Hive job name

[Alex Bohr <al...@gradientx.com>]

Since going from Hive 0.11 to 0.13 the Job Names, in the JobTracker UI, for
my Hive queries in the job tracker changed.

It used to be the first 200 characters or so of the Hive query, now it's
just "hive-exec-0.13.1-cdh5.3.1.jar."

Seeing the beginning of the query in the job Name was super helpful to
figure out which queries are taking super long or using heavy # of slots.
Now I need to drill down to the job config file and check the hive.query.string
key.

Is there a config setting I can use to make the Hive Job name reflect the
query being executed?

I'm still using the JobTracker, haven't switched to Yarn yet.  Does this
change with Yarn?

Thanks

Re: Hive job name

[Stephen Sprague <sp...@gmail.com>]

Gotta say i don't particular like this either.   We run thousands of
queries a day and by using a comment before the query you could manipulate
the job.name hive would use.   Now like the OP stated that's ignored -
every job is named "hive-exec-0.13.1-cdh5.3.1.jar"   Sure setting
mapreduce.job.name explicitly is a workaround but... that's a boat load of
code changes!

Would not there be a "fix" to roll this back to how it got the job.name
before?

Thanks,
Stephen Sprague

On Wed, Mar 11, 2015 at 1:38 PM, Viral Bajaria <vi...@gmail.com>
wrote:

> I haven't used Tez but it's in my list of experimentation...
>
> Thanks,
> Viral
>
> On Tue, Mar 10, 2015 at 8:38 AM, P lva <ru...@gmail.com> wrote:
>
>> Hey Viral,
>>
>> Is there a similar config for tez ?
>>
>> Thanks
>>
>>
>>
>>
>> On Mon, Mar 9, 2015 at 6:36 PM, Viral Bajaria <vi...@gmail.com>
>> wrote:
>>
>>> We use the hive.job.name property to set meaningful job names. Look
>>> into using that before submitting queries.
>>>
>>> Thanks,
>>> Viral
>>>
>>>
>>> On Mon, Mar 9, 2015 at 2:47 PM, Alex Bohr <al...@gradientx.com> wrote:
>>>
>>>> Since going from Hive 0.11 to 0.13 the Job Names, in the JobTracker UI,
>>>> for my Hive queries in the job tracker changed.
>>>>
>>>> It used to be the first 200 characters or so of the Hive query, now
>>>> it's just "hive-exec-0.13.1-cdh5.3.1.jar."
>>>>
>>>> Seeing the beginning of the query in the job Name was super helpful to
>>>> figure out which queries are taking super long or using heavy # of slots.
>>>> Now I need to drill down to the job config file and check the hive.query.string
>>>> key.
>>>>
>>>> Is there a config setting I can use to make the Hive Job name reflect
>>>> the query being executed?
>>>>
>>>> I'm still using the JobTracker, haven't switched to Yarn yet.  Does
>>>> this change with Yarn?
>>>>
>>>> Thanks
>>>>
>>>
>>>
>>
>

Re: Hive job name

[Viral Bajaria <vi...@gmail.com>]

I haven't used Tez but it's in my list of experimentation...

Thanks,
Viral

On Tue, Mar 10, 2015 at 8:38 AM, P lva <ru...@gmail.com> wrote:

> Hey Viral,
>
> Is there a similar config for tez ?
>
> Thanks
>
>
>
>
> On Mon, Mar 9, 2015 at 6:36 PM, Viral Bajaria <vi...@gmail.com>
> wrote:
>
>> We use the hive.job.name property to set meaningful job names. Look into
>> using that before submitting queries.
>>
>> Thanks,
>> Viral
>>
>>
>> On Mon, Mar 9, 2015 at 2:47 PM, Alex Bohr <al...@gradientx.com> wrote:
>>
>>> Since going from Hive 0.11 to 0.13 the Job Names, in the JobTracker UI,
>>> for my Hive queries in the job tracker changed.
>>>
>>> It used to be the first 200 characters or so of the Hive query, now it's
>>> just "hive-exec-0.13.1-cdh5.3.1.jar."
>>>
>>> Seeing the beginning of the query in the job Name was super helpful to
>>> figure out which queries are taking super long or using heavy # of slots.
>>> Now I need to drill down to the job config file and check the hive.query.string
>>> key.
>>>
>>> Is there a config setting I can use to make the Hive Job name reflect
>>> the query being executed?
>>>
>>> I'm still using the JobTracker, haven't switched to Yarn yet.  Does this
>>> change with Yarn?
>>>
>>> Thanks
>>>
>>
>>
>

Re: Hive job name

[P lva <ru...@gmail.com>]

Hey Viral,

Is there a similar config for tez ?

Thanks




On Mon, Mar 9, 2015 at 6:36 PM, Viral Bajaria <vi...@gmail.com>
wrote:

> We use the hive.job.name property to set meaningful job names. Look into
> using that before submitting queries.
>
> Thanks,
> Viral
>
>
> On Mon, Mar 9, 2015 at 2:47 PM, Alex Bohr <al...@gradientx.com> wrote:
>
>> Since going from Hive 0.11 to 0.13 the Job Names, in the JobTracker UI,
>> for my Hive queries in the job tracker changed.
>>
>> It used to be the first 200 characters or so of the Hive query, now it's
>> just "hive-exec-0.13.1-cdh5.3.1.jar."
>>
>> Seeing the beginning of the query in the job Name was super helpful to
>> figure out which queries are taking super long or using heavy # of slots.
>> Now I need to drill down to the job config file and check the hive.query.string
>> key.
>>
>> Is there a config setting I can use to make the Hive Job name reflect the
>> query being executed?
>>
>> I'm still using the JobTracker, haven't switched to Yarn yet.  Does this
>> change with Yarn?
>>
>> Thanks
>>
>
>

Re: Hive job name

[Viral Bajaria <vi...@gmail.com>]

We use the hive.job.name property to set meaningful job names. Look into
using that before submitting queries.

Thanks,
Viral


On Mon, Mar 9, 2015 at 2:47 PM, Alex Bohr <al...@gradientx.com> wrote:

> Since going from Hive 0.11 to 0.13 the Job Names, in the JobTracker UI,
> for my Hive queries in the job tracker changed.
>
> It used to be the first 200 characters or so of the Hive query, now it's
> just "hive-exec-0.13.1-cdh5.3.1.jar."
>
> Seeing the beginning of the query in the job Name was super helpful to
> figure out which queries are taking super long or using heavy # of slots.
> Now I need to drill down to the job config file and check the hive.query.string
> key.
>
> Is there a config setting I can use to make the Hive Job name reflect the
> query being executed?
>
> I'm still using the JobTracker, haven't switched to Yarn yet.  Does this
> change with Yarn?
>
> Thanks
>