same across, seems we have issues in general yet to be flagged

same

same

facing issue here

Seeing connection issues to Railway-hosted Redis here too since around 0945 GMT

Same with redis for me

same

redis

@railway Can you please call an incident?! Why does it always take you such a long time??

What region are such services hosted in?

EU West

Have you tried redeploying the affected services?

No, will do and report back

A lil update, the team is currently investigating the elevated rate of TCP timeouts in EU West.

Same here! It seems every week we have something as a present from Railway!! That's great!

Possibly the same incident (or related): from US-East, TCP proxy connections via tramway.proxy.rlwy.net are reset on idle within 10–45s. Started ~09:45 UTC today.

TCP-level repro:

# DIES in 10-45s:
ssh -o ServerAliveInterval=5 user@tramway.proxy.rlwy.net "sleep 45"
# → client_loop: send disconnect: Connection reset

# LIVES indefinitely (continuous stdout):
ssh ... "for i in $(seq 30); do date; sleep 2; done"

Eliminated locally:

• Container health (no OOM, CPU not throttled, sshd auth.log clean, tmux survives → not restarting)
• Container-specific bug (reproduces on two services with different ports/images: 35946 and 41991)
• Client network (reproduces from mobile-data hotspot AND another physical device — different ISPs)

So this is upstream of the container. Variable timing (10-45s, not fixed) suggests probabilistic middlebox eviction. Redeploy won't help — already verified container is healthy.

Hi @0x5b62656e5d — this has happened a few times now where I noticed and reported incident-like behavior hours before it was acknowledged on your side.

Is there a better escalation path or internal-facing channel where I can send early signals when I see this happening? I’m happy to help provide timely heads-ups if that’s useful.

A re-deployment did not help by the way

An incident has been called: https://status.railway.com/incident/V08OD1KI

Hi,

Sorry about this issue, incident is updated with details, it should be mitigated now.

I wanted to confirm, are folks are seeing recovery?

It looks like it's sorted for now, yes.