pankajkoti commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1823904072
@amoghrajesh just if you wonder if Jed is jot able to respond anytime soon,
Jed is on parental leave 👶
--
This is an automated message from the Apache Git Service.
To respond to th
amoghrajesh commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1823829483
> I think we have ways to generate different resources for 1.28+ . I believe
helm takes K8s version into account when generating resources. Look in our
chart I think we have a few c
potiuk commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1820964266
> Thanks for letting me know about this. Was not very aware of this. If this
is the case, we should consider this as "THE ONLY" solution along with lock
file as you mentioned for earlier
amoghrajesh commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1820451617
>So I'd say we should treat native side-containers as THE solution and the
others are merely workarounds for versions of kubernwtes that do not have
native side containers yet.
amoghrajesh commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1820450496
>I also think native sidecar is the long term way to go. It's been long time
in the making - KEP for it has been discussed for 3 years at least - and it
handles many edge cases and
potiuk commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1819032429
I think we have ways to generate different resources for 1.28+ . I believe
helm takes K8s version into account when generating resources. Look in our
chart I think we have a few cases o
amoghrajesh commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1818986629
>we should figure out a different way how to communicate with the kerberos
refreshing process and stop it - one of the options is to let the running
process be stoppable by sending
amoghrajesh commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1818979974
Thank you for the detailed review @potiuk. This is extremely helpful to me.
> Another option (probably simpler to implement) is to keep some kind of
"shutdown" lock file in t
potiuk commented on PR #35748:
URL: https://github.com/apache/airflow/pull/35748#issuecomment-1818945067
Shared process namespace in this case kind of violates the whole assumption
the separate kerberos container here is introduced for. The whole idea is that
"airflow" containers do not see
amoghrajesh commented on code in PR #35748:
URL: https://github.com/apache/airflow/pull/35748#discussion_r1398696229
##
chart/templates/workers/worker-deployment.yaml:
##
@@ -98,6 +98,9 @@ spec:
{{- toYaml $podAnnotations | nindent 8 }}
{{- end }}
spec:
amoghrajesh opened a new pull request, #35748:
URL: https://github.com/apache/airflow/pull/35748
As a part of the effort to enable kerberos to work e2e with airflow in a
kubernetes deployment, some changes related to the pod template and worker
deployment have gone in recent
11 matches
Mail list logo
