Call an optional webhook before killing a pod #149
Conversation
Miouge1
force-pushed
the
webhook
branch
3 times, most recently
from
e19e2a5 to
a08884a
Compare
|
It took a couple of tries, but I've added unit tests and fixed the comments from @golangcibot |
|
@Miouge1 Interesting idea. I'll take a look. |
|
Merged in latest master. |
|
@linki thanks for the rebase, what do you think about the PR? Anything I can do to help? |
| if resp.StatusCode == http.StatusOK { | ||
| filteredList = append(filteredList, pod) | ||
| } | ||
| } |
There was a problem hiding this comment.
This will make one request for each candidate pod. If you have a cluster with 1000 pods and none of them aren't filtered out by the previous filters, this will call your endpoint 1000 times.
Is that what you had in mind?
If we want to go with the current behaviour we should document it.
There was a problem hiding this comment.
Yes that was the idea, I understand that with a large number of pods that might not be workable.. Another option would be to call the webhook only for the randomly selected pod in Victim().
Either way would work for my use case, so maybe it's more wise to move things to Victim()? What do you think?
Call an optional POST webhook before killing a pod.
The webhook sends the PodSpec as a JSON payload to the webhook. This could allow an external webhook service to make an informed decision about killing the pod.
This could be used to have more complex logic or simply checking the Docker Hub status page.