You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@zookeeper.apache.org by ma...@apache.org on 2011/12/09 20:03:26 UTC
svn commit: r1212578 - in /zookeeper/trunk: ./
src/java/main/org/apache/zookeeper/server/quorum/
src/java/test/org/apache/zookeeper/server/quorum/
src/java/test/org/apache/zookeeper/test/
Author: mahadev
Date: Fri Dec 9 19:03:26 2011
New Revision: 1212578
URL: http://svn.apache.org/viewvc?rev=1212578&view=rev
Log:
ZOOKEEPER-1319. Missing data after restarting+expanding a cluster. (phunt and breed via mahadev)
Modified:
zookeeper/trunk/CHANGES.txt
zookeeper/trunk/src/java/main/org/apache/zookeeper/server/quorum/Leader.java
zookeeper/trunk/src/java/test/org/apache/zookeeper/server/quorum/Zab1_0Test.java
zookeeper/trunk/src/java/test/org/apache/zookeeper/test/FollowerResyncConcurrencyTest.java
Modified: zookeeper/trunk/CHANGES.txt
URL: http://svn.apache.org/viewvc/zookeeper/trunk/CHANGES.txt?rev=1212578&r1=1212577&r2=1212578&view=diff
==============================================================================
--- zookeeper/trunk/CHANGES.txt (original)
+++ zookeeper/trunk/CHANGES.txt Fri Dec 9 19:03:26 2011
@@ -73,6 +73,9 @@ BUGFIXES:
(Akira Kitada via mahadev)
ZOOKEEPER-1317. Possible segfault in zookeeper_init. (Akira Kitada via mahadev)
+
+ ZOOKEEPER-1319. Missing data after restarting+expanding a cluster.
+ (phunt and breed via mahadev)
IMPROVEMENTS:
Modified: zookeeper/trunk/src/java/main/org/apache/zookeeper/server/quorum/Leader.java
URL: http://svn.apache.org/viewvc/zookeeper/trunk/src/java/main/org/apache/zookeeper/server/quorum/Leader.java?rev=1212578&r1=1212577&r2=1212578&view=diff
==============================================================================
--- zookeeper/trunk/src/java/main/org/apache/zookeeper/server/quorum/Leader.java (original)
+++ zookeeper/trunk/src/java/main/org/apache/zookeeper/server/quorum/Leader.java Fri Dec 9 19:03:26 2011
@@ -313,11 +313,9 @@ public class Leader {
zk.setZxid(ZxidUtils.makeZxid(epoch, 0));
- /*
synchronized(this){
lastProposed = zk.getZxid();
}
- */
newLeaderProposal.packet = new QuorumPacket(NEWLEADER, zk.getZxid(),
null, null);
Modified: zookeeper/trunk/src/java/test/org/apache/zookeeper/server/quorum/Zab1_0Test.java
URL: http://svn.apache.org/viewvc/zookeeper/trunk/src/java/test/org/apache/zookeeper/server/quorum/Zab1_0Test.java?rev=1212578&r1=1212577&r2=1212578&view=diff
==============================================================================
--- zookeeper/trunk/src/java/test/org/apache/zookeeper/server/quorum/Zab1_0Test.java (original)
+++ zookeeper/trunk/src/java/test/org/apache/zookeeper/server/quorum/Zab1_0Test.java Fri Dec 9 19:03:26 2011
@@ -18,6 +18,8 @@
package org.apache.zookeeper.server.quorum;
+import static org.junit.Assert.assertEquals;
+
import java.io.ByteArrayOutputStream;
import java.io.File;
import java.io.FileNotFoundException;
@@ -36,8 +38,8 @@ import org.apache.jute.InputArchive;
import org.apache.jute.OutputArchive;
import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;
-import org.apache.zookeeper.ZooDefs;
import org.apache.zookeeper.Watcher.Event.EventType;
+import org.apache.zookeeper.ZooDefs;
import org.apache.zookeeper.data.Stat;
import org.apache.zookeeper.server.ByteBufferInputStream;
import org.apache.zookeeper.server.ByteBufferOutputStream;
@@ -506,7 +508,8 @@ public class Zab1_0Test {
oa.writeRecord(qp, null);
readPacketSkippingPing(ia, qp);
- Assert.assertEquals(Leader.DIFF, qp.getType());
+ Assert.assertEquals(Leader.SNAP, qp.getType());
+ deserializeSnapshot(ia);
readPacketSkippingPing(ia, qp);
Assert.assertEquals(Leader.NEWLEADER, qp.getType());
@@ -518,11 +521,28 @@ public class Zab1_0Test {
oa.writeRecord(qp, null);
readPacketSkippingPing(ia, qp);
+ Assert.assertEquals(Leader.NEWLEADER, qp.getType());
+ Assert.assertEquals(ZxidUtils.makeZxid(1, 0), qp.getZxid());
+ Assert.assertEquals(1, l.self.getAcceptedEpoch());
+ Assert.assertEquals(1, l.self.getCurrentEpoch());
+
+ qp = new QuorumPacket(Leader.ACK, qp.getZxid(), null, null);
+ oa.writeRecord(qp, null);
+
+ readPacketSkippingPing(ia, qp);
Assert.assertEquals(Leader.UPTODATE, qp.getType());
}
});
}
-
+
+ private void deserializeSnapshot(InputArchive ia)
+ throws IOException {
+ ZKDatabase zkdb = new ZKDatabase(null);
+ zkdb.deserializeSnapshot(ia);
+ String signature = ia.readString("signature");
+ assertEquals("BenWasHere", signature);
+ }
+
@Test
public void testLeaderBehind() throws Exception {
testLeaderConversation(new LeaderConversation() {
@@ -545,12 +565,23 @@ public class Zab1_0Test {
qp = new QuorumPacket(Leader.ACKEPOCH, 0, new byte[4], null);
oa.writeRecord(qp, null);
readPacketSkippingPing(ia, qp);
- Assert.assertEquals(Leader.DIFF, qp.getType());
+ Assert.assertEquals(Leader.SNAP, qp.getType());
+ deserializeSnapshot(ia);
+
readPacketSkippingPing(ia, qp);
Assert.assertEquals(Leader.NEWLEADER, qp.getType());
Assert.assertEquals(ZxidUtils.makeZxid(21, 0), qp.getZxid());
+
qp = new QuorumPacket(Leader.ACK, qp.getZxid(), null, null);
oa.writeRecord(qp, null);
+
+ readPacketSkippingPing(ia, qp);
+ Assert.assertEquals(Leader.NEWLEADER, qp.getType());
+ Assert.assertEquals(ZxidUtils.makeZxid(21, 0), qp.getZxid());
+
+ qp = new QuorumPacket(Leader.ACK, qp.getZxid(), null, null);
+ oa.writeRecord(qp, null);
+
readPacketSkippingPing(ia, qp);
Assert.assertEquals(Leader.UPTODATE, qp.getType());
}
Modified: zookeeper/trunk/src/java/test/org/apache/zookeeper/test/FollowerResyncConcurrencyTest.java
URL: http://svn.apache.org/viewvc/zookeeper/trunk/src/java/test/org/apache/zookeeper/test/FollowerResyncConcurrencyTest.java?rev=1212578&r1=1212577&r2=1212578&view=diff
==============================================================================
--- zookeeper/trunk/src/java/test/org/apache/zookeeper/test/FollowerResyncConcurrencyTest.java (original)
+++ zookeeper/trunk/src/java/test/org/apache/zookeeper/test/FollowerResyncConcurrencyTest.java Fri Dec 9 19:03:26 2011
@@ -40,6 +40,7 @@ import org.apache.zookeeper.ZooKeeper;
import org.apache.zookeeper.server.ZKDatabase;
import org.apache.zookeeper.server.quorum.Leader;
import org.apache.zookeeper.test.ClientBase.CountdownWatcher;
+import org.junit.Assert;
import org.junit.Test;
@@ -50,6 +51,83 @@ public class FollowerResyncConcurrencyTe
private volatile int counter = 0;
/**
+ * See ZOOKEEPER-1319 - verify that a lagging follwer resyncs correctly
+ *
+ * 1) start with down quorum
+ * 2) start leader/follower1, add some data
+ * 3) restart leader/follower1
+ * 4) start follower2
+ * 5) verify data consistency across the ensemble
+ *
+ * @throws Exception
+ */
+ @Test
+ public void testLaggingFollowerResyncsUnderNewEpoch() throws Exception {
+ CountdownWatcher watcher1 = new CountdownWatcher();
+ CountdownWatcher watcher2 = new CountdownWatcher();
+ CountdownWatcher watcher3 = new CountdownWatcher();
+
+ QuorumUtil qu = new QuorumUtil(1);
+ qu.shutdownAll();
+
+ qu.start(1);
+ qu.start(2);
+ Assert.assertTrue("Waiting for server up", ClientBase.waitForServerUp("127.0.0.1:"
+ + qu.getPeer(1).clientPort, ClientBase.CONNECTION_TIMEOUT));
+ Assert.assertTrue("Waiting for server up", ClientBase.waitForServerUp("127.0.0.1:"
+ + qu.getPeer(2).clientPort, ClientBase.CONNECTION_TIMEOUT));
+
+ ZooKeeper zk1 =
+ createClient(qu.getPeer(1).peer.getClientPort(), watcher1);
+ LOG.info("zk1 has session id 0x" + Long.toHexString(zk1.getSessionId()));
+
+ final String resyncPath = "/resyncundernewepoch";
+ zk1.create(resyncPath, null, ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
+ zk1.close();
+
+ qu.shutdown(1);
+ qu.shutdown(2);
+ Assert.assertTrue("Waiting for server down", ClientBase.waitForServerDown("127.0.0.1:"
+ + qu.getPeer(1).clientPort, ClientBase.CONNECTION_TIMEOUT));
+ Assert.assertTrue("Waiting for server down", ClientBase.waitForServerDown("127.0.0.1:"
+ + qu.getPeer(2).clientPort, ClientBase.CONNECTION_TIMEOUT));
+
+ qu.start(1);
+ qu.start(2);
+ Assert.assertTrue("Waiting for server up", ClientBase.waitForServerUp("127.0.0.1:"
+ + qu.getPeer(1).clientPort, ClientBase.CONNECTION_TIMEOUT));
+ Assert.assertTrue("Waiting for server up", ClientBase.waitForServerUp("127.0.0.1:"
+ + qu.getPeer(2).clientPort, ClientBase.CONNECTION_TIMEOUT));
+
+ qu.start(3);
+ Assert.assertTrue("Waiting for server up", ClientBase.waitForServerUp("127.0.0.1:"
+ + qu.getPeer(3).clientPort, ClientBase.CONNECTION_TIMEOUT));
+
+ zk1 = createClient(qu.getPeer(1).peer.getClientPort(), watcher1);
+ LOG.info("zk1 has session id 0x" + Long.toHexString(zk1.getSessionId()));
+
+ assertNotNull("zk1 has data", zk1.exists(resyncPath, false));
+
+ final ZooKeeper zk2 =
+ createClient(qu.getPeer(2).peer.getClientPort(), watcher2);
+ LOG.info("zk2 has session id 0x" + Long.toHexString(zk2.getSessionId()));
+
+ assertNotNull("zk2 has data", zk2.exists(resyncPath, false));
+
+ final ZooKeeper zk3 =
+ createClient(qu.getPeer(3).peer.getClientPort(), watcher3);
+ LOG.info("zk3 has session id 0x" + Long.toHexString(zk3.getSessionId()));
+
+ assertNotNull("zk3 has data", zk3.exists(resyncPath, false));
+
+ zk1.close();
+ zk2.close();
+ zk3.close();
+
+ qu.shutdownAll();
+ }
+
+ /**
* See ZOOKEEPER-962. This tests for one of the bugs hit while fixing this,
* setting the ZXID of the SNAP packet
* Starts up 3 ZKs. Shut down F1, write a node, restart the one that was shut down