wehub-resource-sync
53 lines
1.2 KiB
YAML
53 lines
1.2 KiB
YAML
---
|
|
apiVersion: apps/v1
|
|
kind: DaemonSet
|
|
metadata:
|
|
name: lmcache-server
|
|
namespace: multi-process
|
|
spec:
|
|
selector:
|
|
matchLabels:
|
|
app: lmcache-server
|
|
template:
|
|
metadata:
|
|
labels:
|
|
app: lmcache-server
|
|
spec:
|
|
hostNetwork: true
|
|
volumes:
|
|
- name: dev-shm
|
|
hostPath:
|
|
path: /dev/shm
|
|
type: Directory
|
|
containers:
|
|
- name: lmcache-server
|
|
image: lmcache/standalone:nightly
|
|
command:
|
|
- /opt/venv/bin/python3
|
|
- -m
|
|
- lmcache.v1.multiprocess.server
|
|
- --host
|
|
- "0.0.0.0"
|
|
- --port
|
|
- "6555"
|
|
- --l1-size-gb
|
|
- "60"
|
|
- --eviction-policy
|
|
- LRU
|
|
- --max-workers
|
|
- "4"
|
|
volumeMounts:
|
|
- name: dev-shm
|
|
mountPath: /dev/shm
|
|
resources:
|
|
requests:
|
|
# 65Gi = 60GB l1-size-gb + ~5GB overhead for Python runtime and server operations
|
|
memory: "65Gi"
|
|
cpu: "4"
|
|
limits:
|
|
memory: "120Gi"
|
|
cpu: "8"
|
|
env:
|
|
- name: LMCACHE_LOG_LEVEL
|
|
value: "DEBUG"
|