-
Notifications
You must be signed in to change notification settings - Fork 39k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add support for custom stop signal to send to containers #30051
Comments
If STOPSIGNAL is already in the image, shouldn't the docker daemon kill the container with the right signal automatically? It'd only be broken if we bypass docker daemon or use a non-docker runtime. |
When we signal a container, do we just 'docker kill' or do we explicitly On Thu, Aug 4, 2016 at 12:51 PM, Yu-Ju Hong notifications@github.com
|
We use
|
ACK, then I think we can close this. thanks! On Thu, Aug 4, 2016 at 2:32 PM, Yu-Ju Hong notifications@github.com wrote:
|
This continues to be a problem for people who use pre-built images that handle SIGTERM by closing their listening socket. If they could simply change the signal to SIGUSR1, they would not have to deal with this. |
I realized that this request is not as clear as it could be. Dockerfile supports STOPSIGNAL and we support that transparently. However, not everyone builds their own images, and telling folks they have to build a custome image just to set this is pretty hostile. Docker supports |
Issues go stale after 90d of inactivity. If this issue is safe to close now please do so with Send feedback to sig-testing, kubernetes/test-infra and/or fejta. |
/lifecycle frozen |
This commit changes the default behaviour when stopping a Docker container from fast shutdowns to graceful ones. It is a potentially breaking change for some users and warrants a discussion. The background for this is that Kubernetes doesn't currently allow configuring the termination signal it sends to pods. For more on that, see kubernetes/kubernetes#30051 While this change increases the time it takes for a container to stop, Docker still sends an additional `SIGKILL` ten seconds after triggering the termination. This means that the default behaviour still ensures that the container is terminated _relatively_ quickly. There are potential workarounds if we decide against changing the Dockerfile. We could have a seperate image, use `PreStop` hooks, or or work around the issue in Teleport itself. However, this change is by far the simplest.
/kind feature |
@thockin it seems like this ticket is now misnamed. Perhaps you or someone could rename it so as not to confuse those who stumble upon it "Open". |
/retitle Add support for custom stop signal to send to containers |
How to get current |
@thockin Close this? dockerd is removed and kubelet reads the STOPSIGNAL from the image config: |
"Kubernetes has no equivalent as part of Pod or Container APIs. I posit that we should have something." |
Docker added STOPSIGNAL to let container images change which signal is delivered to kill the container. Should we support that? It would be bad if an image expected SIGUSR1 and we sent SIGTERM.
They support ints or strings. I think we should only support strings.
The text was updated successfully, but these errors were encountered: