Skip to content

Commit c1fa007

Browse files
shaneuttvMaroon
authored andcommitted
feat: add kubernetes dev deployment
Signed-off-by: Shane Utt <shaneutt@linux.com>
1 parent 8574f5f commit c1fa007

File tree

2 files changed

+55
-0
lines changed

2 files changed

+55
-0
lines changed
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
apiVersion: kustomize.config.k8s.io/v1beta1
2+
kind: Kustomization
3+
4+
namespace: ${NAMESPACE}
5+
6+
resources:
7+
- ../../../components/vllm-sim/
8+
- ../../../components/inference-gateway/
9+
10+
images:
11+
- name: quay.io/vllm-d/vllm-sim
12+
newName: ${VLLM_SIM_IMAGE}
13+
newTag: ${VLLM_SIM_TAG}
14+
- name: quay.io/vllm-d/gateway-api-inference-extension/epp
15+
newName: ${EPP_IMAGE}
16+
newTag: ${EPP_TAG}
17+
18+
patches:
19+
- path: patch-deployments.yaml
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,36 @@
1+
apiVersion: apps/v1
2+
kind: Deployment
3+
metadata:
4+
name: endpoint-picker
5+
spec:
6+
template:
7+
spec:
8+
imagePullSecrets:
9+
- name: ${REGISTRY_SECRET}
10+
containers:
11+
- name: epp
12+
args:
13+
- -poolNamespace
14+
- ${NAMESPACE}
15+
- -refreshMetricsInterval
16+
- "500ms"
17+
- -poolName
18+
- "vllm-llama3-8b-instruct"
19+
- -v
20+
- "4"
21+
- --zap-encoder
22+
- "json"
23+
- -grpcPort
24+
- "9002"
25+
- -grpcHealthPort
26+
- "9003"
27+
---
28+
apiVersion: apps/v1
29+
kind: Deployment
30+
metadata:
31+
name: vllm-sim
32+
spec:
33+
template:
34+
spec:
35+
imagePullSecrets:
36+
- name: ${REGISTRY_SECRET}

0 commit comments

Comments
 (0)