Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Hive, mail # user - Hive job fails on hive client even though all map-red stages finish but succeeds on hive server


Copy link to this message
-
Re: Hive job fails on hive client even though all map-red stages finish but succeeds on hive server
Jagat Singh 2012-08-11, 07:49
Hi Anurag,

How much space is for /user and /tmp directory on client.

Did you check that part? , anything which might stop move task from
finishing.

-----------
Sent from Mobile , short and crisp.
On 11-Aug-2012 1:37 PM, "Anurag Tangri" <[EMAIL PROTECTED]> wrote:

> Hi,
> We are facing this issue where we run a hive job over huge data about ~6
> TB input.
>
> We run this from hive client and hive metastore server is on another
> machine.
>
>
> If we have smaller input, this job succeeds but for above input size, it
> fails with error :
>
> 2012-08-11 01:34:01,722 Stage-1 map = 100%,  reduce = 100%
>
> 2012-08-11 01:35:02,195 Stage-1 map = 100%,  reduce = 100%
>
> 2012-08-11 01:36:02,682 Stage-1 map = 100%,  reduce = 100%
>
> 2012-08-11 01:37:03,215 Stage-1 map = 100%,  reduce = 100%
>
> 2012-08-11 01:38:03,719 Stage-1 map = 100%,  reduce = 100%
>
> 2012-08-11 01:39:04,311 Stage-1 map = 100%,  reduce = 100%
>
> Ended Job = job_201207072204_34432
>
> Loading data to table default.atangri_test_1
>
> Failed with exception Unable to fetch table atangri_test_1
>
> FAILED: Execution Error, return code 1 from
> org.apache.hadoop.hive.ql.exec.MoveTask
>
>
> If we have smaller input (~2 TB), this job succeeds but for above input
> size, it fails with error : We have set
> hive.metastore.client.socket.timeout to big value like 86400  but still it
> fails after about 8-9 hours.
>
> Does anyone face the same issue or any pointers ?
>
> The job succeeds if it is directly run on hive server.
>
> Thanks,
> Anurag Tangri
>