hive-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Edward Capriolo <edlinuxg...@gmail.com>
Subject Re: Closing long running hive query forcefully
Date Thu, 07 Feb 2013 16:31:57 GMT
That is a good way to do it. We do it with comment sometimes.

select /* myid bla bla*/ x,y,z

Edward

On Thu, Feb 7, 2013 at 11:12 AM, Gabor Makrai <makrai.list@gmail.com> wrote:
> Hi,
>
> We solved this problem in the following way (this is really not a simple
> solution):
> - start hive query in a different thread
> - we generated an unique id for the query and used the SET key=value;
> command (before the long query command) to give this unique id to the MR
> jobs related to the query
> - main thread waits for the completion or the cancellation sign (comes from
> the gui or anywhere else)
> - if cancellation sign comes, then we will kill the running query through
> the JobTracker
>   - list all jobs and search the job which contains the previously
> configured unique id
>   - kill that job
>   - that will cause exception on the query thread
>
> It's a little bit ugly, but thrift doesn't support async communication and
> so "statement.cancel()" function can't be supported/implemented.
>
> Gabor
>
>
>
> On Thu, Feb 7, 2013 at 2:21 PM, Nitin Pawar <nitinpawar432@gmail.com> wrote:
>>
>> hive waits till the hadoop job is completed. so unless you kill the job of
>> jdbc connection is dropped I don't see any other way to reduce the load on
>> application.
>> best when you think its  long enough, you will need to find a way to kill
>> the hadoop job. That will automatically release the resources in the
>> pipeline
>>
>>
>> On Thu, Feb 7, 2013 at 6:35 PM, Kugathasan Abimaran
>> <abimarank@hsenidmobile.com> wrote:
>>>
>>> Hi All,
>>>
>>>
>>> Are there anyways to close the long running hive query through hive-jdbc?
>>> Since when ever Hive hangs, my application also hang, So I want to close the
>>> hive connection forcefully after a certain time.
>>>
>>>
>>>
>>> --
>>> Thanks,
>>> With Regards,
>>>
>>> Abimaran
>>
>>
>>
>>
>> --
>> Nitin Pawar
>
>

Mime
View raw message