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.
055 * Expected result is that the master will abort with an informative
056 * error message describing the set of its loaded coprocessors for crash diagnosis.
057 * (HBASE-4014).
058 */
059@Category({CoprocessorTests.class, MediumTests.class})
060public class TestMasterCoprocessorExceptionWithAbort {
061
062  @ClassRule
063  public static final HBaseClassTestRule CLASS_RULE =
064      HBaseClassTestRule.forClass(TestMasterCoprocessorExceptionWithAbort.class);
065
066  public static class MasterTracker extends ZKNodeTracker {
067    public boolean masterZKNodeWasDeleted = false;
068
069    public MasterTracker(ZKWatcher zkw, String masterNode, Abortable abortable) {
070      super(zkw, masterNode, abortable);
071    }
072
073    @Override
074    public synchronized void nodeDeleted(String path) {
075      if (path.equals("/hbase/master")) {
076        masterZKNodeWasDeleted = true;
077      }
078    }
079  }
080
081  public static class CreateTableThread extends Thread {
082    HBaseTestingUtility UTIL;
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,
157        BuggyMasterObserver.class.getName());
158    conf.setBoolean(CoprocessorHost.ABORT_ON_ERROR_KEY, true);
159    conf.setInt(HConstants.HBASE_CLIENT_RETRIES_NUMBER, 2);  // Fail fast
160    UTIL.startMiniCluster();
161  }
162
163  @AfterClass
164  public static void teardownAfterClass() throws Exception {
165    UTIL.shutdownMiniCluster();
166  }
167
168  @Test
169  public void testExceptionFromCoprocessorWhenCreatingTable()
170      throws IOException {
171    MiniHBaseCluster cluster = UTIL.getHBaseCluster();
172
173    HMaster master = cluster.getMaster();
174    MasterCoprocessorHost host = master.getMasterCoprocessorHost();
175    BuggyMasterObserver cp = host.findCoprocessor(BuggyMasterObserver.class);
176    assertFalse("No table created yet", cp.wasCreateTableCalled());
177
178    // set a watch on the zookeeper /hbase/master node. If the master dies,
179    // the node will be deleted.
180    ZKWatcher zkw = new ZKWatcher(UTIL.getConfiguration(),
181      "unittest", new Abortable() {
182      @Override
183      public void abort(String why, Throwable e) {
184        throw new RuntimeException("Fatal ZK error: " + why, e);
185      }
186      @Override
187      public boolean isAborted() {
188        return false;
189      }
190    });
191
192    MasterTracker masterTracker = new MasterTracker(zkw,"/hbase/master",
193        new Abortable() {
194          @Override
195          public void abort(String why, Throwable e) {
196            throw new RuntimeException("Fatal ZK master tracker error, why=", e);
197          }
198          @Override
199          public boolean isAborted() {
200            return false;
201          }
202        });
203
204    masterTracker.start();
205    zkw.registerListener(masterTracker);
206
207    // Test (part of the) output that should have be printed by master when it aborts:
208    // (namely the part that shows the set of loaded coprocessors).
209    // In this test, there is only a single coprocessor (BuggyMasterObserver).
210    assertTrue(HMaster.getLoadedCoprocessors().
211      contains(TestMasterCoprocessorExceptionWithAbort.BuggyMasterObserver.class.getName()));
212
213    CreateTableThread createTableThread = new CreateTableThread(UTIL);
214
215    // Attempting to create a table (using createTableThread above) triggers an NPE in BuggyMasterObserver.
216    // Master will then abort and the /hbase/master zk node will be deleted.
217    createTableThread.start();
218
219    // Wait up to 30 seconds for master's /hbase/master zk node to go away after master aborts.
220    for (int i = 0; i < 30; i++) {
221      if (masterTracker.masterZKNodeWasDeleted == true) {
222        break;
223      }
224      try {
225        Thread.sleep(1000);
226      } catch (InterruptedException e) {
227        fail("InterruptedException while waiting for master zk node to "
228            + "be deleted.");
229      }
230    }
231
232    assertTrue("Master aborted on coprocessor exception, as expected.",
233        masterTracker.masterZKNodeWasDeleted);
234
235    createTableThread.interrupt();
236    try {
237      createTableThread.join(1000);
238    } catch (InterruptedException e) {
239      assertTrue("Ignoring InterruptedException while waiting for " +
240          " createTableThread.join().", true);
241    }
242  }
243
244}
245