[
https://issues.apache.org/jira/browse/ZOOKEEPER-2988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16384210#comment-16384210
]
ASF GitHub Bot commented on ZOOKEEPER-2988:
-------------------------------------------
Github user afine commented on a diff in the pull request:
https://github.com/apache/zookeeper/pull/476#discussion_r171972608
--- Diff:
src/java/test/org/apache/zookeeper/server/quorum/QuorumPeerMainTest.java ---
@@ -1012,4 +1012,113 @@ public void testFailedTxnAsPartOfQuorumLoss()
throws Exception {
Assert.assertNull("server " + i + " should not have /zk" +
leader, servers.zk[i].exists("/zk" + leader, false));
}
}
+
+ /**
+ * Verify that a node without the leader in its view will not attempt
to connect to the leader.
+ */
+ @Test
+ public void testLeaderOutOfView() throws Exception {
+ ClientBase.setupTestEnv();
+
+ Layout layout = new PatternLayout("%d{ISO8601} [,yid:%X{myid}] -
%5p [%t:%C{1}@%L] - %m%n");
+ ByteArrayOutputStream os = new ByteArrayOutputStream();
+ WriterAppender appender = new WriterAppender(layout, os);
+ appender.setThreshold(Level.DEBUG);
+ Logger qlogger =
Logger.getLogger("org.apache.zookeeper.server.quorum");
+ qlogger.addAppender(appender);
+
+ try {
+ final int CLIENT_PORT_QP1 = PortAssignment.unique();
+ final int CLIENT_PORT_QP2 = PortAssignment.unique();
+ final int CLIENT_PORT_QP3 = PortAssignment.unique();
+
+ String quorumCfgIncomplete = getUniquePortCfgForId(1) + "\n" +
getUniquePortCfgForId(2);
+ String quorumCfgComplete = quorumCfgIncomplete + "\n" +
getUniquePortCfgForId(3);
+
+ // Node 1 is started without the leader (3) in its config view
+ MainThread q1 = new MainThread(1, CLIENT_PORT_QP1,
quorumCfgIncomplete);
+ MainThread q2 = new MainThread(2, CLIENT_PORT_QP2,
quorumCfgComplete);
+ MainThread q3 = new MainThread(3, CLIENT_PORT_QP3,
quorumCfgComplete);
+
+ // Node 1 must be started first, before quorum is formed, to
trigger the attempted invalid connection to 3
+ q1.start();
+ QuorumPeer quorumPeer1 = waitForQuorumPeer(q1,
CONNECTION_TIMEOUT);
+ Assert.assertTrue(quorumPeer1.getPeerState() ==
QuorumPeer.ServerState.LOOKING);
+
+ // Node 3 started second to avoid 1 and 2 forming a quorum
before 3 starts up
+ q3.start();
+ QuorumPeer quorumPeer3 = waitForQuorumPeer(q3,
CONNECTION_TIMEOUT);
+ Assert.assertTrue(quorumPeer3.getPeerState() ==
QuorumPeer.ServerState.LOOKING);
+
+ // Node 2 started last, kicks off leader election
+ q2.start();
+
+ // Nodes 2 and 3 now form quorum and fully start. 1 attempts
to vote for 3, fails, returns to LOOKING state
+ Assert.assertTrue("waiting for server 2 to start",
+ ClientBase.waitForServerUp("127.0.0.1:" +
CLIENT_PORT_QP2, CONNECTION_TIMEOUT));
+ Assert.assertTrue("waiting for server 3 to start",
+ ClientBase.waitForServerUp("127.0.0.1:" +
CLIENT_PORT_QP3, CONNECTION_TIMEOUT));
+
+ Assert.assertTrue(q1.getQuorumPeer().getPeerState() ==
QuorumPeer.ServerState.LOOKING);
+ Assert.assertTrue(q2.getQuorumPeer().getPeerState() ==
QuorumPeer.ServerState.FOLLOWING);
+ Assert.assertTrue(q3.getQuorumPeer().getPeerState() ==
QuorumPeer.ServerState.LEADING);
+
+ q1.shutdown();
--- End diff --
is there a way we can use the existing tearDown code. So we make sure we
never leave any servers running even if an assertion fails?
> NPE triggered if server receives a vote for a server id not in their voting
> view
> --------------------------------------------------------------------------------
>
> Key: ZOOKEEPER-2988
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2988
> Project: ZooKeeper
> Issue Type: Bug
> Components: leaderElection
> Affects Versions: 3.5.3, 3.4.11
> Reporter: Brian Nixon
> Priority: Minor
>
> We've observed the following behavior in elections when a node is lagging
> behind the quorum in its view of the ensemble topology.
> - Node A is operating with node B in its voting view, but without view of
> node C.
> - B votes for C.
> - A then switches its vote to C, but throws a NPE when attempting to connect.
> This causes the QuorumPeer to spin up a Follower only to immediately have it
> shutdown by the exception.
> Ideally, A would not advertise a vote for a server that it will not follow.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
