[HDDS-1709] TestScmSafeNode is flaky

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 0.4.1
• Component/s: SCM, test
• Labels: pull-request-available
• Target Version/s: 0.4.1
• Sprint: Done
• Assignee: Marton Elek
• Reporter: Marton Elek
• Created: 20/Jun/19 08:17
• Updated: 05/Aug/19 14:46
• Resolved: 25/Jun/19 21:07

Description

org.apache.hadoop.ozone.om.TestScmSafeMode.testSCMSafeMode is failed at last night with the following error:

java.lang.AssertionError at org.junit.Assert.fail(Assert.java:86) at org.junit.Assert.assertTrue(Assert.java:41) at org.junit.Assert.assertTrue(Assert.java:52) at org.apache.hadoop.ozone.om.TestScmSafeMode.testSCMSafeMode(TestScmSafeMode.java:285) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:498) at org.junit.runners.model.FrameworkMethod$1.runReflectiveCall(FrameworkMethod.java:47) at org.junit.internal.runners.model.ReflectiveCallable.run(ReflectiveCallable.java:12) at org.junit.runners.model.FrameworkMethod.invokeExplosively(FrameworkMethod.java:44) at org.junit.internal.runners.statements.InvokeMethod.evaluate(InvokeMethod.java:17) at org.junit.internal.runners.statements.FailOnTimeout$StatementThread.run(FailOnTimeout.java:74)

Locally it can be tested but it's very easy to reproduce by adding an additional sleep DataNodeSafeModeRule:

+++ b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/safemode/DataNodeSafeModeRule.java
@@ -63,7 +63,11 @@ protected boolean validate() {
 
   @Override
   protected void process(NodeRegistrationContainerReport reportsProto) {
-
+    try {
+      Thread.sleep(3000);
+    } catch (InterruptedException e) {
+      e.printStackTrace();
+    }

This is a clear race condition:

DatanodeSafeModeRule and ContainerSafeModeRule are processing the same events but it can be possible (in case of an accidental sleep) that the container safe mode rule is done, but DatanodeSafeModeRule didn't process the new event (yet).

As a result the test execution will continue:

GenericTestUtils
    .waitFor(() -> scm.getCurrentContainerThreshold() == 1.0, 100, 20000);

(This line is waiting ONLY for the ContainerSafeModeRule).

The fix is easy, let's wait for the processing of all the async events:

EventQueue eventQueue =
    (EventQueue) cluster.getStorageContainerManager().getEventQueue();
eventQueue.processAll(5000L);

As we are sure that the events are already sent to the EventQueue (because we have the previous waitFor), it should be enough.

Issue Links

• links to: GitHub Pull Request #993