Skip to content

Add LLMInferenceService support #5

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.

Sign up for GitHub

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

Merged
wseaton merged 2 commits into from
Aug 31, 2026
Merged

Add LLMInferenceService support #5

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4722f96
add LLMInferenceService support (serving.kserve.io/v1alpha1)
wseaton Apr 1, 2026
ed11ce7
resolve LLMISVC-owned LeaderWorkerSets, flatten owner-chain walk
wseaton Aug 31, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# gpu-pruner

The `gpu-pruner` is a non-destructive idle culler that works with Red Hat OpenShift AI/Kubeflow provided APIs (`InferenceService` and `Notebook`), as well as generic `Deployment`, `ReplicaSet`, `StatefulSet` and `LeaderWorkerSet` (`leaderworkerset.x-k8s.io`).
The `gpu-pruner` is a non-destructive idle culler that works with Red Hat OpenShift AI/Kubeflow provided APIs (`InferenceService` and `Notebook`), as well as generic `Deployment`, `ReplicaSet`, `StatefulSet`, `LeaderWorkerSet` (`leaderworkerset.x-k8s.io`) and `LLMInferenceService` (`serving.kserve.io/v1alpha1`).

The way it works is by querying cluster NVIDIA DCGM metrics and looking at a window of GPU utilization per pod. A scaling decision is made by looking up the pods metadata, and using owner-references to figure out the owning resource.

Expand Down Expand Up @@ -33,9 +33,9 @@ Options:
-e, --enabled-resources <ENABLED_RESOURCES>
Specifcy enabled resources with a string of letters

- `d` for Deployment - `r` for ReplicaSet - `s` for StatefulSet - `i` for InferenceService - `n` for Notebook - `l` for LeaderWorkerSet
- `d` for Deployment - `r` for ReplicaSet - `s` for StatefulSet - `i` for InferenceService - `n` for Notebook - `l` for LeaderWorkerSet - `m` for LLMInferenceService

[default: drsinl]
[default: drsinlm]

-c, --check-interval <CHECK_INTERVAL>
interval in seconds to check for idle pods, only used in daemon mode
Expand Down
Loading
Loading