gpu-manager
gpu-manager copied to clipboard
No metric collected when no "tencent.com/vcuda-core" assigned
I deployed my app via SeldonCore and Triton Server in an k8s, and found no metrics collected when tencent.com/vcuda-core is skipped in deployment.yaml. Is that a bug or the tencent.com/vcuda-core must be configed for each deployment?
### metrics are collected when vcude-core set
resources:
limits:
tencent.com/vcuda-core: 50
tencent.com/vcuda-memory: 3
requests:
tencent.com/vcuda-core: 50
tencent.com/vcuda-memory: 3
### metrics are not collected when no vcuda-core set
resources:
limits:
tencent.com/vcuda-memory: 3
requests:
tencent.com/vcuda-memory: 3