Use generated resource allocation options for veda hub #3566
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
While starting to work on 2i2c-org#3565, I realized that VEDA was still using the older style 'node share' rather than the generated 'resource allocation' options. I've swapped over the options to now be based on images for users to choose and resource allocation options generated by our resource allocation script. This matches openscapes, and there has generally been pretty big positive feedback on this mode. I've kept the initial cloning to only happen on the pangeo image as it currently exists, without making any changes. That should be cleaned up as part of 2i2c-org#3565
|
Merging this PR will trigger the following deployment actions. Support and Staging deployments
Production deployments
|
|
I'd suggest just rounding down the numbers to the nearest 0.5? |
|
Positive feedback from Alex Mandel on slack: 
Although there was a suggestion to do better rounding of the numbers. |
|
This does seems like an improvement, though I wonder why more than one section is necessary: Also linking to external descriptions/docs on the what software is installed or how to use the custom image choice. |
|
@wildintellect that issue is pending jupyterhub/kubespawner#778 getting fixed. |
sgibson91
approved these changes
Jan 4, 2024
|
I've approved pending any tweaks asked for by the community :) |
|
I'll open another issue to deal with the rounding as it's reasonably complex, and deploy this now. The community seemed to not object to this on slack :) |
|
🎉🎉🎉🎉 Monitor the deployment of the hubs here 👉 https://github.com/2i2c-org/infrastructure/actions/runs/7413484076 |
yuvipanda
added a commit
to yuvipanda/pilot-hubs
that referenced
this pull request
Jan 4, 2024
2i2c-org#3569 changed the cryptnono daemonset to have different resource requests for the init containers as well as the container. While working on 2i2c-org#3566, I noticed this was generating wrong choices - the overhead was calculated wrong (too small). We were intentionally ignoring init containers while calculating overhead, and turns out the scheduler and the autoscaler both do take it into consideration. The effective resource request for a pod is the higher of the resource requests for the containers *or* the init containers - this ensures that a pod with higher requests for init containers than containers (like our cryptnono pod!) will actually run. This is documented at https://kubernetes.io/docs/concepts/workloads/pods/init-containers/#resource-sharing-within-containers, and implemented in Kubernetes itself at https://github.com/kubernetes/kubernetes/blob/9bd0ef5f173de3cc2d1d629a4aee499d53690aee/pkg/api/v1/resource/helpers.go#L50 (this is the library code that the cluster autoscaler uses). This PR updates the two places we currently have that calculate effective resource requests (I assume eventually these will be merged into one - I haven't kept up with the team's work last quarter here). I've updated the node-capacity-info.json file, which is what seems to be used by the generator script right now.
yuvipanda
added a commit
to yuvipanda/pilot-hubs
that referenced
this pull request
Jan 4, 2024
2i2c-org#3569 changed the cryptnono daemonset to have different resource requests for the init containers as well as the container. While working on 2i2c-org#3566, I noticed this was generating wrong choices - the overhead was calculated wrong (too small). We were intentionally ignoring init containers while calculating overhead, and turns out the scheduler and the autoscaler both do take it into consideration. The effective resource request for a pod is the higher of the resource requests for the containers *or* the init containers - this ensures that a pod with higher requests for init containers than containers (like our cryptnono pod!) will actually run. This is documented at https://kubernetes.io/docs/concepts/workloads/pods/init-containers/#resource-sharing-within-containers, and implemented in Kubernetes itself at https://github.com/kubernetes/kubernetes/blob/9bd0ef5f173de3cc2d1d629a4aee499d53690aee/pkg/api/v1/resource/helpers.go#L50 (this is the library code that the cluster autoscaler uses). This PR updates the two places we currently have that calculate effective resource requests (I assume eventually these will be merged into one - I haven't kept up with the team's work last quarter here). I've updated the node-capacity-info.json file, which is what seems to be used by the generator script right now.
|
@wildintellect I opened #3584 with the explanation for why we show the odd numbers in memory. |
yuvipanda
added a commit
to yuvipanda/pilot-hubs
that referenced
this pull request
Jan 11, 2024
Brings in 2i2c-org#3566 (and follow-ups) to the GHG hub Ref 2i2c-org#3565
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.
While starting to work on #3565, I realized that VEDA was still using the older style 'node share' rather than the generated 'resource allocation' options. I've swapped over the options to now be based on images for users to choose and resource allocation options generated by our resource allocation script. This matches openscapes, and there has generally been pretty big positive feedback on this mode. It also gives more visibility to the R & QGIS options.
I've kept the initial cloning to only happen on the pangeo image as it currently exists, without making any changes. That should be cleaned up as part of #3565
I also had to update the node-info.json file by running the appropriate command.
Old:
New: