[ACCUMULO-2764] Stopping MAC before it's processes have fully started causes an indefinite hang

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.5.1, 1.6.0
• Fix Version/s: 1.5.2, 1.6.1, 1.7.0
• Component/s: mini

Environment

OpenJDK 1.6.0, CentOS 6.5, 2CPU, 6GB RAM (virtual hardware)

• Assignee: Josh Elser
• Reporter: Christopher Tubbs
• Created: 30/Apr/14 18:48
• Updated: 11/Jul/14 17:31
• Resolved: 15/May/14 18:36

Description

I saw this testing 1.6.0-RC5.

Calling process.destroy() and then process.waitFor(), as MiniAccumuloCluster does in it's stop method, before the process is fully started, appears to create an indefinite hang.

I saw this most recently in MiniAccumuloClusterGCTest.testAccurateProcessListReturned, which gets a ProcessReference and then immediately shuts down MAC, though it was also the root cause of ACCUMULO-2756. In this instance, the test got stuck in the MAC teardown.

"main" prio=10 tid=0x00007f3cf4008800 nid=0x2b19 in Object.wait() [0x00007f3cf8f9c000]
   java.lang.Thread.State: WAITING (on object monitor)
        at java.lang.Object.wait(Native Method)
        - waiting on <0x00000000e29dd2e8> (a java.lang.UNIXProcess)
        at java.lang.Object.wait(Object.java:502)
        at java.lang.UNIXProcess.waitFor(UNIXProcess.java:181)
        - locked <0x00000000e29dd2e8> (a java.lang.UNIXProcess)
        at org.apache.accumulo.minicluster.impl.MiniAccumuloClusterImpl.stop(MiniAccumuloClusterImpl.java:607)
        at org.apache.accumulo.minicluster.impl.MiniAccumuloClusterGCTest.tearDownMiniCluster(MiniAccumuloClusterGCTest.java:74)
        at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
        at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
        at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
        at java.lang.reflect.Method.invoke(Method.java:622)
        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.RunAfters.evaluate(RunAfters.java:33)
        at org.junit.runners.ParentRunner.run(ParentRunner.java:309)
        at org.apache.maven.surefire.junit4.JUnit4Provider.execute(JUnit4Provider.java:264)
        at org.apache.maven.surefire.junit4.JUnit4Provider.executeTestSet(JUnit4Provider.java:153)
        at org.apache.maven.surefire.junit4.JUnit4Provider.invoke(JUnit4Provider.java:124)
        at org.apache.maven.surefire.booter.ForkedBooter.invokeProviderInSameClassLoader(ForkedBooter.java:200)
        at org.apache.maven.surefire.booter.ForkedBooter.runSuitesInProcess(ForkedBooter.java:153)
        at org.apache.maven.surefire.booter.ForkedBooter.main(ForkedBooter.java:103)

It appears that destroy() doesn't actually succeed in destroying a process which is still starting, so the waitFor() waits indefinitely. I haven't debugged further. It may be a JVM bug, or a limitation in the java Process API, or some UNIX signal handling quirk with process instantiation that destroy() cannot know.

One fix could be to make start() wait until the metadata table can be scanned before it returns, to ensure all processes are actually running and ready. Another fix would be to have the teardown code try another destroy if waitFor() doesn't return after a reasonable amount of time.

Issue Links

• breaks: ACCUMULO-2985 MAC doesn't stop cleanly in 1.6.1-SNAPSHOT
• relates to: ACCUMULO-2756 multiTableTest is timing out
• relates to: ACCUMULO-2763 Repeated hang of build in MiniAccumuloClusterImplTest