[
https://issues.apache.org/jira/browse/HBASE-18085?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16020685#comment-16020685
]
Yu Li commented on HBASE-18085:
-------------------------------
bq. In ur test method for tryLock, there is no logic other than just try lock
and release. If that is been removed as dead code by compiler?
Ok got your point now.
Have changed the testLock method to use Blackhole like:
{code}
@Benchmark
public void testLock(Blackhole bh) {
if (purgeLock.tryLock()) {
try {
// do purge
bh.consume(0);
return;
} finally {
purgeLock.unlock();
}
}
bh.consume(1);
}
{code}
and the result turned out:
{noformat}
Benchmark Mode Cnt Score
Error Units
AtomicLockBenchmark.testAtomicBoolean thrpt 100 47677994.547 ±
1295778.841 ops/s
AtomicLockBenchmark.testLock thrpt 100 2277364613.072 ±
49185975.985 ops/s
AtomicLockBenchmark.testVolatileBoolean thrpt 100 44386308.984 ±
644407.794 ops/s
{noformat}
Checking the compiled class:
{noformat}
public void testLock(org.openjdk.jmh.infra.Blackhole);
Code:
0: aload_0
1: getfield #7 // Field
purgeLock:Ljava/util/concurrent/locks/Lock;
4: invokeinterface #11, 1 // InterfaceMethod
java/util/concurrent/locks/Lock.tryLock:()Z
9: ifeq 39
12: aload_1
13: iconst_0
14: invokevirtual #12 // Method
org/openjdk/jmh/infra/Blackhole.consume:(I)V
17: aload_0
18: getfield #7 // Field
purgeLock:Ljava/util/concurrent/locks/Lock;
21: invokeinterface #13, 1 // InterfaceMethod
java/util/concurrent/locks/Lock.unlock:()V
26: return
27: astore_2
28: aload_0
29: getfield #7 // Field
purgeLock:Ljava/util/concurrent/locks/Lock;
32: invokeinterface #13, 1 // InterfaceMethod
java/util/concurrent/locks/Lock.unlock:()V
37: aload_2
38: athrow
39: aload_1
40: iconst_1
41: invokevirtual #12 // Method
org/openjdk/jmh/infra/Blackhole.consume:(I)V
44: return
Exception table:
from to target type
12 17 27 any
{noformat}
Let me know your thoughts sir [~anoop.hbase]
> Prevent parallel purge in ObjectPool
> ------------------------------------
>
> Key: HBASE-18085
> URL: https://issues.apache.org/jira/browse/HBASE-18085
> Project: HBase
> Issue Type: Bug
> Reporter: Yu Li
> Assignee: Yu Li
> Attachments: e89l05465.st3.jstack, HBASE-18085.patch
>
>
> Parallel purge in ObjectPool is meaningless and will cause contention issue
> since {{ReferenceQueue#poll}} has synchronization (source code shown below)
> {code}
> public Reference<? extends T> poll() {
> if (head == null)
> return null;
> synchronized (lock) {
> return reallyPoll();
> }
> }
> {code}
> We observed threads blocking on the purge method while using offheap bucket
> cache, and we could easily reproduce this by testing the 100% cache hit case
> in bucket cache with enough reading threads.
> We propose to add a purgeLock and use tryLock to avoid parallel purge.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
