problem history
[Ray Serve] LLM/GPU deployment stuck: "replicas that have taken more than 30s to be scheduled ... Resources required for each replica: {GPU: 1}, total resources available: {...}"
Revision reasons are not recorded in this slice.