Scenarios
I'm deploying ai Inference service in k8s.So I need to make container to use rdma capability.
When I'm using host-device cni plugin to move eight network devices to container, I got an error message "failed (add): netplugin failed with no error message: signal: killed".
After troubleshooting,the reason is the cni process is too long, but CNITimeoutSec(default 220sec) for cri-dockerd is not enough.
So I think maybe wo should make CNITimeoutSec configurable?
fix
#491
Scenarios
I'm deploying ai Inference service in k8s.So I need to make container to use rdma capability.
When I'm using host-device cni plugin to move eight network devices to container, I got an error message "failed (add): netplugin failed with no error message: signal: killed".
After troubleshooting,the reason is the cni process is too long, but CNITimeoutSec(default 220sec) for cri-dockerd is not enough.
So I think maybe wo should make CNITimeoutSec configurable?
fix
#491