gt; environment?
>
>
> Best,
> Yun Tang
>
> From: David Morávek
> Sent: Wednesday, January 5, 2022 19:51
> To: dev
> Subject: Re: [DISCUSS] FLIP-201: Persist local state in working directory
>
> +1 the general direction here see
: [DISCUSS] FLIP-201: Persist local state in working directory
+1 the general direction here seems pretty solid
D.
On Wed, Jan 5, 2022 at 11:57 AM Till Rohrmann wrote:
> If there is no other larger feedback, I would start the vote soonish.
>
> Cheers,
> Till
>
> On Thu, Dec 3
+1 the general direction here seems pretty solid
D.
On Wed, Jan 5, 2022 at 11:57 AM Till Rohrmann wrote:
> If there is no other larger feedback, I would start the vote soonish.
>
> Cheers,
> Till
>
> On Thu, Dec 30, 2021 at 4:28 PM Till Rohrmann
> wrote:
>
> > Hi David,
> >
> > Thanks for
If there is no other larger feedback, I would start the vote soonish.
Cheers,
Till
On Thu, Dec 30, 2021 at 4:28 PM Till Rohrmann wrote:
> Hi David,
>
> Thanks for your feedback.
>
> With the graceful shutdown I mean a way to stop the TaskManager and to
> clean up the working directory. At the
Hi David,
Thanks for your feedback.
With the graceful shutdown I mean a way to stop the TaskManager and to
clean up the working directory. At the moment, I think we always kill the
process via SIGTERM or SIGKILL. This won't clean up the working directory
because it could also originate from a
Hi Till,
thanks for drafting the FLIP, it looks really good. I did a quick pass over
the PR and it seems to be heading in a right direction.
It might be required to introduce a graceful shutdown of the TaskExecutor
> in order to support proper cleanup of resources.
>
This is actively being
I've created draft PR for the desired changes [1]. It might be easier to
take a look at than the branch.
[1] https://github.com/apache/flink/pull/18237
Cheers,
Till
On Tue, Dec 28, 2021 at 3:22 PM Till Rohrmann wrote:
> Hi everyone,
>
> I would like to start a discussion about using the
Hi everyone,
I would like to start a discussion about using the working directory to
persist local state for faster recovery (FLIP-201) [1]. Persisting the
local state will be beneficial if a crashed process is restarted with the
same working directory. In this case, Flink does not have to
