Jiaxin Shan
Jiaxin Shan
Engineer support for R1 issue has been done. We can close this issue.
1. what's your kubernetes offering? are you on public cloud or on-prem cluster? 2. could you describe `kubectl describe svc envoy-aibrix-system-aibrix-eg-903790dc -n envoy-gateway-system` to check the service pending information? Seems...
@AlexHe99 can you manually change `LoadBalancer` type to `NodePort` type and use host IP + nodePort to continue the testing? Since this is on-prem cluster, I think the kubernetes distribution...
/cc @varungup90 please help take a look at this issue
the problem is it didn't disable the probe injection successfully. For most of the case, this is ok but for large model, probe may kill the pod and can not...
@varungup90 what's the status of this issue?
this is related to multi-tenancy, let's move from v0.4.0 milestone
@jolfr Simplifying the deployment experience would be super helpful. I actually do not have preference at this moment, I personally use GCP/AWS and our company use volcano engine cloud for...
@googs1025 We can check other solutions as references.
We should use implicitly append `vllm:`, this is really confusing, we should ask user to type the full name to avoid the conversion