hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Junheng Gong (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-13678) Hbase client block on region call back
Date Wed, 13 May 2015 05:06:59 GMT

     [ https://issues.apache.org/jira/browse/HBASE-13678?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Junheng Gong updated HBASE-13678:
---------------------------------
    Description: 
When batch get some data from cluster, the client will block some threads base on region server
amount

In code client use Future.get block a thread for each related region servers and wait them
finish one by one.

In our case we have 12 region servers, so most of time the client will cost 12 threads for
a batch get, if we have many concurrent request it will OOM by can not create native threads.

I suggest we use some kind async or call back to process region server's response.

  was:
when batch get some data from cluster, the client will block some threads base on region server
amount

In code client use Future.get block a thread for each related region servers and wait them
finish one by one.

In our case we have 12 region servers, so most of time the client will cost 12 threads for
a batch get, if we have many concurrent request it will OOM by can not create native threads.

I suggest we use some kind async or call back to process region server's response.


> Hbase client block on region call back
> --------------------------------------
>
>                 Key: HBASE-13678
>                 URL: https://issues.apache.org/jira/browse/HBASE-13678
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 0.94.15
>         Environment: CentOS 6.4 scala 2.11/java 1.7
>            Reporter: Junheng Gong
>
> When batch get some data from cluster, the client will block some threads base on region
server amount
> In code client use Future.get block a thread for each related region servers and wait
them finish one by one.
> In our case we have 12 region servers, so most of time the client will cost 12 threads
for a batch get, if we have many concurrent request it will OOM by can not create native threads.
> I suggest we use some kind async or call back to process region server's response.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message