Add Slurm autoscaling #18
Draft
Conversation
This file contains hidden or 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
Convert Rook NFS to a Helm chart and install it as a Slurm chart dependency
Moved secret generation from scripts to helm
Only permit one slurmd pod per k8s node
Use host networking
Replaced mounted kubeconfig with service account
|
NB: this CANNOT make use of #27 |
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Outdated suggestions cannot be applied.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.
Requires
KUBECONFIGto be defined in the shell used to runhelm. This is injected as a secret to use to create pods on demand.Uses host network for slurmd pods. These pods have a hostPort defined (for slurmd) which means they don't get scheduled onto the same k-node.
Note there's no ResumeFailProgram defined; assuming the slurmd pod definition is ok the the most likely reason for "resume failure" from slurm's PoV is the pod pending beyond ResumeTimeout, due to e.g. not enough k-nodes. While the s-node does then show DOWN, if k-resources become available later the pod will launch, at which time the s-node changes from DOWN to IDLE (this has been tested). Note this is different from e.g. autoscale on openstack, where a failed VM launch due to cloud resources will not keep retrying and the s-node state needs to be reset to allow slurm to try to launch it again. So really leaving the s-node showing as DOWN accurately reflects the state of cloud resources here.