Access sites — Argo CD, Forgejo, Nextcloud, Euro-Office and Grafana

This guide describes how to reach the services after the installation in installation-steps.md has completed.

The cluster lives on the isolated 10.10.50.0/24 network. Only the management container is dual-homed, so it is the single entry point from the home network:

ServiceURL on the home networkHow it is exposed
Argo CDhttps://192.168.0.192:8080kubectl port-forward on the management container
Forgejohttp://192.168.0.192:3000kubectl port-forward on the management container
nginx demohttp://nginx-demo.kubernetes.localTraefik NodePort + systemd socket proxy on the management container
Nextcloudhttp://nextcloud.kubernetes.localTraefik ingress + the same socket proxy
Euro-Office Docshttp://euro-office.kubernetes.localTraefik ingress + the same socket proxy
Grafanahttp://grafana.kubernetes.localTraefik ingress + the same socket proxy
Talos / Kubernetes APIhttps://10.10.50.22:6443only from the management container (or its LAN side)
Forgejo Git (for Argo CD)http://forgejo-http.forgejo.svc.cluster.local:3000/gitea_admin/gitops.gitin-cluster DNS, not resolvable from the home network

192.168.0.192 is the management container’s DHCP address on the home network. It was 192.168.0.201 during the initial installation; the lease changed on 2026-10-29. If it changes again, look it up in the Proxmox UI (LXC 105 → Network) or on the container with ip -4 -o addr show eth0. The OpenTofu state can be stale after a DHCP change; the Proxmox UI is authoritative.

When the address changes, also update ansible2/hosts.yml and the traefik-ingress-http.socket / traefik-ingress-https.socket units (or re-run ansible2/traefik.yml), otherwise port 80/443 stays bound to the old address.


1. Argo CD

Argo CD’s service is ClusterIP, so it is not directly reachable from the home network. Forward its HTTPS port from the management container:

ssh root@192.168.0.192 \
  'kubectl -n argocd port-forward --address 0.0.0.0 svc/argocd-server 8080:443'

Keep that SSH session open and browse to:

https://192.168.0.192:8080
ssh root@192.168.0.192 \
  "kubectl -n argocd get secret argocd-initial-admin-secret \
   -o jsonpath='{.data.password}' | base64 -d; echo"

The password is stored in the cluster, not in this repository.

Alternative without exposing the port on the LAN — bind the forward to localhost on the management container and tunnel it over SSH:

# terminal 1
ssh root@192.168.0.192 'kubectl -n argocd port-forward svc/argocd-server 8080:443'

# terminal 2
ssh -L 8080:127.0.0.1:8080 root@192.168.0.192

Then open https://localhost:8080.

The port-forward stops when the SSH session ends. For a long-lived setup, run it inside tmux/screen, or add a small systemd unit on the management container.


2. Forgejo

Forgejo is also a ClusterIP service (forgejo-http, port 3000), so use a port-forward as well:

ssh root@192.168.0.192 \
  'kubectl -n forgejo port-forward --address 0.0.0.0 svc/forgejo-http 3000:3000'

Browse to:

http://192.168.0.192:3000

Admin credentials (the chart generates them once, in the forgejo-admin secret):

ssh root@192.168.0.192 \
  "kubectl -n forgejo get secret forgejo-admin \
   -o jsonpath='{.data.username}' | base64 -d; echo"

ssh root@192.168.0.192 \
  "kubectl -n forgejo get secret forgejo-admin \
   -o jsonpath='{.data.password}' | base64 -d; echo"

The GitOps repository Argo CD reads is:

http://forgejo-http.forgejo.svc.cluster.local:3000/gitea_admin/gitops.git

That hostname only resolves inside the cluster (Argo CD’s repo-server uses it). From your workstation, use the Forgejo web UI above.

Git over SSH (optional). Forward the SSH service to a high port and use it:

ssh root@192.168.0.192 \
  'kubectl -n forgejo port-forward --address 0.0.0.0 svc/forgejo-ssh 2222:22'

git clone ssh://git@192.168.0.192:2222/gitea_admin/gitops.git

Forgejo is deliberately not published through Traefik in this setup; only the four application hostnames in section 3 use the ingress path.


3. Kubernetes ingresses (*.kubernetes.local)

Four hostnames are served through Traefik on the management container:

HostnameApplication
nginx-demo.kubernetes.localnginx demo
nextcloud.kubernetes.localNextcloud 32
euro-office.kubernetes.localEuro-Office Docs
grafana.kubernetes.localGrafana

Where does the /etc/hosts entry go?

On the machine where the browser (or curl) runs — not on the Proxmox host.

/etc/hosts is a per-machine file. It is consulted by the operating system that makes the request. Your laptop asks “what is the IP of nextcloud.kubernetes.local?”, so the entry must exist on your laptop. Adding it to 192.168.0.141 (the Proxmox host) only affects commands executed on the Proxmox host itself; it does not change name resolution for your laptop, phone, or any other device.

Where you browse fromWhere to add the entry
Your laptop/desktop (normal case)/etc/hosts on that laptop/desktop
The Proxmox host shell (curl there)/etc/hosts on the Proxmox host
A phone/tabletNot possible without root; use the Host header trick below or local DNS
Whole home networkAdd a DNS record on your router (better than per-device hosts files)

Linux / macOS

Add all four names on one line:

sudo sh -c 'echo "192.168.0.192  nginx-demo.kubernetes.local nextcloud.kubernetes.local euro-office.kubernetes.local grafana.kubernetes.local" >> /etc/hosts'

Verify:

getent hosts nginx-demo.kubernetes.local nextcloud.kubernetes.local euro-office.kubernetes.local grafana.kubernetes.local    # Linux
dscacheutil -q host -a name grafana.kubernetes.local   # macOS (repeat per name)

Windows

Edit C:\Windows\System32\drivers\etc\hosts as Administrator and add:

192.168.0.192nginx-demo.kubernetes.localnextcloud.kubernetes.localeuro-office.kubernetes.localgrafana.kubernetes.local

Then flush the DNS cache in an Administrator prompt:

ipconfig /flushdns

Browse

URLExpected
http://nginx-demo.kubernetes.localthe “nginx demo — ArgoCD” page
http://nextcloud.kubernetes.localNextcloud login page (/login)
http://euro-office.kubernetes.localredirect to /welcome/, title “Euro-Office Docs Community Edition”
http://euro-office.kubernetes.local/example/example/test page, title “Euro-Office Document editors”
http://grafana.kubernetes.localGrafana login page

Nextcloud admin credentials:

ssh root@192.168.0.192 \
  "kubectl -n nextcloud get secret nextcloud-admin \
   -o jsonpath='{.data.username}' | base64 -d; echo"
ssh root@192.168.0.192 \
  "kubectl -n nextcloud get secret nextcloud-admin \
   -o jsonpath='{.data.password}' | base64 -d; echo"

The username is normally admin. Euro-Office itself has no web login; the /welcome/ page is just the document server landing page. Documents are opened from Nextcloud, which authenticates the user and signs the document-server request.

Grafana admin credentials are in the stable monitoring-grafana-admin Secret, not in Git:

ssh root@192.168.0.192 \
  "kubectl -n monitoring get secret monitoring-grafana-admin \
   -o jsonpath='{.data.admin-user}' | base64 -d; echo"
ssh root@192.168.0.192 \
  "kubectl -n monitoring get secret monitoring-grafana-admin \
   -o jsonpath='{.data.admin-password}' | base64 -d; echo"

See monitoring.md for the Prometheus/Grafana architecture, retention and verification commands.

Without editing /etc/hosts

Send the Host header manually:

curl -H 'Host: nginx-demo.kubernetes.local' http://192.168.0.192/
curl -I -H 'Host: nextcloud.kubernetes.local' http://192.168.0.192/
curl -I -H 'Host: euro-office.kubernetes.local' http://192.168.0.192/
curl -I -H 'Host: grafana.kubernetes.local' http://192.168.0.192/

or let curl resolve the name for one request. Add -L when the application redirects, otherwise the second request tries to resolve the hostname normally:

curl -sSL --resolve nginx-demo.kubernetes.local:80:192.168.0.192 \
  http://nginx-demo.kubernetes.local/
curl -sSL --resolve nextcloud.kubernetes.local:80:192.168.0.192 \
  -o /dev/null -w '%{http_code} %{url_effective}\n' \
  http://nextcloud.kubernetes.local/
curl -sSL --resolve euro-office.kubernetes.local:80:192.168.0.192 \
  -o /dev/null -w '%{http_code} %{url_effective}\n' \
  http://euro-office.kubernetes.local/
curl -sSL --resolve grafana.kubernetes.local:80:192.168.0.192 \
  -o /dev/null -w '%{http_code} %{url_effective}\n' \
  http://grafana.kubernetes.local/login

How the request flows

b11<r90no│▼2│▼.│▼aw.1ms10eeD6s.TsrN8y5rpS.s0aao:0t.ecn.e2fe<1m2i>yh9d:k/oo2-3<us:s0rsrt8o0oen0c8urlak0tvameeipetsct>-eopo>prnoxt1yh9de2.(H1mtToI6arrsn8naatg.aeer0gffhe.eiies1mkkas9e-d2niNetnorgdcreS/oePeensorttsrvca-ti/ihchntoeoetnsrpt,.tssa)selPyroosvsdti-ecnmeod)des-o3cket
HostnameIngressService
nginx-demo.kubernetes.localnginx-demo/nginxnginx-demo/nginx
nextcloud.kubernetes.localnextcloud/nextcloudnextcloud/nextcloud:8080
euro-office.kubernetes.localeuro-office/documentservereuro-office/documentserver:8888
grafana.kubernetes.localmonitoring/monitoring-grafanamonitoring/monitoring-grafana:80

The HTTPS equivalent is 192.168.0.192:443 → 10.10.50.22:30443. The socket proxy is managed by ansible2/traefik.yml and runs as traefik-ingress-http.socket / traefik-ingress-https.socket on the management container.

Verify from the management container

ssh root@192.168.0.192 \
  "curl -s -o /dev/null -w '%{http_code}\n' \
   -H 'Host: nginx-demo.kubernetes.local' http://192.168.0.192/"
ssh root@192.168.0.192 \
  "curl -s -o /dev/null -w '%{http_code}\n' \
   -H 'Host: nextcloud.kubernetes.local' http://192.168.0.192/"
ssh root@192.168.0.192 \
  "curl -s -o /dev/null -w '%{http_code}\n' \
   -H 'Host: euro-office.kubernetes.local' http://192.168.0.192/"
ssh root@192.168.0.192 \
  "curl -s -o /dev/null -w '%{http_code}\n' \
   -H 'Host: euro-office.kubernetes.local' \
   http://192.168.0.192/healthcheck"
ssh root@192.168.0.192 \
  "curl -s -o /dev/null -w '%{http_code}\n' \
   -H 'Host: grafana.kubernetes.local' \
   http://192.168.0.192/login"

Use 192.168.0.192, not 127.0.0.1: the systemd socket binds specifically to the container’s LAN address, so localhost refuses the connection. Expected: 200, 302, 302, 200, 200.


4. Quick reference

# Management container IP (run on the control host)
tofu output -json management \
  | python3 -c 'import json,sys; print(json.load(sys.stdin)["wan_ips"]["eth0"])'

# Argo CD
ssh root@192.168.0.192 'kubectl -n argocd port-forward --address 0.0.0.0 svc/argocd-server 8080:443'
# -> https://192.168.0.192:8080  (admin / secret argocd-initial-admin-secret)

# Forgejo
ssh root@192.168.0.192 'kubectl -n forgejo port-forward --address 0.0.0.0 svc/forgejo-http 3000:3000'
# -> http://192.168.0.192:3000   (secret forgejo-admin)

# nginx demo (after the /etc/hosts entry on the client)
# browse to http://nginx-demo.kubernetes.local

# Nextcloud
# browse to http://nextcloud.kubernetes.local (admin / secret nextcloud-admin)

# Euro-Office Docs
# browse to http://euro-office.kubernetes.local
# example/test page: http://euro-office.kubernetes.local/example/
# documents are opened from Nextcloud; there is no separate login

# Grafana
# browse to http://grafana.kubernetes.local
# admin / secret monitoring-grafana-admin; see monitoring.md

# Garage S3 backup target (internal only; not a browser UI)
# from the management container:
# curl -sS -o /dev/null -w '%{http_code}\n' http://10.10.50.2:3900/
# expected: 403 without credentials; see backup.md and backup-restore-howto.md

5. Troubleshooting

SymptomCheck
bind: address already in useAnother port-forward is still running; kill it or pick another local port (8081:443, 3001:3000).
Argo CD/Forgejo page does not loadConfirm the SSH session running kubectl port-forward is still open, and kubectl -n <ns> get pods is healthy on the management container.
*.kubernetes.local does not resolveThe /etc/hosts entry is missing on the client; re-check with getent hosts / ping.
502/503 from TraefikCheck the pods and Ingress in the matching namespace (nginx-demo, nextcloud, euro-office, monitoring).
Grafana shows a login page but dashboards are emptyCheck the Argo CD monitoring Application, the Prometheus targets, and the Prometheus datasource in Grafana; see monitoring.md.
Nextcloud returns 400 with “Trusted domain error”Add the hostname to trusted_domains; see installation-steps.md section 7.9.
Connector: “Error while downloading the document file”Check StorageUrl uses port 8080, ALLOW_PRIVATE_IP_ADDRESS=true, and the internal Nextcloud hostname is trusted (section 7.9).
Example editor: “The document security token is not correctly formed”JWT_ENABLED=true or EXAMPLE_URL=http://example:3000 is missing; see installation-steps.md section 7.11. After the rollout, hard-refresh and create a new document.
Example: “File not found: new.docx” after a pod restartExample files are ephemeral; create a new Blank document. Avoid With sample content until the /create handler is patched; see section 7.12.
Connection refused on 192.168.0.192:80The socket proxy is not active: systemctl status traefik-ingress-http.socket on the management container.
Wrong management IPThe address is DHCP; refresh it from the Proxmox API or the Proxmox UI and update /etc/hosts (and ansible2/hosts.yml).

Security note: the only application ports permanently exposed to the home network are 80/443 (nginx demo, Nextcloud, Euro-Office and Grafana). Argo CD and Forgejo stay on ClusterIP and are only reachable while a kubectl port-forward started over SSH is running. SSH (22) remains open for administration.