set up for new gpu class #19
Conversation
memalhot
force-pushed
the
gpu_class
branch
4 times, most recently
from
55c27ca to
45a2bde
Compare
|
This is a good start! Left some inline comments. In addition, could you rename the |
gpu-class/resource.yaml
Outdated
| annotations: | ||
| notebooks.opendatahub.io/inject-oauth: 'true' | ||
| notebooks.opendatahub.io/last-image-selection: ${IMAGE_NAME} | ||
| notebooks.opendatahub.io/last-size-selection: X Small |
There was a problem hiding this comment.
I think it should be Small
gpu-class/resource.yaml
Outdated
| preferredDuringSchedulingIgnoredDuringExecution: | ||
| - preference: | ||
| matchExpressions: | ||
| - key: nvidia.com/gpu.present |
There was a problem hiding this comment.
I don't think we need gpu in dev notebook.
gpu-class/resource.yaml
Outdated
| - Tesla-V100-PCIE-32GB | ||
| weight: 1 | ||
| containers: | ||
| - resources: |
There was a problem hiding this comment.
The resources for Small container are: Limits: 2 CPU, 8GiB Memory; Requests: 1 CPU, 8GiB Memory
gpu-class/rb.yaml
Outdated
| name: ${SERVICE_ACCOUNT} | ||
| namespace: ${NAMESPACE} | ||
|
|
||
| # CREATE CLUSTERQUEUE ROLE |
There was a problem hiding this comment.
the cluster role is a static role. You'll just need to create once rather than create each clusterrole per namespace.
gpu-class/gpu-class-setup.sh
Outdated
| namespace=$1 | ||
|
|
||
| #get student sername from namespace | ||
| username=$(echo "$ns" | awk -F'-' '{print $2}') |
There was a problem hiding this comment.
It might be possible that TAs and instructors also are added to student namespace.
There was a problem hiding this comment.
should it be $namespace instead of $ns?
memalhot
force-pushed
the
gpu_class
branch
3 times, most recently
from
d3009e9 to
9f7fcff
Compare
gpu-class/rb.yaml
Outdated
| - apiVersion: rbac.authorization.k8s.io/v1 | ||
| kind: RoleBinding | ||
| metadata: | ||
| name: edit-jobs |
There was a problem hiding this comment.
it might be better to name the RB as ${SERVICE_ACCOUNT_NB}-edit-jobs, in case when a new RB is created, it will raise an error indicating RB name already exists
gpu-class/rb.yaml
Outdated
| roleRef: | ||
| apiGroup: rbac.authorization.k8s.io | ||
| kind: Role | ||
| name: default-kueue-localqueue-reader |
There was a problem hiding this comment.
I'd rename it to ${SERVICE_ACCOUNT_NB}-kueue-localqueue-reader
…int to clusterqueues, and observability for jobs through rolebinding
setting up for new gpu class by creating notebooks for students to use, setting up localqueue so they can submit gpu jobs, and setting up rolebindings so students can see what nodes are using gpus, what jobs are running, etc