Severe latency spikes and ECONNRESET errors on Directus service
build76
PROOP

a month ago

Can't reach station.railway.com to submit a status ticket either

Project Region: europe-west4

Issue Started: ~11:15 AM GMT, January 14, 2026

Symptoms:

  • Response times spiking to 20-55 seconds (p90/p99)

  • Error rate spiking to 80-100% during incidents

  • ECONNRESET errors from client applications

  • Affects ALL routes, not just specific endpoints

What we've ruled out:

  • Not application code - rolled back recent changes, no improvement

  • Not database load - Postgres queries return instantly (<50ms), only 1 active connection

  • Not resource exhaustion - CPU at 0.3 vCPU (limit 32), Memory at 250MB (limit 32GB)

  • Not extension issues - removed custom extension, still slow

Diagnostics performed:

  • Database responds in <50ms via direct psql connection

  • Health endpoint takes 2-5 seconds (should be <100ms)

  • pg:responseTime warnings at 278-316ms (threshold 150ms)

  • Service restart did not resolve the issue

Impact: Admin UI unusable, customer portal experiencing connection resets

7 Replies

brody
EMPLOYEE

a month ago

Can you go more in depth when you say that you cannot reach the station?


There was an incident earlier (~16h ago) where no services were able to be reached (incl Railway's sites), would time out/econnreset, or get a cloudflare ssl error
might be related to that
https://discord.com/channels/713503345364697088/727689277219012669/1460747113549009141


brody
EMPLOYEE

a month ago

That's known, but this report doesn't seem to line up with anything.


brody
EMPLOYEE

a month ago

Unless I am just bad at reconciling time zones


Is this still ongoing?


Are there connection issues when connecting to a service from a service via the private network?


Or are you experiencing these symptons when connecting to the service over HTTP from e.g. your home network?


Loading...