pkuwm commented on a change in pull request #1035:
URL: https://github.com/apache/helix/pull/1035#discussion_r435701339
##########
File path:
helix-core/src/test/java/org/apache/helix/integration/TestZkCallbackHandlerLeak.java
##########
@@ -317,6 +324,150 @@ public boolean verify() throws Exception {
System.out.println("END " + clusterName + " at " + new
Date(System.currentTimeMillis()));
}
+ @Test
+ public void testDanglingCallbackHanlderFix() throws Exception {
+ String className = TestHelper.getTestClassName();
+ String methodName = TestHelper.getTestMethodName();
+ String clusterName = className + "_" + methodName;
+ final int n = 3;
+ final String zkAddr = ZK_ADDR;
+ System.out.println("START " + clusterName + " at " + new
Date(System.currentTimeMillis()));
+
+ TestHelper.setupCluster(clusterName, zkAddr, 12918, "localhost", "TestDB",
1, // resource
+ 32, // partitions
+ n, // nodes
+ 2, // replicas
+ "MasterSlave", true);
+
+ final ClusterControllerManager controller =
+ new ClusterControllerManager(zkAddr, clusterName, "controller_0");
+ controller.syncStart();
+
+ MockParticipantManager[] participants = new MockParticipantManager[n];
+ for (int i = 0; i < n; i++) {
+ String instanceName = "localhost_" + (12918 + i);
+ participants[i] = new MockParticipantManager(zkAddr, clusterName,
instanceName);
+ participants[i].syncStart();
+ }
+
+ Boolean result =
+ ClusterStateVerifier
+ .verifyByZkCallback(new
ClusterStateVerifier.BestPossAndExtViewZkVerifier(zkAddr,
+ clusterName));
+ Assert.assertTrue(result);
+
+ //HelixManager rpManager = HelixManagerFactory
+ // .getZKHelixManager(clusterName, "", InstanceType.SPECTATOR, ZK_ADDR);
+ //rpManager.connect();
+ ClusterSpectatorManager rpManager = new ClusterSpectatorManager(ZK_ADDR,
clusterName, "router");
+ rpManager.syncStart();
+ RoutingTableProvider rp = new RoutingTableProvider(rpManager,
PropertyType.CURRENTSTATES);
+
+ Thread.sleep(5000);
Review comment:
It doesn't have to explicitly sleep to wait for the new session
established because expireSession() already guarantees new session established.
I guess next sleep may be waiting for session handling or callback handling?
But is it necessary? If necessary, could we find a better/deterministic way to
verify it? Could you help understand what each of these sleeps is for.
----------------------------------------------------------------
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:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
