TableDescriptor htd = TableDescriptorBuilder.newBuilder(tableName)
244          .setColumnFamily(ColumnFamilyDescriptorBuilder.of(family)).build();
245      admin.createTable(htd, SPLIT_KEYS);
246    }
247    TEST_UTIL.waitUntilNoRegionsInTransition(60000);
248    HMaster m = cluster.getMaster();
249    try (Table table = TEST_UTIL.getConnection().getTable(tableName)) {
250      int numberOfPuts;
251      for (numberOfPuts = 0; numberOfPuts < 6; numberOfPuts++) {
252        Put p = new Put(Bytes.toBytes(numberOfPuts));
253        p.addColumn(Bytes.toBytes("col"), Bytes.toBytes("ql"),
254          Bytes.toBytes("value" + numberOfPuts));
255        table.put(p);
256      }
257      m.abort("Test recovery from zk session expired",
258        new KeeperException.SessionExpiredException());
259      assertTrue(m.isStopped()); // Master doesn't recover any more
260      cluster.killRegionServer(TEST_UTIL.getRSForFirstRegionInTable(tableName).getServerName());
261      // Without patch for HBASE-6046 this test case will always timeout
262      // with patch the test case should pass.
263      int numberOfRows = 0;
264      try (ResultScanner scanner = table.getScanner(new Scan())) {
265        while (scanner.next() != null) {
266          numberOfRows++;
267        }
268      }
269      assertEquals("Number of rows should be equal to number of puts.", numberOfPuts, numberOfRows);
270    }
271  }
272
273  static class MockLoadBalancer extends SimpleLoadBalancer {
274    static boolean retainAssignCalled = false;
275
276    @Override
277    public Map<ServerName, List<RegionInfo>> retainAssignment(
278        Map<RegionInfo, ServerName> regions, List<ServerName> servers) throws HBaseIOException {
279      retainAssignCalled = true;
280      return super.retainAssignment(regions, servers);
281    }
282  }
283
284}
285