ecerulm commented on issue #21087:
URL: https://github.com/apache/airflow/issues/21087#issuecomment-1121343043
> once that issue is fixed you kube client, you won't be getting 410 errors
, **at least not so frequently**, when you do a watch.
so like I said, we will still see 410 errors. If we handle 410 in
kubernetes_executor.py this issue will be solved regardless of kubernetes
python client fixes. So those fixes are inconsequential for this issue
> in the rare case of a 410 error.
well, it's not that rare.
> like i mentioned in my previous comment. BOOKMARK is required for
preventing 410 error.
I thought that we agreed that this is not preventing 410, you will still get
it if enough time passed between the disconnect and the when the next watch is
goes through because at that time the resource version that
kuberentes_executor.py is sending in the watch has already expired. Again, the
bookmark thing is not preventing that at all.
I think it's going to be very confusing for people reading this thread and
thinking that some fix is required in kubernetes python client , because that
fix is not required at all to fix this issue, the only thing needed to fix this
issue is to handle 410 properly.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
