N8N workflows simply won't load, they stop working CONSTANTLY

Questions
1

My n8n just simply stops EVERY DAY randomly working for MONTHS. Sometimes it works, sometimes it doesnt, and I can’t even load the workflow site. It saysn site not avalible “ERR_CONNECTION_TIMED_OUT” and thats it.
After sometimes 5 mins, 15 mins., completely random it comesback and I can access it and it works.
Im on Hostinger self hosted, what am I missing?

Im trying now for 15 mins. and my apps and forms using my n8n as a backend not working… somebody help please!

2

n8n Version

2.3.6

Source Code

https://github.com/n8n-io/n8n

License

Sustainable Use License + n8n Enterprise License

Third-Party Licenses

View all third-party licenses

Instance ID

c474895fe057acee9893b2dca7cd5c020a4eed53c9e8c1cc192e650fa1855d4d

Debug

Copy debug information

3

Self-Hosted Application – Instance Availability and Scaling

Context

For self-hosted applications, especially on platforms that support autoscaling or serverless-like deployment, it is common for the minimum instance count to be set to 0 to save compute and cost.

  • Minimum instance = 0:

    • When there are no active sessions or no API requests hitting the application, the instance may sleep.

    • This reduces operational cost but increases the cold-start time when a new request comes in.

Recommended Setup for Always-On Availability

To ensure the application is always available and ready to serve requests immediately:

  • Set the minimum instance count to 1.

    • This ensures there is at least one active instance running at all times.

    • Reduces cold-start delays and guarantees availability for users or API calls.

Additional Notes

  • You may still implement autoscaling for high traffic, but the base instance must remain at 1 to guarantee uptime.

  • Cost will be slightly higher than a zero-minimum instance setup, but user experience and reliability are significantly improved.

1 Like
4

Thank you!

As a not programmer myself, how can I find this option and where exactly?

5

This looks less like “workflows won’t load” and more like intermittent service/network downtime (ERR_CONNECTION_TIMED_OUT usually means n8n or the reverse proxy/VPS is unreachable). Quick questions:

  • is “Hostinger self-hosted” a VPS vs shared/PaaS?
  • how is n8n running (Docker/Compose vs systemd/Cloudron)?
  • what’s in front (Nginx/Traefik/Cloudflare/WAF) and which DB (SQLite vs Postgres/MySQL)?
  • During the outage, can you still SSH and do webhooks fail too, or only the UI?

If it’s a VPS, please add host monitoring (Zabbix/Munin/Netdata) to correlate outages with RAM+swap/OOM, disk/inodes, iowait, CPU steal, network drops, and container/service restarts.

Also monitor n8n health endpoints (e.g., /healthz and /healthz/readiness once enabled) so we can see whether it’s reachability vs DB readiness. Share VPS specs + install method + proxy snippet + logs around the last downtime and we can narrow the root cause fast.

1 Like
6

Well, thats a lot to comprehend rn, as i said, Im just a regular citizen who made an awesome numerology reading automatization, and pissed because its always “unavalible” after a while :smiley: What you wrote needs time to settle now haha :smile: