Homepage dashboard — Phase 2/3 follow-ups #68

New issue

Open

opened 2026-09-10 19:06:16 +00:00 by claude-bot · 0 comments

claude-bot commented

2026-09-10 19:06:16 +00:00

Collaborator

Tracking follow-ups after ccnet-homepage-swarm/ Phase 1 (PR #67) is deployed.

Phase 2 — live Docker / Swarm discovery

Add a pinned ghcr.io/tecnativa/docker-socket-proxy to the stack, read-only
(CONTAINERS/SERVICES/TASKS/NODES/NETWORKS/INFO=1, POST=0), node.role == manager.
config/docker.yaml → swarm: true against the proxy over the overlay.
Keep services.yaml authoritative for tiles; Docker supplies only the
running/stopped dot (no deploy.labels: churn across 6 stacks).
Optional: a second socket-proxy inside the mailcow compose stack + a
docker-compose.override.yml, exposed on the vlan50 IP, to show mailcow's
~10 containers as a group. Needs the Step 6 firewall question resolved first.

Phase 3 — as other work lands

*arr widgets (Sonarr/Radarr/Prowlarr/SABnzbd/qBittorrent) when
docs/arr-media-stack.md comes off hold — add tiles to services.yaml.
prometheus / grafana widgets + a linked Ceph panel once
docs/ccnet-monitoring-stack-design.md is built.
Consider a glances agent per Proxmox node + Unraid for per-host resource
tiles (the closest thing to the Unraid-GUI feel for the hypervisors).

Tooling

Teach service_onboard.py a --homepage-group / --homepage-widget flag so
the "add a Homepage tile" convention step is automated alongside DNS/Caddy.

Open question carried from the plan

Whether the swarm can already reach 10.10.50.10:443 (mailcow, vlan50 DMZ),
and if not, whether a narrow pass rule is warranted. Check + decide before
the mailcow widget will work.

Tracking follow-ups after `ccnet-homepage-swarm/` Phase 1 (PR #67) is deployed. ## Phase 2 — live Docker / Swarm discovery - Add a pinned `ghcr.io/tecnativa/docker-socket-proxy` to the stack, read-only (`CONTAINERS/SERVICES/TASKS/NODES/NETWORKS/INFO=1`, `POST=0`), `node.role == manager`. - `config/docker.yaml` → `swarm: true` against the proxy over the overlay. - Keep `services.yaml` authoritative for tiles; Docker supplies only the running/stopped dot (no `deploy.labels:` churn across 6 stacks). - Optional: a second socket-proxy *inside the mailcow compose stack* + a `docker-compose.override.yml`, exposed on the vlan50 IP, to show mailcow's ~10 containers as a group. Needs the Step 6 firewall question resolved first. ## Phase 3 — as other work lands - *arr widgets (Sonarr/Radarr/Prowlarr/SABnzbd/qBittorrent) when `docs/arr-media-stack.md` comes off hold — add tiles to `services.yaml`. - `prometheus` / `grafana` widgets + a linked Ceph panel once `docs/ccnet-monitoring-stack-design.md` is built. - Consider a `glances` agent per Proxmox node + Unraid for per-host resource tiles (the closest thing to the Unraid-GUI feel for the hypervisors). ## Tooling - Teach `service_onboard.py` a `--homepage-group` / `--homepage-widget` flag so the "add a Homepage tile" convention step is automated alongside DNS/Caddy. ## Open question carried from the plan - Whether the swarm can already reach `10.10.50.10:443` (mailcow, vlan50 DMZ), and if not, whether a narrow pass rule is warranted. Check + decide before the mailcow widget will work.

claude-bot referenced this issue

2026-09-17 20:15:37 +00:00

ccnet-homepage-swarm: internal-only Homepage dashboard (Phase 1) #67