[
https://issues.apache.org/jira/browse/HBASE-17951?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15985883#comment-15985883
]
Guangxu Cheng edited comment on HBASE-17951 at 4/27/17 2:32 AM:
----------------------------------------------------------------
Yes, LongAdder is faster than AtomicLong.
I want to use incrementAndGet, but LongAdder is not implemented.
If we use LongAdder, can we do like this?
{code}
long tmpBlockedRequestsCount;
synchronized (blockedRequestsCount) {
blockedRequestsCount.increment();
tmpBlockedRequestsCount = getBlockedRequestsCount();
}
{code}
Maybe It could be worse than using AtomicLong.
By the way, it will block the client's request affecting the efficiency of the
client when memstoreDataSize is greater than blockingMemStoreSize.
Do you have any suggestions? Thanks. [~anoop.hbase]
was (Author: andrewcheng):
Yes, LongAdder is faster than AtomicLong.
I want to use incrementAndGet, but LongAdder is not implemented.
If we use LongAdder, can we do like this?
{code}
long tmpBlockedRequestsCount;
synchronized (blockedRequestsCount) {
blockedRequestsCount.increment();
tmpBlockedRequestsCount = getBlockedRequestsCount();
}
{code}
Do you have any suggestions? Thanks. [~anoop.hbase]
> Log the region's information in HRegion#checkResources
> ------------------------------------------------------
>
> Key: HBASE-17951
> URL: https://issues.apache.org/jira/browse/HBASE-17951
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.0.0, 1.3.0, 1.2.5
> Reporter: Guangxu Cheng
> Assignee: Guangxu Cheng
> Attachments: HBASE-17951-master-v1.patch, HBASE-17951-master-v2.patch
>
>
> We throw RegionTooBusyException in HRegion#checkResources if above memstore
> limit. On the server side，we just increment the value of blockedRequestsCount
> and do not print the region's information.
> I think we should also print the region's information in server-side, so that
> we can easily find the busy regions.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
