Hi Junrui,
I think you understood correctly. What I'm seeing is that each vertex has a
single subtask, but multiple vertices are started in parallel in different
slots. That is not a problem in my case, I _want_ to parallelize the work,
it's just that this mechanism is very different from streamin
Hi Alexis,
Could you clarify what you mean by "If I add more slots to the task
manager, I see the transformations actually start in parallel even though I
submit the job with 'flink run -p 1'"?
Are you asking if multiple slots are working simultaneously, or if a single
JobVertex contains multiple
Hi Junrui,
Thanks for the confirmation. I tested some more and I'm seeing a strange
behavior.
I'm currently testing a single source stream that is fed to 6 identical
transformations. The state processor api requires batch mode and, from what
I can tell, I must specify a parallelism of 1 in the jo
Hi Alexis,
For the SavepointWriter, I've briefly looked over the code and the write
operation is enforced as non-parallel.
Best,
Junrui
Alexis Sarda-Espinosa 于2024年7月6日周六 01:27写道：
> Hi Gabor,
>
> Thanks for the quick response. What about SavepointWriter? In my case I'm
> actually writing a job
Hi Gabor,
Thanks for the quick response. What about SavepointWriter? In my case I'm
actually writing a job that will read from an existing savepoint and modify
some of its data to write a new one.
Regards,
Alexis.
Am Fr., 5. Juli 2024 um 17:37 Uhr schrieb Gabor Somogyi <
gabor.g.somo...@gmail.co
Hi Alexis,
It depends. When one uses SavepointLoader to read metadata only then it's
non-parallel.
SavepointReader however is basically a normal batch job with all its
features.
G
On Fri, Jul 5, 2024 at 5:21 PM Alexis Sarda-Espinosa <
sarda.espin...@gmail.com> wrote:
> Hello,
>
> Really quick
Hello,
Really quick question, when using the state processor API, are all
transformations performed in a non-parallel fashion?
Regards,
Alexis.
