-
Notifications
You must be signed in to change notification settings - Fork 67
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
daskhub: provide worker resource options for 16CPU/128GB nodes on GKE/EKS #3344
Conversation
This comment was marked as resolved.
This comment was marked as resolved.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This looks good to me @consideRatio! Thank you!
I left a few comments about adding more notes about where certain numbers are coming from, so we can trace back in the future if need be.
Related to docs: do you believe it makes sense to add a small (rather generic) note about these new instance-dependent dask worker options in https://docs.2i2c.org/about/distributions/research/#a-brief-overview-of-scalable-computing-hubs?
a74ab02
to
b0f1283
Compare
Deployed to leap's staging for now, with a 👍 from @jbusecke who actively makes use of dask-gateway atm I'll go for a merge. |
8eda3d7
to
42de740
Compare
42de740
to
8ac8085
Compare
Rebased - I'll go for a merge at this point! |
🎉🎉🎉🎉 Monitor the deployment of the hubs here 👉 https://github.com/2i2c-org/infrastructure/actions/runs/6934459391 |
All GKE/EKS clusters with daskhubs have
n2-highmem-16
/r5.4xlarge
nodes setup for them (#2687). This makes us able to easily standardize pre-defined options to choose from for all clusters with regards to how large CPU/memory requests should be made for each pod.Review notes
daskhub's new default
The daskhub chart comes with these options by default now, adjusting to use
n2-highmem-16
orr5.4xlarge
based on the k8s distribution that is detected automatically.With this, jupyter-meets-the-earth no longer specifies its own custom logic.
basehub-cluster-info
ConfigMap addedIt is what makes us have the
BASEHUB_K8S_DIST
environment variable available for the hub and dask-gateway pod, where the dask-gateway pod relies on it in this PR and the hub pod got it mounted as its likely useful.The ConfigMap gets labels aligning with modern seemingly now stable helm chart best practices.
Test with
pange/base-notebook:latest
at https://dask-staging.2i2c.cloudUX for users
@jbusecke I've deployed this to leap's staging hub for now, if you have feedback its most welcome!