thank you!

Can you talk a bit more about what you're running into?

sql: error: connection to server at "ballast.proxy.rlwy.net" (66.33.22.248), port 46255 failed: server closed the connection unexpectedly

This probably means the server terminated abnormally

before or while processing the request.

railway Cli also unable to connect

railway connect "Store Management DB"

psql: error: connection to server at "ballast.proxy.rlwy.net" (66.33.22.248), port 46255 failed: server closed the connection unexpectedly

And was this working before? Did you change anything recently?

Nothing changed

And i'm able to connect through the railway web portal which is so odd?

attempting redeploy

Sounds good let me know! Yeah a bit odd but looking into it on our side

Update - railway UI unable to connect to DB now

keep disconnecting every couple seconds as well.

this fucking sucks

this is probably

like not even exaggerating

like the 4th downtime

in like 3 weeks

or 2 weeks

AI code

nah this is fucked up

yeah its bad

AI doesnt write that bad of a code lol

opus 4.6 would never lol

@lunix I feel your pain man

truly I do

worst thing about it is sitting and watching it happen

not being able to do anything

exactly

genuinely thinking of just

dipping

are you a small team

and hosting using dokploy or coolify or smth

i'm on a 2 person team so I can't waste time on infra

im Me lol.

just me, building a product

self hosting has issues too though no?

wouldn't be that bad

imo

I mean fair

atp its either self hosting or fly.io

i'm just thinking rihgt now about even just the databases

running those on something more stable

and connecting via public urls

while we're at it - give me ur honest, brutal opinion on https://wegotwork.co/

UI is clean

needs some color

yah idk what colors to add and where

Features

Pricing

About these would look better centered imo

sorry bro I gotta get back to my dead database

good eye

yea nws

nice chatting though and I truly feel your pain

likewise. we'll probably go through this again in a day or two - see u then!

See you then brother!

And dm me anytime for feedback on your site

added u as a friend. whenever you're done with the DB hassle - we could talk again and discuss some stuff, if you're down ofc.

sounds good

we may have one thing in common : railway downtime

Yeah we can chat during railway downtime while my management emails me and customers can't do basic stuff

lmao

what are u working on, if u dont mind me asking, and if its not confidential

same

no

might have to do the same if this keeps happening

Same - my site is down

I mean who cares what its related to at this point

apps are down its that simple

and it needs to be fixed

We are working on it right now!

Now is a good time to add some nice messaging to my app so my users don't freak out and think all their data is gone 😅

working!

appreciate you

My site is able to connect to my postgres again too

That was the worst downtime I've had with railway ever.

Appreciate the 'we're working on it', but honestly, that's not good enough anymore.

This is at least the 3rd or 4th database outage in the past few weeks. Not a one-off. Not bad luck. A pattern. And every single time, it's paying PRO customers sitting there watching their apps burn while their users churn and their management breathes down their necks.

So I want a real answer: what is Railway actually doing to make sure this doesn't keep happening? Not 'we're looking into it', a concrete plan. Are there architectural changes being made to the proxy layer? Is there a post-mortem being published? Is there any kind of SLA or compensation being offered to customers whose production apps have been down multiple times this month?

Because right now, the community consensus in this very thread is that Railway is only suitable for hobby projects. People are openly discussing moving to Fly.io, self-hosting with Dokploy, or just pulling their databases out entirely. That should alarm you.

What's the actual commitment here to retain customers who trusted Railway with production workloads?

heartattack

most important message of the thread. More important than my initial flagging of the issue. thank you

The thread is genuinely rough reading. Multiple PRO customers, all hitting the same connection drops simultaneously, redeployments not fixing it, people literally making friends in the thread because they're bonding over shared Railway trauma.

So true.

lol

lol

The green shield of assurance

💪

Nathan, did you read this

Ah yeah you did

Yeah - I haven't experienced a full outage like this since using Railway (about a year). But not good that others are seeing this happen a lot.

Consider yourself lucky my man!

Its been weekly for us for the past 6-8 weeks

Yikes

We need duplicity - but not sure Railway has anything like that build in...

Are you using internal networking on your DB

You found anything?

It seemed unimpacted during this

Internal networking mostly

Damn, my internal networking Db connections seemed fine but public was cooked due to TCP

Haven't heard of it. Does it let you swap between hosting providers or something?

No, meaning we need something that will do that.

right right

just sayin

AWS is just too damn expensive and maintence is too high.

Yes, I love the simplicity of Railway

Have you used it much?

Yeah for a few things, not much though, a bit more cumbersome but not by a lot.

In the end the most important thing is uptime

This seems nice:

That would be what we need - thanks @Tomgang

all that matters

Yeah will look into it too. thank you Tom!

Not just downtime it's also what it does to us in front of our customers.

They don't know what Railway is. They don't care about TCP proxy failures or whatever is happening under the hood. All they see is that THEY CAN'T ACCESS THEIR DATA. And who do they blame? Us. Not Railway. Us.

We look like amateurs. We look like we don't know what we're doing. We're out here building serious products and we're being made to look like we deployed something on a raspberry pi in someone's basement.

Railway gets to quietly fix it and move on. We're the ones doing damage control with our customers, explaining outages we had zero control over, and trying to hold onto trust we worked hard to build. That cost is real and Railway doesn't seem to acknowledge it exists.

This. This. This.

our users are not technical. the expectation shouldn't be for them to be technical. they shouldn't need to understand networking. TCP means nothing to them.

We can only say "There was a downstream outage from our hosting provider" so many times. They don't care. It means nothing

@Tomgang can I ask what your current gameplan is(

How do costs compare?

Although to be fair I don’t care about costs at this point I care about uptime.

Boot licking incoming:

I agree with a lot of this sentiment (and I have also mentioned that some acknowledgement would be appreciated: https://discord.com/channels/713503345364697088/1479490167277490268/1479504082157961257), but there is also the shared responsibility model in the cloud. Connecting to your DB from your application via the proxy is really bad practice (egress fees, security, an additional failure point).

Of course it shouldn't happen, and the downtime recently has been problematic, but I think this thread is somewhat poor taste. It is possible to be constructive with your feedback, even if it is letting it be known that you're migrating off of Railway

@Jasper I completely agree, some of our services are using the internal url, but not all services are hosted on railway and have the VPC access

@McKay Current gameplan is to migrate production data to Render. It'll require a planned maintenance window but at least that downtime is on my terms and my schedule, not Railway's.

I'll stay on Railway with my staging env for now but only until I hear something concrete from them about how they're actually going to fix this. Not 'we're working on it', not 'very true', a real plan. Post-mortem, architectural changes, something.

Because right now the trust is gone. Every deployment feels like a gamble. Every morning I open my laptop and wonder if today is the day something just disappears. That's not a headspace you can build a product in.

Very well said!!

My internal connection to the database was what went down this time

Also well said. We are moving to production tomorrow and will be rolled out to hundreds of users with thousands of downstream customers per month. I really am rooting for Railway and there’s nothing I want less than to have to go provision all infrastructure on another platform. So I’ve let it string along for the past couple of months, but unfortunately, my trust is also lost.

All I’d ask for is just actual actionable items that will be taken for these incidents to see how things are being resolved because unfortunately that isn’t happening. that is where my trust falls.

@Jasper And on the 'poor taste' thing, I get it, but I'd push back a little. When this is the 4th outage in a few weeks what does constructive feedback even look like at that point? I'm not attacking anyone personally, I'm not spreading misinformation. I'm describing what is happening to me and others in this very thread.

The feedback has been constructive. I asked specific questions. I laid out exactly what I need to hear from Railway to keep my trust.

Very true. We’ve all been really professional throughout these Incidents. It just feels like things are at a boiling point now where people feel as if there’s no other option @Jasper

For my team, this all stems from a lack of communication. Flagging an incident means nothing. I need to see what and why it happened, otherwise I have no faith in it not happening again.

We are asking the exact same thing of Railway that our customers are asking of us. How did this happen, and how do we make sure it doesn't happen again?

When our product goes down, we don't get to say 'we're working on it' and leave it at that. We owe our customers a real explanation and a plan. All we're asking is that Railway holds itself to the same standard.

We had downtime also because of this but it was a Redis service that was unreachable

Had to re-deploy

How close are we to the end of this drama?

It's been more than 1h now... Any news?

@CallMeFred this was from the incident earlier

but was resolved by railway team

if you're still having issues i'd open a new thread

deployments still queued

it has been 1h

Yeah thats unrelated to this incident

open a new ticket in #✋｜help

Insane.

If only someone could reply in the other thread 🙂

I get it man

I dont work for Railway i'm sorry

1h+... once a week at least

I know dude

its very frustrating

and support goes quiet

Customers would deserve more transparency

@McKay just curious, did you ever get this resolved?

Looks like Railway responded: https://railway.com/changelog/2026-03-13-highly-available-postgres

@McKay @Tomgang

This is great news! Thank you

Will wait until its prod ready to upgrade of course.