Hello,
And if you want to go for a deployment, consider using a statefulset
instead. This way you are certain that, on upgrade, it will let the old
version exit before it starts the new version.
Greetings,
Frank
On 22.02.22 00:23, Austin Cawley-Edwards wrote:
Hey Marco,
There’s unfortunately no perfect fit here, at least that I know of. A
Deployment will make it possible to upgrade the image, but does not
support container exits (eg if the Flink job completes, even
successfully, K8s will still restart the container). If you are only
running long lived streaming jobs, this may be acceptable for you, but
makes it difficult to stop the job with a final savepoint (since it
will exit after completion).
What you could look into with the Job approach is building an upgrade
procedure that takes the final savepoint, allows the Job to exit, then
deploys a new helm release with the upgraded image and savepoint path.
It is more expensive, but may be more flexible.
Hope that helps,
Austin
On Mon, Feb 21, 2022 at 2:40 PM marco andreas
<marcoandreas...@gmail.com> wrote:
Hello flink community,
I am deploying a flink application cluster using a helm chart ,
the problem is that the jobmanager component type is a "Job" , and
with helm i can't do an upgrade of the chart in order to change
the application image version because helm is unable to
upgrade the docker image of the kind "Job" so i am wondering if i
can change the jobmanager kind to a "deployment" and if there are
any drawbacks in doing do.
Thanks.
