Skip to content

armgit / Pulse

Public Imported from GitHub
0

Real-time monitoring dashboard for Proxmox VE, PBS, Docker, Kubernetes, TrueNAS and vSphere. Self-hosted, with alerts for failed backups, full disks, unhealthy containers and offline nodes.

7.4 KB
Raw Blame

πŸ”„ Reverse Proxy Setup

Pulse uses WebSockets for real-time updates. Your proxy MUST support WebSockets.

Before configuring the proxy

Use HTTPS at the public proxy and keep Pulse's backend private. The examples below are configuration fragments, not complete TLS or authentication setups. Keep Pulse authentication enabled; TLS termination does not authenticate users. For identity supplied by an IdP/proxy, follow the separate proxy authentication guide, including its Header Trust Boundary.

A same-origin reverse proxy needs no CORS exception. Keep the UI and API on the same public origin rather than adding * to repair a login or routing error. For a separate trusted browser app, follow the exact-origin CORS checks.

Forwarded HTTPS headers are trusted only from a configured immediate peer. Set PULSE_TRUSTED_PROXY_CIDRS in Pulse's deployment configuration to the proxy address as seen by Pulse, not the browser's address or the public hostname. For a proxy on the same host connecting to 127.0.0.1:7655, use:

PULSE_TRUSTED_PROXY_CIDRS=127.0.0.1/32

Use ::1/128 instead if that connection actually uses IPv6 loopback. A Docker proxy normally reaches Pulse from its private container/network address, not loopback; trust only that proxy's fixed address or a narrowly isolated proxy network. Do not trust a whole LAN or all Docker networks to make a warning go away. Wildcard ranges 0.0.0.0/0 and ::/0 are rejected at startup.

At the public edge, replace client-supplied forwarded scheme, host and client-IP headers with values established by the proxy; do not append an untrusted value. With multiple proxy hops, configure the upstream trust boundary on each hop and trust only the final peer in Pulse. PULSE_TRUSTED_NETWORKS is not a substitute for proxy trust, and proxy trust is not an authentication bypass. See configuration overrides.

⚑ Quick Configs

Nginx

Inside your HTTPS server block, for a single public-edge proxy:

location / {
    proxy_pass http://127.0.0.1:7655;
    proxy_http_version 1.1;
    proxy_set_header Upgrade $http_upgrade;
    proxy_set_header Connection "upgrade";
    proxy_set_header Host $host;
    proxy_set_header X-Forwarded-Host $host;
    proxy_set_header X-Forwarded-Port $server_port;
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header X-Forwarded-For $remote_addr;
    proxy_set_header X-Forwarded-Proto $scheme;
    proxy_set_header X-Forwarded-Scheme "";
    proxy_set_header Forwarded "";

    # Critical for WebSockets
    proxy_read_timeout 86400; # 24h
}

Caddy

Caddy handles the WebSocket upgrade and sets forwarded client-IP, host and scheme headers automatically. The removals below discard other client-supplied forwarding metadata. Pulse still needs the matching trusted-peer setting above:

pulse.example.com {
    reverse_proxy 127.0.0.1:7655 {
        header_up -Forwarded
        header_up -X-Forwarded-Scheme
        header_up -X-Forwarded-Port
    }
}

Traefik (Docker Compose)

These labels assume an existing HTTPS entrypoint named websecure and a configured certificate. Keep Pulse on the private proxy network; do not publish its backend port publicly. Apply trusted-peer configuration to Pulse separately. Keep Traefik's forwardedHeaders.insecure disabled; upstream forwarded headers need their own narrow trusted-IP configuration, not blanket client trust.

labels:
  - "traefik.enable=true"
  - "traefik.http.routers.pulse.rule=Host(`pulse.example.com`)"
  - "traefik.http.routers.pulse.entrypoints=websecure"
  - "traefik.http.routers.pulse.tls=true"
  - "traefik.http.routers.pulse.service=pulse"
  - "traefik.http.services.pulse.loadbalancer.server.port=7655"

Apache

Inside a TLS-enabled virtual host, with the required proxy, rewrite and headers modules enabled:

ProxyPreserveHost On
ProxyAddHeaders On
RequestHeader unset X-Forwarded-For
RequestHeader unset X-Forwarded-Host
RequestHeader unset X-Forwarded-Port
RequestHeader unset X-Real-IP
RequestHeader set X-Forwarded-Proto "https"
RequestHeader unset X-Forwarded-Scheme
RequestHeader unset Forwarded
RewriteEngine On
RewriteCond %{HTTP:Upgrade} websocket [NC]
RewriteCond %{HTTP:Connection} upgrade [NC]
RewriteRule ^/?(.*) "ws://127.0.0.1:7655/$1" [P,L]

ProxyPass / http://127.0.0.1:7655/
ProxyPassReverse / http://127.0.0.1:7655/

⚠️ Common Issues

"HTTPS: HTTP only" in Security Posture

If your reverse proxy terminates SSL but Pulse shows "HTTPS: HTTP only" in Settings β†’ Security:

Pulse's secure-cookie and request-origin logic recognises direct TLS, or X-Forwarded-Proto: https from an immediate peer in PULSE_TRUSTED_PROXY_CIDRS; a header from any other peer is ignored for those decisions, even when the browser used HTTPS. Adding the header alone is not sufficient. The Security Posture HTTPS indicator is different: it is true when the request has direct TLS or any X-Forwarded-Proto: https header, whatever the peer. A passing indicator therefore does not prove that proxy trust is configured correctly, and "HTTP only" means neither TLS nor that header reached Pulse.

Check the proxy's actual backend connection address and the effective Pulse setting first, then check that the TLS proxy sends exactly https. For Nginx, keep proxy_set_header X-Forwarded-Proto $scheme inside the HTTPS server's location; do not label a plain-HTTP request as HTTPS. Caddy normally needs no header override. Any Caddy header_up override belongs inside its reverse_proxy block, not directly in the site block.

After applying configuration through your normal deployment path, sign in through the real HTTPS URL and recheck Security Posture and live dashboard updates. If using SSO, check its redirect/callback on that same URL. A successful /api/health request alone does not verify authentication, WebSockets or proxy trust. Do not disable authentication, certificate verification or origin checks, or expose the backend, to test this. Share only redacted configuration and HTTP/error status, not cookies, proxy secrets or a full network export.

Other Issues

  • "Connection lost" (Workloads), "Disconnected", "Reconnecting...", "Sync reconnecting", "Backend healthy" or a stuck "Connecting to Pulse" screen: separate the WebSocket upgrade from state delivery. If an already-observed upgrade failed through this proxy, check its Upgrade and Connection forwarding. An HTTP 101 or welcome message without subsequent state is a different failure; changing those headers is not an established remedy. Use the stalled-dashboard guidance, even if /api/health returns 200. Keep authentication, TLS and origin checks enabled; do not expose the backend to test a bypass.
  • 502 Bad Gateway: Check Pulse's actual listener and the proxy's upstream address, port and network reachability; this does not by itself prove Pulse is stopped. Container localhost refers to that container, not another service.
  • CORS Errors: Do not add CORS headers in the proxy; Pulse handles them. Set Settings β†’ System β†’ Network β†’ Allowed Origins or use ALLOWED_ORIGINS if needed.
  • OIDC redirects fail: Ensure X-Forwarded-Proto is set (see above).
  • Wrong client IPs: Check the immediate peer and header replacement above. Do not broaden proxy trust or change authentication to repair attribution.