We have upgraded from CDH3b2 to CDH3b3 recently, and our job tracker process now
consistently deadlocks when running a particular map reduce job that worked with
the previous version.

Here is a stacktrace showing the deadlocked threads.  

http://pastebin.com/AMEig0Z0

Any ideas?

-Luke Forehand

Please use the Cloudera support lists for Cloudera specific questions.

bcc: common-dev
to: cdh-user

Hi Luke,

Please send CDH-specific mail only to the CDH user/dev lists. The Apache
lists are intended for discussion of the upstream project.

--
Aaron T. Myers
Software Engineer, Cloudera

On Thu, Dec 16, 2010 at 8:59 AM, Luke Forehand <
[EMAIL PROTECTED]> wrote:

> We have upgraded from CDH3b2 to CDH3b3 recently, and our job tracker
> process now
> consistently deadlocks when running a particular map reduce job that worked
> with
> the previous version.
>
> Here is a stacktrace showing the deadlocked threads.
>
> http://pastebin.com/AMEig0Z0
>
> Any ideas?
>
> -Luke Forehand
>
>
>

Replying back on common-dev list because I've determined this issue
exists upstream. Configuration.writeXml will deadlock if the XML
output is larger than one block (or I think if pipeline recovery
happens mid-write)

I filed HDFS-1542.

-Todd

On Thu, Dec 16, 2010 at 8:59 AM, Luke Forehand
<[EMAIL PROTECTED]> wrote:
> We have upgraded from CDH3b2 to CDH3b3 recently, and our job tracker process now
> consistently deadlocks when running a particular map reduce job that worked with
> the previous version.
>
> Here is a stacktrace showing the deadlocked threads.
>
> http://pastebin.com/AMEig0Z0
>
> Any ideas?
>
> -Luke Forehand
>
>
>

--
Todd Lipcon
Software Engineer, Cloudera