[GitHub] [hbase] sandeepvinayak commented on a change in pull request #2960: HBASE-25582: Handle the running replication source gracefully with replication nodes deleted

[GitBox <gi...@apache.org>]

sandeepvinayak commented on a change in pull request #2960:
URL: https://github.com/apache/hbase/pull/2960#discussion_r578612002



##########
File path: hbase-server/src/test/java/org/apache/hadoop/hbase/replication/regionserver/TestReplicationSourceWithoutReplicationZnodes.java
##########
@@ -0,0 +1,234 @@
+/*
+ *
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.hbase.replication.regionserver;
+
+import static org.apache.hadoop.hbase.replication.ReplicationSourceDummy.fakeExceptionMessage;
+
+import java.io.IOException;
+import java.net.URLEncoder;
+import java.util.ArrayList;
+import java.util.List;
+
+import org.apache.commons.logging.Log;
+import org.apache.commons.logging.LogFactory;
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.FileSystem;
+import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hbase.ClusterId;
+import org.apache.hadoop.hbase.HBaseConfiguration;
+import org.apache.hadoop.hbase.HBaseTestingUtility;
+import org.apache.hadoop.hbase.HColumnDescriptor;
+import org.apache.hadoop.hbase.HConstants;
+import org.apache.hadoop.hbase.HRegionInfo;
+import org.apache.hadoop.hbase.HTableDescriptor;
+import org.apache.hadoop.hbase.KeyValue;
+import org.apache.hadoop.hbase.TableName;
+import org.apache.hadoop.hbase.regionserver.MultiVersionConcurrencyControl;
+import org.apache.hadoop.hbase.regionserver.wal.WALActionsListener;
+import org.apache.hadoop.hbase.regionserver.wal.WALEdit;
+import org.apache.hadoop.hbase.replication.ReplicationException;
+import org.apache.hadoop.hbase.replication.ReplicationSourceDummyWithNoTermination;
+import org.apache.hadoop.hbase.replication.ReplicationSourceWithoutPeerException;
+import org.apache.hadoop.hbase.replication.ReplicationStateZKBase;
+import org.apache.hadoop.hbase.replication.regionserver.helper.DummyServer;
+import org.apache.hadoop.hbase.testclassification.MediumTests;
+import org.apache.hadoop.hbase.util.Bytes;
+import org.apache.hadoop.hbase.util.FSUtils;
+import org.apache.hadoop.hbase.wal.WAL;
+import org.apache.hadoop.hbase.wal.WALFactory;
+import org.apache.hadoop.hbase.wal.WALKey;
+import org.apache.hadoop.hbase.zookeeper.ZKClusterId;
+import org.apache.hadoop.hbase.zookeeper.ZKUtil;
+import org.apache.hadoop.hbase.zookeeper.ZooKeeperWatcher;
+import org.junit.After;
+import org.junit.AfterClass;
+import org.junit.Assert;
+import org.junit.Before;
+import org.junit.BeforeClass;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.experimental.categories.Category;
+import org.junit.rules.TestName;
+
+@Category(MediumTests.class)
+public class TestReplicationSourceWithoutReplicationZnodes {
+
+  private static final Log LOG =
+    LogFactory.getLog(TestReplicationSourceManager.class);
+
+  private static Configuration conf;
+
+  private static HBaseTestingUtility utility;
+
+  private static Replication replication;
+
+  private static ReplicationSourceManager manager;
+
+  private static ZooKeeperWatcher zkw;
+
+  private static HTableDescriptor htd;
+
+  private static HRegionInfo hri;
+
+  private static final byte[] r1 = Bytes.toBytes("r1");
+
+  private static final byte[] r2 = Bytes.toBytes("r2");
+
+  private static final byte[] f1 = Bytes.toBytes("f1");
+
+  private static final byte[] f2 = Bytes.toBytes("f2");
+
+  private static final TableName test =
+    TableName.valueOf("test");
+
+  private static final String slaveId = "1";
+
+  private static FileSystem fs;
+
+  private static Path oldLogDir;
+
+  private static Path logDir;
+
+  private static DummyServer server;
+
+  @BeforeClass
+  public static void setUpBeforeClass() throws Exception {
+
+    conf = HBaseConfiguration.create();
+    conf.set("replication.replicationsource.implementation",
+      ReplicationSourceDummyWithNoTermination.class.getCanonicalName());
+    conf.setBoolean(HConstants.REPLICATION_ENABLE_KEY,
+      HConstants.REPLICATION_ENABLE_DEFAULT);
+    conf.setLong("replication.sleep.before.failover", 2000);
+    conf.setInt("replication.source.maxretriesmultiplier", 10);
+    utility = new HBaseTestingUtility(conf);
+    utility.startMiniZKCluster();
+
+    zkw = new ZooKeeperWatcher(conf, "test", null);
+    ZKUtil.createWithParents(zkw, "/hbase/replication");
+    ZKUtil.createWithParents(zkw, "/hbase/replication/peers/1");
+    ZKUtil.setData(zkw, "/hbase/replication/peers/1",
+      Bytes.toBytes(conf.get(HConstants.ZOOKEEPER_QUORUM) + ":"
+        + conf.get(HConstants.ZOOKEEPER_CLIENT_PORT) + ":/1"));
+    ZKUtil.createWithParents(zkw, "/hbase/replication/peers/1/peer-state");
+    ZKUtil.setData(zkw, "/hbase/replication/peers/1/peer-state",
+      ReplicationStateZKBase.ENABLED_ZNODE_BYTES);
+    ZKUtil.createWithParents(zkw, "/hbase/replication/state");
+    ZKUtil.setData(zkw, "/hbase/replication/state", ReplicationStateZKBase.ENABLED_ZNODE_BYTES);
+
+    ZKClusterId.setClusterId(zkw, new ClusterId());
+    FSUtils.setRootDir(utility.getConfiguration(), utility.getDataTestDir());
+    fs = FileSystem.get(conf);
+    oldLogDir = new Path(utility.getDataTestDir(),
+      HConstants.HREGION_OLDLOGDIR_NAME);
+    logDir = new Path(utility.getDataTestDir(),
+      HConstants.HREGION_LOGDIR_NAME);
+    server = new DummyServer(conf, "example.hostname.com", zkw);
+    replication = new Replication(server, fs, logDir, oldLogDir);
+    manager = replication.getReplicationManager();
+
+    manager.addSource(slaveId);
+
+    htd = new HTableDescriptor(test);
+    HColumnDescriptor col = new HColumnDescriptor(f1);
+    col.setScope(HConstants.REPLICATION_SCOPE_GLOBAL);
+    htd.addFamily(col);
+    col = new HColumnDescriptor(f2);
+    col.setScope(HConstants.REPLICATION_SCOPE_LOCAL);
+    htd.addFamily(col);
+
+    hri = new HRegionInfo(htd.getTableName(), r1, r2);
+  }
+
+  @AfterClass
+  public static void tearDownAfterClass() throws Exception {
+    try {
+      manager.join();
+    } catch (RuntimeException re) {
+      if (re.getMessage().equals(fakeExceptionMessage)) {
+        LOG.info("It is fine");
+      }
+    }
+    utility.shutdownMiniCluster();
+  }
+
+  @Rule
+  public TestName testName = new TestName();
+
+  private void cleanLogDir() throws IOException {
+    fs.delete(logDir, true);
+    fs.delete(oldLogDir, true);
+  }
+
+  @Before
+  public void setUp() throws Exception {
+    LOG.info("Start " + testName.getMethodName());
+    cleanLogDir();
+  }
+
+  @After
+  public void tearDown() throws Exception {
+    LOG.info("End " + testName.getMethodName());
+    cleanLogDir();
+  }
+
+  /**
+   * When the peer is removed, hbase remove the peer znodes and there is zk watcher
+   * which terminates the replication sources. In case of zk watcher not getting invoked
+   * or a race condition between source deleting the log znode and zk watcher
+   * terminating the source, we might get the NoNode exception. In that case, the right
+   * thing is to terminate the replication source.

Review comment:
       >The only place I see updating log position (eventually calling ReplicationQueuesZKImpl.removeLog) is the SourceShipperThread, which is terminated in ReplicationSouceManager.removePeer.
   
   You are right, but the `ReplicationSouceManager.removePeer` is invoked by a `PeersWatcher` zk listerner.
   So, this means there is a race condition between `ReplicationQueuesZKImpl.removeLog` and `ReplicationSouceManager.removePeer` which terminates the source, even though race condition won't impact multiple region servers at once. 
   But if the zk listerner is not invoked at all due to some zk issue, this will effect all the region servers which are replicating and will badly affect the availability. We have seen this issue recently and it seems like ZK issue. But handling the NoNode exception gracefully will make RS resilient to this issue. 




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org