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.rsgroup;
019
020import static org.junit.Assert.assertEquals;
021import static org.junit.Assert.assertFalse;
022
023import org.apache.hadoop.hbase.HBaseClassTestRule;
024import org.apache.hadoop.hbase.HBaseClusterInterface;
025import org.apache.hadoop.hbase.HBaseTestingUtil;
026import org.apache.hadoop.hbase.SingleProcessHBaseCluster;
027import org.apache.hadoop.hbase.StartTestingClusterOption;
028import org.apache.hadoop.hbase.TableName;
029import org.apache.hadoop.hbase.Waiter;
030import org.apache.hadoop.hbase.client.Admin;
031import org.apache.hadoop.hbase.client.RegionInfo;
032import org.apache.hadoop.hbase.master.HMaster;
033import org.apache.hadoop.hbase.master.ServerManager;
034import org.apache.hadoop.hbase.regionserver.HRegionServer;
035import org.apache.hadoop.hbase.testclassification.MediumTests;
036import org.apache.hadoop.hbase.testclassification.RSGroupTests;
037import org.apache.hadoop.hbase.util.Bytes;
038import org.junit.AfterClass;
039import org.junit.BeforeClass;
040import org.junit.ClassRule;
041import org.junit.Rule;
042import org.junit.Test;
043import org.junit.experimental.categories.Category;
044import org.junit.rules.TestName;
045import org.slf4j.Logger;
046import org.slf4j.LoggerFactory;
047
048import org.apache.hbase.thirdparty.com.google.common.collect.Sets;
049
050/**
051 * This tests that GroupBasedBalancer will use data in zk to do balancing during master startup.
052 * This does not test retain assignment.
053 * <p/>
054 * The tests brings up 3 RS, creates a new RS group 'my_group', moves 1 RS to 'my_group', assigns
055 * 'hbase:rsgroup' to 'my_group', and kill the only server in that group so that 'hbase:rsgroup'
056 * table isn't available. It then kills the active master and waits for backup master to come
057 * online. In new master, RSGroupInfoManagerImpl gets the data from zk and waits for the expected
058 * assignment with a timeout.
059 */
060@Category({ RSGroupTests.class, MediumTests.class })
061public class TestRSGroupsOfflineMode extends TestRSGroupsBase {
062
063  @ClassRule
064  public static final HBaseClassTestRule CLASS_RULE =
065    HBaseClassTestRule.forClass(TestRSGroupsOfflineMode.class);
066
067  private static final Logger LOG = LoggerFactory.getLogger(TestRSGroupsOfflineMode.class);
068  private static HMaster master;
069  private static Admin hbaseAdmin;
070  private static HBaseTestingUtil TEST_UTIL;
071  private static HBaseClusterInterface cluster;
072  private final static long WAIT_TIMEOUT = 60000 * 5;
073
074  @Rule
075  public TestName name = new TestName();
076
077  @BeforeClass
078  public static void setUp() throws Exception {
079    TEST_UTIL = new HBaseTestingUtil();
080    RSGroupUtil.enableRSGroup(TEST_UTIL.getConfiguration());
081    TEST_UTIL.getConfiguration().set(ServerManager.WAIT_ON_REGIONSERVERS_MINTOSTART, "1");
082    StartTestingClusterOption option =
083      StartTestingClusterOption.builder().numMasters(2).numRegionServers(3).numDataNodes(3).build();
084    TEST_UTIL.startMiniCluster(option);
085    cluster = TEST_UTIL.getHBaseCluster();
086    master = ((SingleProcessHBaseCluster) cluster).getMaster();
087    master.balanceSwitch(false);
088    hbaseAdmin = TEST_UTIL.getAdmin();
089    // wait till the balancer is in online mode
090    TEST_UTIL.waitFor(WAIT_TIMEOUT, new Waiter.Predicate<Exception>() {
091      @Override
092      public boolean evaluate() throws Exception {
093        return master.isInitialized()
094          && ((RSGroupBasedLoadBalancer) master.getLoadBalancer()).isOnline()
095          && master.getServerManager().getOnlineServersList().size() >= 3;
096      }
097    });
098  }
099
100  @AfterClass
101  public static void tearDown() throws Exception {
102    TEST_UTIL.shutdownMiniCluster();
103  }
104
105  @Test
106  public void testOffline() throws Exception, InterruptedException {
107    // Table should be after group table name so it gets assigned later.
108    final TableName failoverTable = TableName.valueOf(getNameWithoutIndex(name.getMethodName()));
109    TEST_UTIL.createTable(failoverTable, Bytes.toBytes("f"));
110    final HRegionServer killRS = ((SingleProcessHBaseCluster) cluster).getRegionServer(0);
111    final HRegionServer groupRS = ((SingleProcessHBaseCluster) cluster).getRegionServer(1);
112    final HRegionServer failoverRS = ((SingleProcessHBaseCluster) cluster).getRegionServer(2);
113    String newGroup = "my_group";
114    Admin admin = TEST_UTIL.getAdmin();
115    admin.addRSGroup(newGroup);
116    if (
117      master.getAssignmentManager().getRegionStates().getRegionAssignments()
118        .containsValue(failoverRS.getServerName())
119    ) {
120      for (RegionInfo regionInfo : hbaseAdmin.getRegions(failoverRS.getServerName())) {
121        hbaseAdmin.move(regionInfo.getEncodedNameAsBytes(), failoverRS.getServerName());
122      }
123      LOG.info("Waiting for region unassignments on failover RS...");
124      TEST_UTIL.waitFor(WAIT_TIMEOUT, new Waiter.Predicate<Exception>() {
125        @Override
126        public boolean evaluate() throws Exception {
127          return !master.getServerManager().getLoad(failoverRS.getServerName()).getRegionMetrics()
128            .isEmpty();
129        }
130      });
131    }
132
133    // Move server to group and make sure all tables are assigned.
134    admin.moveServersToRSGroup(Sets.newHashSet(groupRS.getServerName().getAddress()), newGroup);
135    TEST_UTIL.waitFor(WAIT_TIMEOUT, new Waiter.Predicate<Exception>() {
136      @Override
137      public boolean evaluate() throws Exception {
138        return groupRS.getNumberOfOnlineRegions() < 1
139          && master.getAssignmentManager().getRegionStates().getRegionsInTransitionCount() < 1;
140      }
141    });
142    // Move table to group and wait.
143    admin.setRSGroup(Sets.newHashSet(RSGroupInfoManagerImpl.RSGROUP_TABLE_NAME), newGroup);
144    LOG.info("Waiting for move table...");
145    TEST_UTIL.waitFor(WAIT_TIMEOUT, new Waiter.Predicate<Exception>() {
146      @Override
147      public boolean evaluate() throws Exception {
148        return groupRS.getNumberOfOnlineRegions() == 1;
149      }
150    });
151
152    groupRS.stop("die");
153    // Race condition here.
154    TEST_UTIL.getHBaseCluster().getMaster().stopMaster();
155    LOG.info("Waiting for offline mode...");
156    TEST_UTIL.waitFor(WAIT_TIMEOUT, new Waiter.Predicate<Exception>() {
157      @Override
158      public boolean evaluate() throws Exception {
159        return TEST_UTIL.getHBaseCluster().getMaster() != null
160          && TEST_UTIL.getHBaseCluster().getMaster().isActiveMaster()
161          && TEST_UTIL.getHBaseCluster().getMaster().isInitialized()
162          && TEST_UTIL.getHBaseCluster().getMaster().getServerManager().getOnlineServers().size()
163              <= 3;
164      }
165    });
166
167    // Get groupInfoManager from the new active master.
168    RSGroupInfoManager groupMgr =
169      ((SingleProcessHBaseCluster) cluster).getMaster().getRSGroupInfoManager();
170    // Make sure balancer is in offline mode, since this is what we're testing.
171    assertFalse(groupMgr.isOnline());
172    // Kill final regionserver to see the failover happens for all tables except GROUP table since
173    // it's group does not have any online RS.
174    killRS.stop("die");
175    master = TEST_UTIL.getHBaseCluster().getMaster();
176    LOG.info("Waiting for new table assignment...");
177    TEST_UTIL.waitFor(WAIT_TIMEOUT, new Waiter.Predicate<Exception>() {
178      @Override
179      public boolean evaluate() throws Exception {
180        return failoverRS.getRegions(failoverTable).size() >= 1;
181      }
182    });
183    assertEquals(0, failoverRS.getRegions(RSGroupInfoManagerImpl.RSGROUP_TABLE_NAME).size());
184
185    // Need this for minicluster to shutdown cleanly.
186    master.stopMaster();
187  }
188}