[
https://issues.apache.org/jira/browse/HBASE-17747?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15923729#comment-15923729
]
Yu Li commented on HBASE-17747:
-------------------------------
bq. This is the GC options we used for CMS
Well, yes, we also have more complicated options online, but for testing, I
don't think all the configurations are relative... Still, please give it a try
if you think we could resolve it through option tuning, and feel free to revert
the commit if you succeed... Or if you'd like me to try the combination you
posted and will admit that we cannot resolve the issue by tuning if this
combination couldn't work, just let me know.
bq. And still, you need to test it with G1 as the problem you described is
highly related to the GC algorithm
Notice that this is weak/soft reference object, not plain object, so we are
talking about how JVM will clean up the object with weak/soft reference
automatically and the difference. I don't think this is GC algorithm related...
No matter which GC algorithm, it should clear the object immediately when
there's no other but a weak reference to it, so with weak reference more
{{IdReadWriteLock}} objects will be created, and under high read concurrency,
such objects are so many that affects GC speed (for whatever GC algorithm),
while soft reference is totally another case. See below codes in
{{ObjectPool#get}}
{code}
public V get(K key) {
Reference<V> ref = referenceCache.get(key);
if (ref != null) {
V obj = ref.get();
if (obj != null) {
return obj;// with soft reference, we will probably return here
}
referenceCache.remove(key, ref);
}
V newObj = objectFactory.createObject(key); // with weak reference, we will
probably create new object here
Reference<V> newRef = createReference(key, newObj);
while (true) {
Reference<V> existingRef = referenceCache.putIfAbsent(key, newRef);
if (existingRef == null) {
return newObj;
}
V existingObject = existingRef.get();
if (existingObject != null) {
return existingObject;
}
referenceCache.remove(key, existingRef);
}
}
{code}
> Support both weak and soft object pool
> --------------------------------------
>
> Key: HBASE-17747
> URL: https://issues.apache.org/jira/browse/HBASE-17747
> Project: HBase
> Issue Type: Improvement
> Affects Versions: 2.0
> Reporter: Yu Li
> Assignee: Yu Li
> Fix For: 2.0
>
> Attachments: HBASE-17747.patch, HBASE-17747.v2.patch,
> HBASE-17747.v3.patch
>
>
> During YCSB testing on embedded mode after HBASE-17744, we found that under
> high read load GC is quite severe even with offheap L2 cache. After some
> investigation, we found it's caused by using weak reference in
> {{IdReadWriteLock}}. In embedded mode the read is so quick that the lock
> might already get promoted to the old generation when the weak reference is
> cleared, which causes dirty card table (old reference get removed and new
> lock object set into {{referenceCache}}, see {{WeakObjectPool#get}}) thus
> slowing YGC. In distributed mode there'll also be more lock object created
> with weak reference than soft reference that slowing down the processing.
> So we proposed to use soft reference for this {{IdReadWriteLock}} used in
> cache, which won't get cleared until JVM memory is not enough, and could
> resolve the issue mentioned above. What's more, we propose to extend the
> {{WeakObjectPool}} to be more generate to support both weak and soft
> reference.
> Note that the GC issue only emerges under embedded mode with DirectOperator,
> in which case all costs on the wire is removed thus produces extremely high
> concurrency.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
