fix(comfyui): port 8888, emptyDir storage, liveness probe #18

Merged
rock merged 2 commits from feat/comfyui-gpu-rebalance into main 2026-09-09 22:38:26 +00:00
Owner
  • Change container port from 8188 to 8888 (Caddy proxy binding)
  • Remove PVC, use emptyDir for ephemeral models/output
  • Replace startup + readiness probes with single liveness probe
  • Remove explicit COMFYUI_FLAGS (container defaults work)
  • Pod now reaches Ready state immediately after image pull

Fixes GPU contention by using ephemeral storage. ComfyUI now runs and is accessible at https://comfy.riotpiao.com

- Change container port from 8188 to 8888 (Caddy proxy binding) - Remove PVC, use emptyDir for ephemeral models/output - Replace startup + readiness probes with single liveness probe - Remove explicit COMFYUI_FLAGS (container defaults work) - Pod now reaches Ready state immediately after image pull Fixes GPU contention by using ephemeral storage. ComfyUI now runs and is accessible at https://comfy.riotpiao.com
rock added 2 commits 2026-09-09 11:37:38 +00:00
GPU rebalance (4× V100 32GB on worker-1):
- reasoning: 2 GPU (unchanged, PP=2 for Qwen3-32B)
- ornith: 2→1 GPU (scale to 1 replica, ornith:35b only)
- comfyui: 0→1 GPU (new)
- embeddings/reranker: 0 GPU (CPU, unchanged)

qwen2.5:3b-instruct moved to CPU on talos-cp-2 (144GB RAM).
Separate Ollama deployment + 5Gi PVC, pulls model on first start.

Gateway config updated in homelab-frontend (separate commit).

Co-authored-by: poimen <[email protected]>
In-cluster pods resolve comfy.riotpiao.com to nginx ingress
controller. Required for Cloudflare tunnel (cloudflared pod
resolves via CoreDNS).

Manual apply needed after merge:
  cd terraform && terraform apply -var-file=terraform.tfvars.local
  cd .. && make apply-cp

Co-authored-by: poimen <[email protected]>
rock merged commit 5b16b882be into main 2026-09-09 22:38:26 +00:00
rock deleted branch feat/comfyui-gpu-rebalance 2026-09-09 22:38:30 +00:00
Sign in to join this conversation.