Your auto scaling algorithm is THROUGHPUT_BASED, it will kicks in only when
it feels the pipeline is not able to keep it up with the incoming source.
How big is your bounded-source and how much pressure (messages per seconds)
your unbounded source is receiving?

Maulik Gandhi <[email protected]> schrieb am Di., 19. März 2019, 21:06:

> Hi Juan,
>
> Thanks for replying.  I believe I am using correct configurations.
>
> I have posted more details with code snippet and Data Flow job template
> configuration on Stack Overflow post:
> https://stackoverflow.com/q/55242684/11226631
>
> Thanks.
> - Maulik
>
> On Tue, Mar 19, 2019 at 2:53 PM Juan Carlos Garcia <[email protected]>
> wrote:
>
>> Hi Maulik,
>>
>> Have you submitted your job with the correct configuration to enable
>> autoscaling?
>>
>> --autoscalingAlgorithm=
>> --maxWorkers=
>>
>> I am on my phone right now and can't tell if the flags name are 100%
>> correct.
>>
>>
>> Maulik Gandhi <[email protected]> schrieb am Di., 19. März 2019, 18:13:
>>
>>>
>>> Maulik Gandhi <[email protected]>
>>> 10:19 AM (1 hour ago)
>>> to user
>>> Hi Beam Community,
>>>
>>> I am working on Beam processing pipeline, which reads data from the
>>> non-bounded and bounded source and want to leverage Beam state management
>>> in my pipeline.  For putting data in Beam state, I have to transfer the
>>> data in key-value (eg: KV<String, Object>.  As I am reading data from the
>>> non-bounded and bounded source, I am forced to perform Window + Triggering,
>>> before grouping data by key.  I have chosen to use GlobalWindows().
>>>
>>> I am able to kick-off the Data Flow job, which would run my Beam
>>> pipeline.  I have noticed Data Flow would use only 1 Worker node to perform
>>> the work, and would not scale the job to use more worker nodes, thus not
>>> leveraging the benefit of distributed processing.
>>>
>>> I have posted the question on Stack Overflow:
>>> https://stackoverflow.com/questions/55242684/join-bounded-and-non-bounded-source-data-flow-job-not-scaling
>>>  but
>>> reaching out on the mailing list, to get some help, or learn what I
>>> am missing.
>>>
>>> Any help would be appreciated.
>>>
>>> Thanks.
>>> - Maulik
>>>
>>

Reply via email to