test: wire cf-orch agent registration into compose stack #48

New issue

Closed

opened 2026-03-31 19:06:14 -07:00 by pyr0ball · 1 comment

pyr0ball commented

2026-03-31 19:06:14 -07:00

Owner

Goal

Verify that Peregrine registers as a node with the cf-orch coordinator so GPU activity appears on the dashboard at http://10.1.10.71:7700/.

What to do

Add a cf-orch agent service to compose.override.yml (alongside the existing ollama/vllm adoption pattern)
The agent should start after the GPU services are up and register with the coordinator at http://heimdall:7700 (or CF_ORCH_COORDINATOR_URL env var)
Confirm peregrine’s GPU card appears on the dashboard with live VRAM readings
Run a cover letter generation task and confirm VRAM usage reflects on the dashboard

Config

cf-orch agent \
  --coordinator http://10.1.10.71:7700 \
  --node-id peregrine \
  --host 0.0.0.0 \
  --port 7701

Acceptance criteria

peregrine GPU card visible on dashboard
VRAM bar updates when vLLM/ollama is active
Agent reconnects after coordinator restart
No impact on existing task scheduler behaviour

## Goal Verify that Peregrine registers as a node with the cf-orch coordinator so GPU activity appears on the dashboard at `http://10.1.10.71:7700/`. ## What to do 1. Add a `cf-orch agent` service to `compose.override.yml` (alongside the existing ollama/vllm adoption pattern) 2. The agent should start after the GPU services are up and register with the coordinator at `http://heimdall:7700` (or `CF_ORCH_COORDINATOR_URL` env var) 3. Confirm peregrine’s GPU card appears on the dashboard with live VRAM readings 4. Run a cover letter generation task and confirm VRAM usage reflects on the dashboard ## Config ```bash cf-orch agent \ --coordinator http://10.1.10.71:7700 \ --node-id peregrine \ --host 0.0.0.0 \ --port 7701 ``` ## Acceptance criteria - [ ] `peregrine` GPU card visible on dashboard - [ ] VRAM bar updates when vLLM/ollama is active - [ ] Agent reconnects after coordinator restart - [ ] No impact on existing task scheduler behaviour

pyr0ball added this to the Paid Tier GA milestone 2026-04-04 16:33:18 -07:00

pyr0ball commented

2026-04-06 10:08:16 -07:00

Author

Owner

Implemented. Changes:

Added cf-orch-agent service to compose.yml on GPU profiles (single-gpu, dual-gpu-*)
Startup wrapper at docker/cf-orch-agent/start.sh — conditionally adds --advertise-host only when CF_ORCH_ADVERTISE_HOST is set (Docker Compose has no native conditional args)
Four new env vars in .env.example: CF_ORCH_COORDINATOR_URL, CF_ORCH_NODE_ID, CF_ORCH_AGENT_PORT, CF_ORCH_ADVERTISE_HOST
Defaults work for same-host coordinator (localhost:7700); set CF_ORCH_ADVERTISE_HOST=10.1.10.71 for remote coordinator

To test: copy the new vars into .env, set CF_ORCH_COORDINATOR_URL=http://10.1.10.71:7700 and CF_ORCH_ADVERTISE_HOST=<this machine LAN IP>, then make start PROFILE=single-gpu.

Implemented. Changes: - Added `cf-orch-agent` service to `compose.yml` on GPU profiles (`single-gpu`, `dual-gpu-*`) - Startup wrapper at `docker/cf-orch-agent/start.sh` — conditionally adds `--advertise-host` only when `CF_ORCH_ADVERTISE_HOST` is set (Docker Compose has no native conditional args) - Four new env vars in `.env.example`: `CF_ORCH_COORDINATOR_URL`, `CF_ORCH_NODE_ID`, `CF_ORCH_AGENT_PORT`, `CF_ORCH_ADVERTISE_HOST` - Defaults work for same-host coordinator (`localhost:7700`); set `CF_ORCH_ADVERTISE_HOST=10.1.10.71` for remote coordinator **To test:** copy the new vars into `.env`, set `CF_ORCH_COORDINATOR_URL=http://10.1.10.71:7700` and `CF_ORCH_ADVERTISE_HOST=<this machine LAN IP>`, then `make start PROFILE=single-gpu`.

pyr0ball closed this issue

2026-04-06 10:08:16 -07:00