Repository navigation
fix: lower proxy CPU request to 950m so a pod stays under 1 CPU - #141
Merged
Merged
Conversation
Resource accounting is per pod: the open-llm-proxy container requested 1000m and the dns-cache sidecar another 10m, so every replica registered as 1.01 CPU. Drop the app request to 950m (pod total 960m). Limits are untouched, so burst headroom is unchanged.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Replicas were registering at 1.01 CPU because resource accounting is per pod, not per container:
open-llm-proxydns-cachesidecarLower the app request to
950m, putting the pod total at 960m — strictly under 1 CPU.Limits are unchanged (2000m app / 200m sidecar), so burst headroom is the same; only the scheduling/quota floor moves.
Applies on the next rollout (both deployments git-clone
mainat pod boot, but this is a manifest change, so it needskubectl apply -f deployment.yamlon NRP). cirrus has its owncirrus/deployment.yaml(250m request) and is unaffected.