You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Victor Xu (JIRA)" <ji...@apache.org> on 2014/06/27 12:57:24 UTC

[jira] [Updated] (HBASE-11420) ThriftServer (version 1) may crash on OOME

     [ https://issues.apache.org/jira/browse/HBASE-11420?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Victor Xu updated HBASE-11420:
------------------------------

    Attachment: HBASE-11420.patch

Attach a simple patch.
I also find the bug only happens when using Binary Protocol, so I change the default Protocol Factory constructor to a strict read& write version.

> ThriftServer (version 1) may crash on OOME
> ------------------------------------------
>
>                 Key: HBASE-11420
>                 URL: https://issues.apache.org/jira/browse/HBASE-11420
>             Project: HBase
>          Issue Type: Bug
>          Components: Thrift
>    Affects Versions: 0.98.2
>         Environment: hadoop-cdh5, hbase-0.98.2, thrift-0.9.0
>            Reporter: Victor Xu
>         Attachments: HBASE-11420.patch
>
>
> When using ThriftServer as a gateway for php & c/c++ applications, I found it very easy to crash of OOME. I analyzed the jprof file and found that the ThriftServer had a 1.2G size byte array before it crashed. It seems to be a memory leak. But when did it happen? I checked the huge byte array and realized that it was a HTTP request string. That means a request in protocols other than Thrift may cause memory allocation exception. 
> We cound easily recur the bug by wget/curl the ThriftServer. And we can check the memory usage infomation using TOP command.



--
This message was sent by Atlassian JIRA
(v6.2#6252)