Hey there! We've found the following might help you get unblocked faster:

• 🧵 n8n - Connection lost

• 🧵 connection lost on my n8n service and now I can't log in

• 🧵 n8n Service - Outbound Egress Traffic Blocked (Timeout)

• 🧵 Connection Delay Issue Between Web Server and MySQL Server

If you find the answer from one of these, please let us know by solving the thread!

After a manual redeploy, my public URLs were not set explicitly. n8n was running behind Railway’s proxy (public HTTPS on 443, app listening on 5678), but the editor/push service was computing the wrong expected origin. That made the push channel (WS/SSE) fail and the UI showed Connection lost. This is a known behavior with recent n8n versions that enforce strict origin checks behind proxies/CDNs.

What fixed it: set the actual public URL and proxy hints explicitly (no placeholders), then redeploy Primary and Worker:

N8N_EDITOR_BASE_URL=https://.up.railway.app
WEBHOOK_URL=https://.up.railway.app
N8N_HOST=.up.railway.app
N8N_PROTOCOL=https
N8N_PROXY_HOPS=1
PORT=5678 (don’t force N8N_PORT=443)

After this, opening workflows worked normally (no more Connection lost). The reasoning matches the docs for reverse proxy webhook/editor URLs and proxy hops.

Optional stability tweaks: if your environment filters WebSockets, set the push backend explicitly and redeploy: N8N_PUSH_BACKEND=sse (or websocket if SSE is the problem). Recent versions default to WS; choosing SSE/WS manually often resolves editor connectivity in proxied setups. Keep N8N_SECURE_COOKIE=true on HTTPS; if debugging odd session behavior through a proxy chain, you can temporarily toggle it to isolate cookie issues, then turn it back on.

How I got back in (if you’re locked out): first make a quick backup from the CLI:

n8n export:workflow --all --output=/tmp/workflows.json
n8n export:credentials --all --output=/tmp/credentials.json

Then reset User Management (this does not delete workflows; it only removes user accounts so you can run the setup again):

n8n user-management:reset

Redeploy, create a new owner account, and all workflows remain intact.

Why this happens behind proxies: n8n builds public URLs from N8N_PROTOCOL/HOST/PORT, which don’t match reality when the app listens on 5678 internally but is exposed as HTTPS 443 via a proxy. The fix is to manually set the public URLs (WEBHOOK_URL, N8N_EDITOR_BASE_URL) and declare proxy hops so the editor and push endpoints validate the correct Origin.

----

if this answer was helpful, please mark it as the solution, thanks