No need to run device-plugin manually anymore. #25
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
It's an addon in GCP now. The manifest for that is here: https://github.com/kubernetes/kubernetes/blob/master/cluster/addons/device-plugins/nvidia-gpu/daemonset.yaml Instead of that a pause container is run, so that the installer can remain an init container. Also, added node affinity to only run on nodes that have a label with key 'cloud.google.com/gke-accelerator'. --- There are three files now: 1) device-plugin-daemonset.yaml 2) daemonset.yaml 3) daemonset-preloaded.yaml (1) already existed but its name no longer makes sense. (2) is copy of (1). (3) is copy of (2) except that the installer image is assumed to be present on the node. (1) would be deleted once tests in kubernetes/kubernetes are updated to point to (2) or (3).
jiayingz
reviewed
Nov 14, 2017
| name: nvidia-driver-installer | ||
| resources: | ||
| requests: | ||
| cpu: 0.15 |
There was a problem hiding this comment.
Nit: should we increase this based on what we measured in PR kubernetes/kubernetes#53541 that installer sometimes can take up to 2 cores?
There was a problem hiding this comment.
I am not sure. That would in effect reduce the capacity of all GPU nodes by 2 cores.
There was a problem hiding this comment.
I also have similar concern. What I am worried about is installer may take much longer time when the node is short of cpu. Perhaps we can discuss this later outside this PR.
/lgtm
k8s-github-robot
pushed a commit
to kubernetes/kubernetes
that referenced
this pull request
Nov 18, 2017
Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>. Update URLs for nvidia gpu device plugin and nvidia driver installer. Device plugin is now an addon and its manifest is now in kubernetes/kubernetes. The manifest on GoogleCloudPlatform/container-engine-accelerators no longer contains device plugin. This is needed after #54826 and GoogleCloudPlatform/container-engine-accelerators#25 **Release note**: ```release-note NONE ``` /sig scheduling
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
It's an addon in GCP now. The manifest for that is here:
https://github.com/kubernetes/kubernetes/blob/master/cluster/addons/device-plugins/nvidia-gpu/daemonset.yaml
Instead of that a pause container is run, so that the installer can remain an init container.
Also, added node affinity to only run on nodes that have a label with key 'cloud.google.com/gke-accelerator'.
There are three files now:
(1) already existed but its name no longer makes sense. (2) is copy of (1).
(3) is copy of (2) except that the installer image is assumed to be present on
the node. (1) would be deleted once tests in kubernetes/kubernetes are updated
to point to (2) or (3).