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.coprocessor;
019
020import static org.junit.Assert.assertEquals;
021import static org.junit.Assert.assertFalse;
022import static org.junit.Assert.assertTrue;
023import static org.junit.Assert.fail;
024
025import java.io.IOException;
026import java.util.Optional;
027import org.apache.hadoop.conf.Configuration;
028import org.apache.hadoop.hbase.Abortable;
029import org.apache.hadoop.hbase.CoprocessorEnvironment;
030import org.apache.hadoop.hbase.HBaseClassTestRule;
031import org.apache.hadoop.hbase.HBaseTestingUtility;
032import org.apache.hadoop.hbase.HColumnDescriptor;
033import org.apache.hadoop.hbase.HConstants;
034import org.apache.hadoop.hbase.HTableDescriptor;
035import org.apache.hadoop.hbase.MiniHBaseCluster;
036import org.apache.hadoop.hbase.TableName;
037import org.apache.hadoop.hbase.client.Admin;
038import org.apache.hadoop.hbase.client.RegionInfo;
039import org.apache.hadoop.hbase.client.TableDescriptor;
040import org.apache.hadoop.hbase.master.HMaster;
041import org.apache.hadoop.hbase.master.MasterCoprocessorHost;
042import org.apache.hadoop.hbase.testclassification.CoprocessorTests;
043import org.apache.hadoop.hbase.testclassification.MediumTests;
044import org.apache.hadoop.hbase.util.Bytes;
045import org.apache.hadoop.hbase.zookeeper.ZKNodeTracker;
046import org.apache.hadoop.hbase.zookeeper.ZKWatcher;
047import org.junit.AfterClass;
048import org.junit.BeforeClass;
049import org.junit.ClassRule;
050import org.junit.Test;
051import org.junit.experimental.categories.Category;
052
053/**
054 * Tests unhandled exceptions thrown by coprocessors running on master. Expected result is that the
055 * master will abort with an informative error message describing the set of its loaded coprocessors
056 * for crash diagnosis. (HBASE-4014).
057 */
058@Category({ CoprocessorTests.class, MediumTests.class })
059public class TestMasterCoprocessorExceptionWithAbort {
060
061  @ClassRule
062  public static final HBaseClassTestRule CLASS_RULE =
063    HBaseClassTestRule.forClass(TestMasterCoprocessorExceptionWithAbort.class);
064
065  public static class MasterTracker extends ZKNodeTracker {
066    public boolean masterZKNodeWasDeleted = false;
067
068    public MasterTracker(ZKWatcher zkw, String masterNode, Abortable abortable) {
069      super(zkw, masterNode, abortable);
070    }
071
072    @Override
073    public synchronized void nodeDeleted(String path) {
074      if (path.equals("/hbase/master")) {
075        masterZKNodeWasDeleted = true;
076      }
077    }
078  }
079
080  public static class CreateTableThread extends Thread {
081    HBaseTestingUtility UTIL;
082
083    public CreateTableThread(HBaseTestingUtility UTIL) {
084      this.UTIL = UTIL;
085    }
086
087    @Override
088    public void run() {
089      // create a table : master coprocessor will throw an exception and not
090      // catch it.
091      HTableDescriptor htd = new HTableDescriptor(TableName.valueOf(TEST_TABLE));
092      htd.addFamily(new HColumnDescriptor(TEST_FAMILY));
093      try {
094        Admin admin = UTIL.getAdmin();
095        admin.createTable(htd);
096        fail("BuggyMasterObserver failed to throw an exception.");
097      } catch (IOException e) {
098        assertEquals("HBaseAdmin threw an interrupted IOException as expected.",
099          "java.io.InterruptedIOException", e.getClass().getName());
100      }
101    }
102  }
103
104  public static class BuggyMasterObserver implements MasterCoprocessor, MasterObserver {
105    private boolean preCreateTableCalled;
106    private boolean postCreateTableCalled;
107    private boolean startCalled;
108    private boolean postStartMasterCalled;
109
110    @Override
111    public Optional<MasterObserver> getMasterObserver() {
112      return Optional.of(this);
113    }
114
115    @Override
116    public void postCreateTable(ObserverContext<MasterCoprocessorEnvironment> env,
117      TableDescriptor desc, RegionInfo[] regions) throws IOException {
118      // cause a NullPointerException and don't catch it: this will cause the
119      // master to abort().
120      Integer i;
121      i = null;
122      i = i++;
123    }
124
125    public boolean wasCreateTableCalled() {
126      return preCreateTableCalled && postCreateTableCalled;
127    }
128
129    @Override
130    public void postStartMaster(ObserverContext<MasterCoprocessorEnvironment> ctx)
131      throws IOException {
132      postStartMasterCalled = true;
133    }
134
135    public boolean wasStartMasterCalled() {
136      return postStartMasterCalled;
137    }
138
139    @Override
140    public void start(CoprocessorEnvironment env) throws IOException {
141      startCalled = true;
142    }
143
144    public boolean wasStarted() {
145      return startCalled;
146    }
147  }
148
149  private static HBaseTestingUtility UTIL = new HBaseTestingUtility();
150  private static byte[] TEST_TABLE = Bytes.toBytes("observed_table");
151  private static byte[] TEST_FAMILY = Bytes.toBytes("fam1");
152
153  @BeforeClass
154  public static void setupBeforeClass() throws Exception {
155    Configuration conf = UTIL.getConfiguration();
156    conf.set(CoprocessorHost.MASTER_COPROCESSOR_CONF_KEY, BuggyMasterObserver.class.getName());
157    conf.setBoolean(CoprocessorHost.ABORT_ON_ERROR_KEY, true);
158    conf.setInt(HConstants.HBASE_CLIENT_RETRIES_NUMBER, 2); // Fail fast
159    UTIL.startMiniCluster();
160  }
161
162  @AfterClass
163  public static void teardownAfterClass() throws Exception {
164    UTIL.shutdownMiniCluster();
165  }
166
167  @Test
168  public void testExceptionFromCoprocessorWhenCreatingTable() throws IOException {
169    MiniHBaseCluster cluster = UTIL.getHBaseCluster();
170
171    HMaster master = cluster.getMaster();
172    MasterCoprocessorHost host = master.getMasterCoprocessorHost();
173    BuggyMasterObserver cp = host.findCoprocessor(BuggyMasterObserver.class);
174    assertFalse("No table created yet", cp.wasCreateTableCalled());
175
176    // set a watch on the zookeeper /hbase/master node. If the master dies,
177    // the node will be deleted.
178    ZKWatcher zkw = new ZKWatcher(UTIL.getConfiguration(), "unittest", new Abortable() {
179      @Override
180      public void abort(String why, Throwable e) {
181        throw new RuntimeException("Fatal ZK error: " + why, e);
182      }
183
184      @Override
185      public boolean isAborted() {
186        return false;
187      }
188    });
189
190    MasterTracker masterTracker = new MasterTracker(zkw, "/hbase/master", new Abortable() {
191      @Override
192      public void abort(String why, Throwable e) {
193        throw new RuntimeException("Fatal ZK master tracker error, why=", e);
194      }
195
196      @Override
197      public boolean isAborted() {
198        return false;
199      }
200    });
201
202    masterTracker.start();
203    zkw.registerListener(masterTracker);
204
205    // Test (part of the) output that should have be printed by master when it aborts:
206    // (namely the part that shows the set of loaded coprocessors).
207    // In this test, there is only a single coprocessor (BuggyMasterObserver).
208    assertTrue(HMaster.getLoadedCoprocessors()
209      .contains(TestMasterCoprocessorExceptionWithAbort.BuggyMasterObserver.class.getName()));
210
211    CreateTableThread createTableThread = new CreateTableThread(UTIL);
212
213    // Attempting to create a table (using createTableThread above) triggers an NPE in
214    // BuggyMasterObserver.
215    // Master will then abort and the /hbase/master zk node will be deleted.
216    createTableThread.start();
217
218    // Wait up to 30 seconds for master's /hbase/master zk node to go away after master aborts.
219    for (int i = 0; i < 30; i++) {
220      if (masterTracker.masterZKNodeWasDeleted == true) {
221        break;
222      }
223      try {
224        Thread.sleep(1000);
225      } catch (InterruptedException e) {
226        fail("InterruptedException while waiting for master zk node to " + "be deleted.");
227      }
228    }
229
230    assertTrue("Master aborted on coprocessor exception, as expected.",
231      masterTracker.masterZKNodeWasDeleted);
232
233    createTableThread.interrupt();
234    try {
235      createTableThread.join(1000);
236    } catch (InterruptedException e) {
237      assertTrue("Ignoring InterruptedException while waiting for " + " createTableThread.join().",
238        true);
239    }
240  }
241
242}