From cf172163722648f42a8c5f61496ee244ec211662 Mon Sep 17 00:00:00 2001 From: Pankaj Kumar Date: Fri, 4 Mar 2016 15:06:17 +0800 Subject: [PATCH] HBASE-14498, Master stuck in infinite loop when all Zookeeper servers are unreachable. Change-Id: I5cf54529803721d093d54d9b105bce8dfe96c78d --- .../hadoop/hbase/zookeeper/ZooKeeperWatcher.java | 86 +++++++++++++++++++++- .../hbase/zookeeper/TestZooKeeperWatcher.java | 47 ++++++++++++ .../hadoop/hbase/regionserver/HRegionServer.java | 2 +- 3 files changed, 132 insertions(+), 3 deletions(-) diff --git a/hbase-client/src/main/java/org/apache/hadoop/hbase/zookeeper/ZooKeeperWatcher.java b/hbase-client/src/main/java/org/apache/hadoop/hbase/zookeeper/ZooKeeperWatcher.java index b665353..64750da 100644 --- a/hbase-client/src/main/java/org/apache/hadoop/hbase/zookeeper/ZooKeeperWatcher.java +++ b/hbase-client/src/main/java/org/apache/hadoop/hbase/zookeeper/ZooKeeperWatcher.java @@ -26,6 +26,9 @@ import java.util.List; import java.util.Map; import java.util.concurrent.CopyOnWriteArrayList; import java.util.concurrent.CountDownLatch; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; +import java.util.concurrent.atomic.AtomicBoolean; import java.util.regex.Matcher; import java.util.regex.Pattern; @@ -39,6 +42,7 @@ import org.apache.hadoop.hbase.HRegionInfo; import org.apache.hadoop.hbase.ZooKeeperConnectionException; import org.apache.hadoop.hbase.classification.InterfaceAudience; import org.apache.hadoop.hbase.security.Superusers; +import org.apache.hadoop.hbase.util.EnvironmentEdgeManager; import org.apache.hadoop.security.UserGroupInformation; import org.apache.zookeeper.KeeperException; import org.apache.zookeeper.WatchedEvent; @@ -76,7 +80,7 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { private RecoverableZooKeeper recoverableZooKeeper; // abortable in case of zk failure - protected Abortable abortable; + protected final Abortable abortable; // Used if abortable is null private boolean aborted = false; @@ -88,6 +92,14 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { // negotiation to complete public CountDownLatch saslLatch = new CountDownLatch(1); + // Connection timeout on disconnect event + private long connWaitTimeOut; + private AtomicBoolean isConnected = new AtomicBoolean(false); + private boolean forceAbortOnZKDisconnect; + + // Execute service for zookeeper disconnect event watcher + private ExecutorService zkEventWatcherExecService = null; + // node names // base znode for this cluster @@ -159,6 +171,24 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { public ZooKeeperWatcher(Configuration conf, String identifier, Abortable abortable, boolean canCreateBaseZNode) throws IOException, ZooKeeperConnectionException { + this(conf, identifier, abortable, canCreateBaseZNode, false); + } + + /** + * Instantiate a ZooKeeper connection and watcher. + * @param conf Configuration + * @param identifier string that is passed to RecoverableZookeeper to be used as identifier for + * this instance. Use null for default. + * @param abortable Can be null if there is on error there is no host to abort: e.g. client + * context. + * @param canCreateBaseZNode whether create base node. + * @param forceAbortOnZKDisconnect abort the watcher if true. + * @throws IOException when any IO exception + * @throws ZooKeeperConnectionException when any zookeeper connection exception + */ + public ZooKeeperWatcher(Configuration conf, String identifier, Abortable abortable, + boolean canCreateBaseZNode, boolean forceAbortOnZKDisconnect) throws IOException, + ZooKeeperConnectionException { this.conf = conf; // Capture a stack trace now. Will print it out later if problem so we can // distingush amongst the myriad ZKWs. @@ -174,10 +204,17 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { this.identifier = identifier + "0x0"; this.abortable = abortable; setNodeNames(conf); + // On Disconnected event a thread will wait for sometime (2/3 of zookeeper.session.timeout), + // it will abort the process if no SyncConnected event reported by the time. + connWaitTimeOut = this.conf.getLong("zookeeper.session.timeout", 90000) * 2 / 3; this.recoverableZooKeeper = ZKUtil.connect(conf, quorum, this, identifier); if (canCreateBaseZNode) { createBaseZNodes(); } + this.forceAbortOnZKDisconnect = forceAbortOnZKDisconnect; + if (this.forceAbortOnZKDisconnect) { + this.zkEventWatcherExecService = Executors.newSingleThreadExecutor(); + } } private void createBaseZNodes() throws ZooKeeperConnectionException { @@ -647,6 +684,7 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { private void connectionEvent(WatchedEvent event) { switch(event.getState()) { case SyncConnected: + isConnected.set(true); // Now, this callback can be invoked before the this.zookeeper is set. // Wait a little while. long finished = System.currentTimeMillis() + @@ -676,7 +714,14 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { // Abort the server if Disconnected or Expired case Disconnected: - LOG.debug(prefix("Received Disconnected from ZooKeeper, ignoring")); + LOG.debug(prefix("Received Disconnected from ZooKeeper.")); + if (forceAbortOnZKDisconnect) { + isConnected.set(false); + ZKDisconnectEventWatcher task = new ZKDisconnectEventWatcher(); + zkEventWatcherExecService.execute(task); + } else { + LOG.debug("Received Disconnected from ZooKeeper, ignoring."); + } break; case Expired: @@ -699,6 +744,39 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { } } + /* + * Task to watch zookeper disconnect event. + */ + class ZKDisconnectEventWatcher implements Runnable { + @Override + public void run() { + if (isConnected.get()) { + return; + } + + long startTime = EnvironmentEdgeManager.currentTime(); + while (EnvironmentEdgeManager.currentTime() - startTime < connWaitTimeOut) { + if (isConnected.get()) { + LOG.debug("Client got reconnected to zookeeper."); + return; + } + try { + Thread.sleep(100); + } catch (InterruptedException e) { + Thread.currentThread().interrupt(); + break; + } + } + + if (!isConnected.get() && abortable != null) { + String msg = + prefix("Couldn't connect to ZooKeeper after waiting " + connWaitTimeOut + + " ms, aborting"); + abortable.abort(msg, new KeeperException.ConnectionLossException()); + } + } + } + /** * Forces a synchronization of this ZooKeeper client connection. *

@@ -759,6 +837,10 @@ public class ZooKeeperWatcher implements Watcher, Abortable, Closeable { if (recoverableZooKeeper != null) { recoverableZooKeeper.close(); } + if (zkEventWatcherExecService != null) { + zkEventWatcherExecService.shutdown(); + zkEventWatcherExecService = null; + } } catch (InterruptedException e) { Thread.currentThread().interrupt(); } diff --git a/hbase-client/src/test/java/org/apache/hadoop/hbase/zookeeper/TestZooKeeperWatcher.java b/hbase-client/src/test/java/org/apache/hadoop/hbase/zookeeper/TestZooKeeperWatcher.java index 10a3816..02b92d0 100644 --- a/hbase-client/src/test/java/org/apache/hadoop/hbase/zookeeper/TestZooKeeperWatcher.java +++ b/hbase-client/src/test/java/org/apache/hadoop/hbase/zookeeper/TestZooKeeperWatcher.java @@ -22,14 +22,22 @@ import static org.junit.Assert.*; import java.io.IOException; +import org.apache.commons.logging.Log; +import org.apache.commons.logging.LogFactory; +import org.apache.hadoop.conf.Configuration; +import org.apache.hadoop.hbase.Abortable; import org.apache.hadoop.hbase.HBaseConfiguration; import org.apache.hadoop.hbase.ZooKeeperConnectionException; import org.apache.hadoop.hbase.testclassification.SmallTests; +import org.apache.hadoop.hbase.util.EnvironmentEdgeManager; +import org.apache.zookeeper.WatchedEvent; +import org.apache.zookeeper.Watcher; import org.junit.Test; import org.junit.experimental.categories.Category; @Category({SmallTests.class}) public class TestZooKeeperWatcher { + private final static Log LOG = LogFactory.getLog(TestZooKeeperWatcher.class); @Test public void testIsClientReadable() throws ZooKeeperConnectionException, IOException { @@ -57,4 +65,43 @@ public class TestZooKeeperWatcher { watcher.close(); } + @Test + public void testConnectionEvent() throws ZooKeeperConnectionException, IOException { + long zkSessionTimeout = 15000l; + Configuration conf = HBaseConfiguration.create(); + conf.set("zookeeper.session.timeout", "15000"); + + Abortable abortable = new Abortable() { + boolean aborted = false; + + @Override + public void abort(String why, Throwable e) { + aborted = true; + LOG.error(why, e); + } + + @Override + public boolean isAborted() { + return aborted; + } + }; + ZooKeeperWatcher watcher = new ZooKeeperWatcher(conf, "testConnectionEvent", abortable, false, true); + + WatchedEvent event = + new WatchedEvent(Watcher.Event.EventType.None, Watcher.Event.KeeperState.Disconnected, null); + + long startTime = EnvironmentEdgeManager.currentTime(); + while (!abortable.isAborted() + && (EnvironmentEdgeManager.currentTime() - startTime < zkSessionTimeout)) { + watcher.process(event); + try { + Thread.sleep(1000); + } catch (InterruptedException e) { + Thread.currentThread().interrupt(); + } + } + + assertTrue(abortable.isAborted()); + watcher.close(); + } } diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java index 4ab2693..b46951b 100644 --- a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java +++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java @@ -595,7 +595,7 @@ public class HRegionServer extends HasThread implements if (!conf.getBoolean("hbase.testing.nocluster", false)) { // Open connection to zookeeper and set primary watcher zooKeeper = new ZooKeeperWatcher(conf, getProcessName() + ":" + - rpcServices.isa.getPort(), this, canCreateBaseZNode()); + rpcServices.isa.getPort(), this, canCreateBaseZNode(), true); this.csm = (BaseCoordinatedStateManager) csm; this.csm.initialize(this); -- 2.6.1.windows.1