[iris] Grant OpenAthena principals full CoreWeave budgets - #8776
Merged
Conversation
Give the seven named OpenAthena identities unlimited INTERACTIVE budgets on every CoreWeave cluster. Restore the unlisted-user default to 1000 so excess work falls to BATCH and the core tier can reclaim cluster capacity.
Contributor
IaC preview
|
Allow one GPU node to stay INTERACTIVE for unlisted CoreWeave users. The H100 configs use 10688 and the GB200 config uses 5680, matching Iris resource_value; additional capacity degrades to BATCH.
Use jobs.submitting_user for budget admission, active-task caps, spend, scheduling fairness, demotion, Kubernetes dispatch, and dashboard buckets. Preserve the friendly job-path owner as the fallback for trusted local and pre-migration rows. Add ben.feuer@openathena.ai to the unlimited INTERACTIVE tier on every CoreWeave cluster.
Describe the fallback in terms of empty submitting_user rows, remove a stale incident reference, and keep the regression tests focused on behavior.
Place each established core job-path alias next to its authenticated OpenAthena email in the unlimited INTERACTIVE tier. Leave non-core aliases in the bounded researcher tier and remove the superseded Benjamin Feuer override.
Use canonical or local identity constants in budget-path tests and keep the budget selector documentation proportional to its behavior.
rjpower
enabled auto-merge (squash)
August 29, 2026 21:10
rjpower
disabled auto-merge
August 29, 2026 21:15
yonromai
reviewed
Aug 29, 2026
| - russell.power@openathena.ai | ||
| - power | ||
| - will.held@openathena.ai | ||
| - held |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Grant both the authenticated email and established job-path alias for each of eight OpenAthena principals an unlimited INTERACTIVE budget on all four CoreWeave clusters. A zero limit keeps their work from degrading to BATCH, so each principal can claim the full cluster after other users exceed the guest allowance.
Live Marin rows paired nickname job paths such as
/power/...with authenticated emails injobs.submitting_user, while budget consumers still used the nickname. Iris now uses the authenticated principal for max-band admission, active-task caps, spend, scheduling fairness and demotion, Kubernetes dispatch, and scheduler dashboard buckets. Friendly job paths remain unchanged. Trusted local jobs and rows with an emptysubmitting_usercontinue to use the job-path owner.Set the unlisted-user limit to one GPU node: 10688 on the H100 clusters and 5680 on the GB200 cluster. Additional unlisted work becomes BATCH and can be preempted by the core tier. Non-core aliases remain in their bounded researcher tier. The prior Benjamin Feuer override is superseded by the unlimited core tier. The roster is explicit in each cluster config because Iris has no cross-file config include.
The budget tiers apply when each controller next starts, and authenticated-principal accounting applies after the updated controller is deployed. No cluster was restarted. The investigation record is https://echo.oa.dev/wiki/288.