[HBASE-22652] Flakey TestLockManager; test timed out after 780 seconds

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.1.5
• Fix Version/s: 3.0.0-alpha-1, 2.3.0, 2.0.6, 2.2.1, 2.1.6
• Component/s: proc-v2
• Hadoop Flags: Reviewed
• Assignee: Michael Stack
• Reporter: Michael Stack
• Created: 03/Jul/19 03:12
• Updated: 16/Jul/19 04:46
• Resolved: 03/Jul/19 14:52

Description

In nightly 1318 on branch-2.1, TestLockManager failed. It was only fail. It fails on occasion. In this instance, test failed like this:

Error Message
test timed out after 780 seconds
Stacktrace
org.junit.runners.model.TestTimedOutException: test timed out after 780 seconds
	at org.apache.hadoop.hbase.master.locking.TestLockManager.tearDown(TestLockManager.java:105)

Digging in log, before the timeout in teardown, you see this:

2019-07-01 22:45:57,849 INFO  [PEWorker-1] procedure2.ProcedureExecutor$WorkerThread(2041): ASSERT pid=11
java.lang.AssertionError
	at org.apache.hadoop.hbase.procedure2.TimeoutExecutorThread.add(TimeoutExecutorThread.java:81)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.execProcedure(ProcedureExecutor.java:1765)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.executeProcedure(ProcedureExecutor.java:1462)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.access$1200(ProcedureExecutor.java:78)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor$WorkerThread.run(ProcedureExecutor.java:2039)

i.e. an assert trips. The assert is doing this:

    assert procedure.getState() == ProcedureState.WAITING_TIMEOUT;

The procedure is expected to be in WAITING_TIMEOUT state when we go to add it to the timeout executor. In the log snippet above, we see the thread is in RUNNABLE state. But even on successful run it is in RUNNABLE state. So what is going on.

I think I figured it accidentally.

I'd added a pause at the assert and when the assert ran, procedure was in RUNNABLE state and it would fail every time. So, it is a timing issue.

I added logging of the setting of state on the procedure. There weren't many procedures so was easy to follow. Here is what I saw (log included emission of stack trace):

2019-07-02 19:50:14,941 DEBUG [PEWorker-1] locking.LockProcedure(313): LOCKED pid=11, state=RUNNABLE; org.apache.hadoop.hbase.master.locking.LockProcedure, namespace=namespace, type=EXCLUSIVE
2019-07-02 19:50:14,943 INFO  [PEWorker-1] procedure2.Procedure(793): SET STATE ON pid=11, state=WAITING_TIMEOUT, locked=true; org.apache.hadoop.hbase.master.locking.LockProcedure, namespace=namespace, type=EXCLUSIVE
java.lang.Exception
	at org.apache.hadoop.hbase.procedure2.Procedure.setState(Procedure.java:793)
	at org.apache.hadoop.hbase.master.locking.LockProcedure.execute(LockProcedure.java:249)
	at org.apache.hadoop.hbase.master.locking.LockProcedure.execute(LockProcedure.java:56)
	at org.apache.hadoop.hbase.procedure2.Procedure.doExecute(Procedure.java:966)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.execProcedure(ProcedureExecutor.java:1723)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.executeProcedure(ProcedureExecutor.java:1462)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.access$1200(ProcedureExecutor.java:78)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor$WorkerThread.run(ProcedureExecutor.java:2039)
2019-07-02 19:50:14,944 INFO  [Time-limited test] procedure2.Procedure(793): SET STATE ON pid=11, state=RUNNABLE, locked=true; org.apache.hadoop.hbase.master.locking.LockProcedure, namespace=namespace, type=EXCLUSIVE
java.lang.Exception
	at org.apache.hadoop.hbase.procedure2.Procedure.setState(Procedure.java:793)
	at org.apache.hadoop.hbase.master.locking.LockProcedure.unlock(LockProcedure.java:225)
	at org.apache.hadoop.hbase.master.locking.LockManager$MasterLock.release(LockManager.java:185)
	at org.apache.hadoop.hbase.master.locking.TestLockManager.testMasterLockAcquire(TestLockManager.java:124)
	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:50)
	at org.junit.internal.runners.model.ReflectiveCallable.run(ReflectiveCallable.java:12)
	at org.junit.runners.model.FrameworkMethod.invokeExplosively(FrameworkMethod.java:47)
	at org.junit.internal.runners.statements.InvokeMethod.evaluate(InvokeMethod.java:17)
	at org.junit.internal.runners.statements.RunAfters.evaluate(RunAfters.java:27)
	at org.junit.rules.TestWatcher$1.evaluate(TestWatcher.java:55)
	at org.junit.rules.RunRules.evaluate(RunRules.java:20)
	at org.junit.runners.ParentRunner.runLeaf(ParentRunner.java:325)
	at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:78)
	at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:57)
	at org.junit.runners.ParentRunner$3.run(ParentRunner.java:290)
	at org.junit.runners.ParentRunner$1.schedule(ParentRunner.java:71)
	at org.junit.runners.ParentRunner.runChildren(ParentRunner.java:288)
	at org.junit.runners.ParentRunner.access$000(ParentRunner.java:58)
	at org.junit.runners.ParentRunner$2.evaluate(ParentRunner.java:268)
	at org.junit.internal.runners.statements.RunBefores.evaluate(RunBefores.java:26)
	at org.junit.internal.runners.statements.RunAfters.evaluate(RunAfters.java:27)
	at org.junit.internal.runners.statements.FailOnTimeout$CallableStatement.call(FailOnTimeout.java:298)
	at org.junit.internal.runners.statements.FailOnTimeout$CallableStatement.call(FailOnTimeout.java:292)
	at java.util.concurrent.FutureTask.run$$$capture(FutureTask.java:266)
	at java.util.concurrent.FutureTask.run(FutureTask.java)
	at java.lang.Thread.run(Thread.java:748)
2019-07-02 19:50:15,037 DEBUG [Time-limited test] procedure.MasterProcedureScheduler(356): Add TableQueue(hbase:namespace, xlock=false sharedLock=1 size=1) to run queue because: pid=11, state=RUNNABLE, locked=true; org.apache.hadoop.hbase.master.locking.LockProcedure, namespace=namespace, type=EXCLUSIVE has lock
2019-07-02 19:53:37,062 WARN  [JvmPauseMonitor] util.JvmPauseMonitor$Monitor(176): Detected pause in JVM or host machine (eg GC): pause of approximately 201656ms
No GCs detected
2019-07-02 19:53:37,063 WARN  [ProcExecTimeout] procedure2.ProcedureExecutor$WorkerMonitor(2147): Worker stuck PEWorker-1(pid=11), run time 3mins, 22.122sec
2019-07-02 19:53:37,063 WARN  [RS:0;localhost:57336] util.Sleeper(86): We slept 202150ms instead of 1000ms, this is likely due to a long garbage collecting pause and it's usually bad, see http://hbase.apache.org/book.html#trouble.rs.runtime.zkexpired
2019-07-02 19:53:37,064 INFO  [master/localhost:0.splitLogManager..Chore.1] hbase.ScheduledChore(176): Chore: SplitLogManager Timeout Monitor missed its start time
Disconnected from the target VM, address: '127.0.0.1:57316', transport: 'socket'
2019-07-02 19:53:37,063 INFO  [regionserver/localhost:0.Chore.1] hbase.ScheduledChore(176): Chore: MemstoreFlusherChore missed its start time
2019-07-02 19:53:37,069 INFO  [PEWorker-1] procedure2.ProcedureExecutor$WorkerThread(2041): ASSERT pid=11
java.lang.AssertionError
	at org.apache.hadoop.hbase.procedure2.TimeoutExecutorThread.add(TimeoutExecutorThread.java:81)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.execProcedure(ProcedureExecutor.java:1765)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.executeProcedure(ProcedureExecutor.java:1462)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor.access$1200(ProcedureExecutor.java:78)
	at org.apache.hadoop.hbase.procedure2.ProcedureExecutor$WorkerThread.run(ProcedureExecutor.java:2039)

So, the worker thread goes to execute the lock procedure. It sets the state to WAITING_TIMEOUT... My break point trips and it holds up the timeout executor. Meantime the main test thread continues and releases lock which sets the procedure state back to RUNNABLE. I let the break point go and the worker thread finds the procedure state is RUNNABLE and the assertion trips.

I think the assert should tolerate RUNNABLE state too.

Looking at all ways into the add to timeout executor, all test in caller in one way or another for WAITING_TIMEOUT except for case where we are loading procedures from stores. This latter is a little unpredictable. The other entrances will have seen the state set to WAITING_TIMEOUT at least just before we get to the assert point.

Let me relax the assert some.

Issue Links

• links to: Review Board (branch-2.1)

Sub-Tasks

• LockProcedure cleanups: Open