I would love to see the following accomplished, or even fixed up a bit...

- Docker-Compose mimetype detection on the template composer is broken... most times, it will just say, "Error Only YAML files are supported."
- Shared Volumes (so that we can finalize and host Supabase on Railway)
- Multiple Volume Support (for when a service needs multiple volumes [very big issue for some containers, such as GitLab])

Great goal! Hope you're able to squash all those bugs!

I've run into two bugs the last few months

1. I can't seem to have a period (.) in shared env variables (works fine in non-shared variables): https://help.railway.app/questions/shared-variables-show-up-as-empty-strin-ccbda77c

2. Watch paths disappearing (I guess Railway couldn't reproduce, but maybe someone wants to take another stab at it): https://help.railway.app/questions/watch-paths-are-removed-automatically-7f91872b

Those are two (somewhat) minor things I've noticed. Otherwise love the platform and excited to see it grow!

100% with Vin regarding volumes, would love to see shared + multiple volumes becoming a thing!

Also would be nice to see monthly earnings on the earnings tab and on individual templates rather than just total earnings

Love many aspects of environments, including PR environments; but I don't love that the active environment is not part of the URL. I often want to compare settings between two environments and I make two tabs, then I switch environment, and go back to the to the other tab and forget that it's a global browser state setting, then get confused that the settings look the same, then... doh! I just can't compare environments like that, unless I use two different browsers

While evaluating a new platform to move my former Google App Engine app to, I almost gave up on Railway because after something like 3 attempts to release my repo in the free trial mode, it never got past the 'Build > Building the image' step. It failed each time after a 15m timeout. I found the production status page, and saw that at one point free-trial builds were paused, but the incident had been closed out days before my attempt.

Seeing that free-trial builds were paused at one point, I figured it was worth $5 to see if a hobby account worked. It immediately did work and I LOVE Railway now. But you almost lost me because of the QoS stuff that was causing free trial accounts to not work.

Now, as a paid user, I understand QoS and of course I want free trials to perform poorly before my paid account. But there was no indication that was what was happening. All I saw was poor service.

So I'd suggest that if the free trial mode has to be QoS-ed significantly, it would be better to notify free trial users that things are paused and try back later, than to make it look like the product performs poorly.

anukari, fwiw even Pro users have experienced the same issues.

When an app goes to sleep the memory metrics graph gets frozen in place and continually shows the memory that was used when the app went to sleep, this is incredibly misleading, ideally when an app is sleeping the metrics would simply report zero at the same frequency as if it was alive.

• +1 for shared volumes to support hosting Supabase (final blocker for moving my team over to Railway)

• +1 for storing the active environment in the URL so that I can compare environments between tabs

• A light mode UI. I find dark text on a light background much easier to read.

• Ability to expose multiple ports per service

• Ability to integrate with multiple Vercel projects - not just one

• The UI for editing an existing service environment variable is currently broken on mobile

• Ability to synchronise deployment of services triggered by the same commit. Each service would build in the queue as they do now, but would only be promoted to live once all services are ready.

lochlan,

Railway does have a light mode ctrl / cmd + k

You can currently expose multiple ports (https) per service with the edge proxy enabled.

That's a great initiative!

• The metrics tab of a service has barely any space dedicated to the tab content while on mobile + landscape mode, it's very hard to view the graphs

• Frequently there are missing logs
  

• I'm not sure if this is an issue or a feature request but I'd like to be able to grant the ability to view logs to project members without giving them edit capabilities

• Setting a default region, so we don't have to migrate volumes when deploying templates

• I love the canvas to visualise our services, would be cool to add virtual services so we can add external things like S3 buckets, or managed DBs, maybe with hooks to create db branches or buckets for PR envs

I would love to be able to manually draw arrows on my dashboard view between services.

I would love to be able to have a more straight-forward way to deploy static frontend sites such as React SPAs and Storybook projects. The current process with nixpacks files is confusing. Other platforms like Render and Vercel make this super easy and put those static assets behind a CDN.

My app lost connectivity to a Railway database because this:

postgresql://user:password@containers-us-west-58.railway.app:port/railway

became this:

postgresql://user:password@viaduct.proxy.rlwy.net:port/railway

Nothing informed me of the change besides loss of connection. Would've been nice to know!

I'd love an extra column here saying which apps/services have access to a Shared Variable. Currently, you have to open the "Share" modal for each of them individually to check or hover over the [x] badge one by one.

iannotian, that change was communicated a very long time ago over several channels over a large time span.

iannotian, that change was communicated a very long time ago over several channels over a large time span.

What was the email subject?

I'm sorry, It was quite a long time ago, I wouldn't remember that.

It was communicated over twitter, changelogs, email, community posts, and even shown on your dashboard, sorry you missed these.

Reliability: It seems like builds will randomly fail due to internal railway stuff, and I have to go in and redeploy it to make it work. Also seems to occasionally hang and deployment (mine is currently at 20 mins vs the usual 10 mins)

PLEASE multiple volumes

PLEASE multiple volumes

Do you mean multiple volumes mounted to a single service?

A couple of security hardening points that I'd love to see.

1) Audit log for railway run commands
2) A way to lock down railway run against production (or other envs) for some/all users, with an override button for emergencies.

1) Audit log for railway run commands

Could you share what this would ideally look like for you?

1) Audit log for railway run commands

Could you share what this would ideally look like for you?

@brody I'd like to know what commands people ran so that we have a history of the changes against the different environments.
I guess it could flow into the regular log queue as long as it was separately searchable.
I know the output is generated on machines far away, so capturing (and filtering sensitive data) would be hard.

Maybe it could flow into the Activity panel in a project?

Anywhere that's searchable would work for me. I'd like to know who ran what when.

Love your product, keep it going!

> A simpler way to deploy React (or other) SPA (like Render)

Re raising that whatever logic you have on the reference variable selector, it should also include full matches.

Full context https://discord.com/channels/713503345364697088/1251243221305266276 (discord link because it turns out that discord feedback doesn't cross-post)

I would also like to see general log improvements. The current system is pretty good. I have two main requests

Can the way default severity be reevaluated? Right now, every single psql request is sent to the logs as an error. I get that it has to do with the channel it comes into. But maybe railway has a custom filter for the common DBs, and auto moves it to info / debug. I find the logs hard to go through errors due to being clogged with db stuff. Well, it isn't that bad. But you did ask for nits ;)

Is there was a way to have fuzzy search on the whole data? IE, if I have JSON data, and want to search for some nested key, it can be a bit of a pain. Being able to do a fuzzy string search over the whole thing would help. Assuming it is even possible based on how you store / process the logs.

Anywhere that's searchable would work for me. I'd like to know who ran what when.

When undergoing our compliance work, we did spend some significant investments into audibility of the platform. I may or may not have spiked a user facing audit log. It requires significant polish and likely won't come this quarter but consider it on our roadmap.

Here are two small ones that drive me bananas when they happen on the wrong occasion:

1. Sometimes the redeploy actions are not present in a previous skipped or removed deployment: https://share.cleanshot.com/hShW2V2D

2. When using watch paths, the containers start to build and then stop. I would like for them not to start until the watch paths have been validated because I fear the detection failed. Especially considering this: https://help.railway.app/questions/watch-paths-are-removed-automatically-7f91872b

For a client side developer who's mainly used to writing frontend apps and building apis on serverless, the knowledge here is infinite. I would love to see more tutorials on youtube on basic things like setting up dbs, reading/writing from dbs using state of the art tech stacks. the docs / videos need a lot of work.

That seems a bit out of the scope of Railways own documentation, and not all that specific to the Railway platform.

Interesting POV. the vars, shared vars and how to run them locally are pretty railway specific.

How to use variables of all kinds is covered in the docs, and the functions of the railway cli are documented too.

Please share what Railway centric information the docs are missing.

Railway interface is very confusing. As I just joined finding many things is so hard. First of all I couldnt find how to delete projects. I found questions in the help station regarding how to delete projects but most of them were answered by how to delete services. Second I dont know how to get the IP for my server. Third, I had to scroll down to the very bottom of all these comments to find a way to reply for this. Maybe its just me, but a good interface shouldnt be so hard to work with. Because it makes railway incredibly hard to use in my case.

Being able to delete projects inside a workspace is essential in my opinion. Currently that is not possible or I didn't find the way with the new interface.

@chap @javikeziah Inside of any project you can go to `Settings › Danger › Delete project`.

Editing shared variables just stopped working completely for me - every time I make a change, it just immediately adds it back as the original version, I'm unable to swap out my (soon to be expiring) api token for my production service

Ok nevermind - turns out that after I deploy with the changes, the values in the shared variables section update to the new value. The UI for editing shared variables is a bit confusing - I would delete a variable from the shared variables, and it would get deleted from the page, only to show up half a second later, so it looked like there was an error with Railway that was preventing it from being deleted. And when I edit a variable, I would expect that when I inspect that variable again after editing (but before deploying), that I would see my newly updated value. But it's the old value (until you hit "Deploy") which made me think that Railway was having issues updating my shared variables.

Would love to have a way to update a project template automatically when updates are made to a project. This way we don't have to create a template everytime a change is made to a project. Something else that would be nice is creating/deleting templates via the API.

This way we don't have to create a template everytime a change is made to a project.

Curious to know why you are creating a template whenever you change your project in the first place.

Something else that would be nice is creating/deleting templates via the API.

If the dashboard can do it, so can you! simply watch the network requests in your dev tools.

1. Multiple global deployments not available to pro users.
2. Env vars overwrite other env vars when merging from different branches
3. Cost, no usage included in the pro plan, railway is expensive.
4. No neon integration with preview branch sync with db url branch sync. (you could win me over with this alone).

1. Heard, we have a response somewhere in the forums on why this isn't the case for now. (Compute liquidity )

2. Oh... can you expand on the behavior that you expect within the product to be to handle those conflicts? Something like a diff per PR env?

3. We hear ya, a lot of internal discussions around our offering. Nothing solid yet but we are floating a few ideas around.

4. We'll have our people talk to their people, can you tell me a bit more on how'd you expect this to work?

Please add support for Kubernetes liveness checks, so we can automatically restart containers that fail health checks.

More insight into different metrics, like network latency etc

These will be great:
- Dashboard page: overviews of services state on project cards, i.e. show if any services in a project have failed builds or are all active. When you have lots of projects it is great to be able to see that all are running well from the same page.
- Prod ready cron jobs. At least guarantee that runs won't be skipped during build time. Will happily migrate all my crons to the railway.
- Noticed the problem: Services don't start automatically after they were stopped due to billing issues. Need to manually rebuild each service. Had it with Node services.
- Sometimes when you are on the project page and refresh it, you see the project page as normal, but you are actually logged out(not 100% sure, but it might happen after I clean cache on my Mac). So it looks like normal, but functions don't work. You need to click on the Dashboard link, to trigger the auth check and get redirected to log in page.
- In the build history for the service it will be great to highlight(or keep on top) the latest successful build(maybe only when current one is failed). When working with monorepos you get a lot of "Skipped" items, so finding the latest successful build to rollback to can be hard.

• Last week railway deployed our "release" branch but a migration from our "main" branch ran. How is this possible and this is very disappointing! :/
  
  Project id: 3546eaff-2bac-4b25-92df-7b14c7279039
  Env: production

It would be so useful if the graphs time axis 1) agreed with each other on the dashboard screen so I can see like-for-like and 2) stuck to the time filters. For example, if the service only has data from 3-4pm but I asked for the last 12 hours, the time axis behaves really unpredictably and I almost always misread the graph the first time I look at it.

1. Multiple global deployments not available to pro users.
2. Env vars overwrite other env vars when merging from different branches
3. Cost, no usage included in the pro plan, railway is expensive.
4. No neon integration with preview branch sync with db url branch sync. (you could win me over with this alone).

1+ 🫡 to remove usage usage, it costs a lot if you have many active services or many volumes

1. Adding the ability to recover a working deployment, for example if the new deployments is removed and you want to recover it, would save the mess made by the person

I would love to have a datatable / list view of all my Railway projects. The unsorted tile view just isn't cutting it if you have more than 20 projects

I would love to be able to change the order of projects on the dashboard, or at least be able to pin certain projects to the top. Thanks for all your hard work!

This way we don't have to create a template everytime a change is made to a project.

Curious to know why you are creating a template whenever you change your project in the first place.

This so we can keep the template up-to-date with the project configuration.

I am missing onedev.io

It would be great if we could specify which project to spawn a template into when deploying a template via:

• The templates page

• The marketplace

• The 'deploy on Railway' purple buttons we can share on ie repository README's

One serving of UX frustrations coming right up. Take them in the good faith in which they are meant.

1) Do not open the assets page when I right click the Railway logo in the top left. I want to "open link in new tab", goddammit. I don't want to copy your icon. Please stop trying to be cleverer than the user.

~2) On that note: DO NOT SYNCHRONIZE THE SELECTED ENVIRONMENT ACROSS BROWSER TABS! I want to have a side-by-side visual comparison of my environments in two different browser tabs. I cannot describe the amount of frustration and anger and hairpulling this causes. Please stop trying to be cleverer than the user.~

3) Use more anchor tags in your website output. I want to be able to right click and "open link in new tab"; or just want to see the href attr. location preview that browsers display on hover - it's an indication that "clicking this takes me to a new location". All these buttons are annoying.
4) Integrate GitLab so that I don't have to mirror my repositories to GitHub

5) Allow disabling the autodeploy-on-commit feature without the need to completely detach the source repository.

~6) How about you give PRO the same $5 usage credit that HOBBY users get? Come on...~

7) I would like more detailed database stats. transactions, connections, timings... As a GCP reseller you could get inspired by the GCP dashboard for Cloud SQL.

8) Please implement standard password login. magic links are an annoying gimmick and shouldn't be used as a primary login method. You are the only service that does this. We have our password managers, guys. Let us use them.

~9) Why isn't EU available to the base plan? I'd even pay a few dollars as an addon for this option. Forcing us to pay the whole 20 USD + entire usage just for the ability to deploy to the EU region is annoying. I don't care about your 32 GB of memory, but I care about the deployment region. Make it customizable.~

10) Let us bill different projects to different credit cards! Please copy GCP and their "Billing Accounts". Don't force me to pay every project under my account in bulk with a single credit card. Let me charge specific projects to specific cards.

Edit 6. 5. 2025: strikethrough got implemented

1) Log exploration is quite painful. Use case: someone copies their current logs URL to an error, places into ticket. Later, following the link, user wants to explore backwards, but this behavior is fully broken: there is no ability to "Load more" past the starting time in the URL; the time selector is not editable. Best workaround: edit the ISO timestamp in the URL back a few minutes to access older logs.

2) similar to a common github actions behavior, the ability to allow only 1 build running for a given environment+service would be very helpful (i.e. push commit, then quick followup push seconds or minutes later – the first commit builds should be canceled (if an option is toggled). Similar behavior would be very nice for production to prevent some network flakes that occur around container switch over

3) dockerfile layer caching is not supported. I have a mix of nixpacks and dockerfiles, and the dockerfiles, which frequently start with a bunch of apt-get install take 10-15 minutes (even though they should take ~30 seconds). PR environments don't support the watch directories, so for PRs to be green, the full docker build must complete without almost any real caching that image repositories all support.

4) there is a fair bit of nixpacks logspam (errors and warnings which users cannot control or interact with) that should be silenced or fixed

5) +1 on current environment being URL state so cross-tab/window comparisons are possible

6) CLI protection of production environment (y/N stdin response prior to executing) was a great CLI feature that got dropped. It should come back!

7) on PR envs, make going back to the PR easy (didn't there used to be a link?)

Ability to pick the region when deploying a new service/template. I now always have to migrate it to our main region (EU) after it's been deployed.

When I filter logs to a time range, it's usually because I am trying to narrow down to an error. It's very frustrating that it does not let me expand the search range by clicking a button at the top of the logs list... instead it just says something like "you've reached the start of the range"... Ok, that's great, can you let me click a button to expand the range back by another minute? Hehe

When Github is linked to Slack for notifications, if I set a railway service to Enable App Sleeping, I receive a deploy notification in Slack every time the railway service wakes up again.

I Love the canvas feature, but I would really love to be able to model external services and link to them. Like AWS S3 buckets or Cloudflare CDNs etc. Or even better have those services in Railway

I configured app sleep with health check, when the container is starting up i still get the placeholder page about app not ready.

I think it should wait for /health when we starting a container after sleep.

I Love the canvas feature, but I would really love to be able to model external services and link to them. Like AWS S3 buckets or Cloudflare CDNs etc. Or even better have those services in Railway

What a great feature idea – cross cloud "core" resource creation and cleanup (buckets, other resources) according to a named pattern.

• Support for common monorepos, e.g NX, all of our /apps are linked to a service in Railway one way or another. It would be fantastic to be able to use something like nx-affected to ensure we're not rebuilding our 40 services within a monorepo based on one push to change a log in a single app (granted we're building our own CD for this, but out of the box support would be awesome!)

• Support for grouping projects e.g by client, etc

• Advanced permissions, environment based, and being able to hide env variables

Speaking of nits: is there a way to unsubscribe to this thread? :D Lots of good stuff here but I don't need to keep getting emails.

Limit public URL access to certain IP addresses -> Scratch that, the tailscale subnets solve this.

Is there a publicly available curated list of accepted proposals from this thread? Which ones are accepted, and which ones are rejected?

I have to agree with others here, why do Hobby users get usage included but going Pro removes that. It feels completely backwards, even if its the same $5 though you could always double it for Pro

Ability to remove the "Webserver detected" -> I don't want to have a public URL to my db for instance.

...

There seems to be an issue when viewing usage per service for an environments where 1 or more service is sleeping. The app completely crashes and shows a "Application error: a client-side exception has occurred (see the browser console for more information)." Error.

Other than that, love the platform.

• A lot of errors that happen especially at runtime are very non-descriptive. It will say things like: "Container Died" or "Failed to start". I'm a developer, give me details!!

• The UI for attaching a Docker image is weird. I press the link to go to the image, and how do I submit? Enter? Okay. It's just not intuitive. Make it make more sense from the get-go please!

One small thing - deploying from template always requires me to press `New from Template` and then type (nothing appears) and then press Esc and start again (this time it works)

+1 on the docker image selection, gets me every time

The Waiting for CI feature is very useful, however we see often these are stuck even when the CI is finished succesfully. At least a certain timeout on this or proper interval checks from Railway should prevent that
.

There seems to be a bug when if my pipeline fails on the main branch the deployment is skipped, makes sense, but when I retry those jobs on the ci and they pass the deployment is not triggered.

Also, some way of filtering the types of webhooks that are sent to specific channels would be great. The hooks now are very noisey so they get ignored. It's be great if I could choose to send failure hooks only to a specific channel.

It should be possible to just upgrade your hobby plan to pro. Right now upgrading to pro creates a separate subscription. There is also no way to get a hold of anyone about anything

I deployed Postgres service. Then I changed ending of the DATABASE_URL env from railway to my database name. After, I changed my app to read DATABASE_URL to connect. I created a Dockerfile for my service and deployed it to the Railway. So, it is running in a docker container on the Railway.
The issue is next: my service can't connect to Postgres service via internal host. It says UnknownHostException. Is the problem that my service is running inside the docker network?

Now that we have GitLab Docker Registry, I'd love to see an integration for deploying from a GitLab repo, just like we can do with GitHub.

My auto sleep services don't sleep at all

Specifically my (postgres db + tailscale) service

It's probably because tailscale has some active connection with their servers, but using the exact same setup on fly.io it auto scales down to zero with no problems

I would love if I can edit the sleep trigger, maybe edit the minimum network traffic to consider a service active, or adjust the minimum time between two individual requests

Could we see ttyd or another terminal over web service as something I could open in a deployment? I'd love to be able to poke around in my container while still in the browser.

It would be nice if the deployment pane remembered whether the history was expanded or collapsed (it defaults to expanded every time)

Speaking of nits: is there a way to unsubscribe to this thread? :D Lots of good stuff here but I don't need to keep getting emails.

We now support thread subscribe / unsubscribe in Help Station (check out the button in the top right of a thread) !

1. (easy): Add some sort of debounce when clicking "Show value" on a variable. If we click again to hide it and click yet again to show the value, it hides after just a couple seconds (10 seconds after the first time we click show value)

2. (moderate): Add a CLI interface in the Data tab for PostgreSQL/MySQL/Redis resources. I want to query/modify data and it's always going to be more convenient than using the baked-in interface. Even Heroku doesn't have that functionality which is surprising. This would make me so happy!

3. (hard): Mentioned by others, but some sort of blob storage service wrapper around AWS S3/Cloudflare R2 would be amazing. I want all my stuff on that nice canvas!

1. I would like to have a command palette item that allows me to abort a deployment

2. Sometimes when using the deploy latest commit from the command palette it doesn't deploy the latest commit but the latest deployed commit

3. It would be nice to have a button for the deploy latest commit thing

4. I would like to have a way to skip deployments besides watch paths. We have WebSocket gaming servers, and those cannot be redeployed, so I have to constantly do the abort and deploy commit thing. This could perhaps be an internal endpoint like a health check or maybe reading the commit message.

I would like to have below:

• Support for MariadB deployment

• Ability to browse Volume content, upload/download and ftp support

• Terminal for docker

• Default IP for root CNAME instead of relying to specific registrars

• Avoid redeploy after env variables change, as most of the times only restart is required to apply the change

• Deployment needs enabling cache, this will make deployment very fast

All great points! but you can deploy MariaDB easily!

https://railway.app/template/Onvy0F

I love your product, but some missing things/features keep me from putting business-critical applications on the platform.

* I need support for log centralization/forwarding/alerting. It's impossible proactive follow up the logs of multiple projects. Some things that could help:

* being able to forward/filter the logs to a log centralization platform for further processing and alerting. It would be great if we could, for example, access the logs with Fluentbit so we can filter and forward them.

* being able to configure alerts on the logs when certain events happen.

* Descent Terraform support. With respect for the maintainer I can say the current Terraform community project is a good start but is in my perspective a litte buggy, not fully documented and sometimes instable. I played a few hours with it and had a few contacts with the maintainer of the project, but in my opinion it's not production ready. I think if Railway want to grow to bigger customers, infrastructure as code support is crucial.

* Native inbound firewall support without the use of Tailscale

* Private IPs for incoming requests so we don't need to rely on Cloudflare

* I don't see why Railway has to change my severity level names; it's application-specific

* An object storage compatible with S3 would be a great addon

Long environment variables such as connection strings have a horizontal scroll bar. The nit I have is that this scrollbar almost takes up half the height of the environment variable when showing the value and obfuscates it. Would it be possible to increase the height of the environment variable rows so the scrollbar does not affect the visibility of the environment variable value?

It would be great to see the name of the branch instead of the ID when having multiple PRs open.
Currently, I have to check GitHub's PR IDs to know which environment each PR is from.

I completely understand why multiple regions is a Pro feature - it's a great way to segregate the audience and offer market pricing to allow global services to host accordingly. However, I have a feeling that putting every non-pro project in a US region is just making it harder and more expensive for small projects outside of the US. Currently all of my requests on my 'hobby' project literally circumnavigate the globe, despite being a small local project. I will buy Pro if my app takes off, but right now it is literally a hobby project.

I will reference a few issues/feedbacks I opened lately, if they can be included in here:

• If I remove an active deployment, an old failed deployment is being "promoted" as active (this failed deployment is 15+ months old): https://discord.com/channels/713503345364697088/1279491632969810084

• Volume capacity not showing on mobile: https://discord.com/channels/713503345364697088/1286418193178431488

• Removed volume are disappearing before "applying the changes": https://discord.com/channels/713503345364697088/1284428336432681042

• Databases, like Redis, have links that points to nowhere in the service settings "Config-as-code" or "Feature-flags": https://discord.com/channels/713503345364697088/1279492563044139068

• Color label should be "purple" and not "pink": https://discord.com/channels/713503345364697088/1277893311444815872

Log checking works really bad for me. When I try to check logs from the past (2 hours ago, 3 hours ago, day ago, specific time etc.) it just never works. I only can manually load previous logs by clicking a "load more" button which is really annoying and what's more, I'm not able to go to the moment I need because my logs are really long (like 100 lines every few seconds)

stage changes is pretty junky, specially when editing or importing env variables from other places. Sometimes it counts more changes and then reverts itself to the correct amount and sometimes i've to double click the apply button changes to make it work.

deployment removal is also pretty junky, sometimes the state of the deploy history rollback or goes completely wrong.

1. Pasting in the environment variable name field overwrites existing text.

2. It would be helpful to be able to rename existing environment variables.

I think the Webhooks tester is broken.

When I clicked the "Test URL" button, it showed a loading state for a bit, but my webhook receiver didn't receive any request (or it didn't receive any that I managed to see - I could've had more transparent logging).

But after I click "Save Webhook" and do a deploy, my webhook receiver receives it without a problem.

Platform is quite unstable. Sometimes cannot even access the projects which is quite stressful. It comes back up after 20-25 mins.

seiyial, this isn't quite the place to report clear platform bugs, please open a new separate help thread.

Being able to redeployed an old removed deploy
ment

The system that railway uses for when you do a git push, that the old deployment is not deactivated until the new one is already running sometimes gives me problems. There are some projects that only 1s running in parallel can cause me account/API bans. So before deploying I always have to go manually and do a remove. It would be useful an option to change that behavior, so that when you do the push, before deploying the new version, the old deployment stops completely and they don't run in parallel. Or even that from the dashboard by right clicking the service there would be a direct access to the remove. Thanks!

rgmvisser, you can, for deployments newer than two weeks - https://docs.railway.app/reference/deployments#redeploy

ciphersight, this is possible, set a RAILWAY_DEPLOYMENT_OVERLAP_SECONDS service variable to 0 - https://docs.railway.app/reference/variables#user-provided-configuration-variables

@brody I see now, however this is only possible it seems for successful deployment. I got an transient error in the build process and was not able to retry/redeploy that commit. (Happy to take this to a separate thread)

ciphersight, this is possible, set a RAILWAY_DEPLOYMENT_OVERLAP_SECONDS service variable to 0 - https://docs.railway.app/reference/variables#user-provided-configuration-variables

Thanks @brody didn't see that!

As @okarmazin already wrote (in his longer post):

Please implement standard password login. magic links are an annoying gimmick and shouldn't be used as a primary login method. You are the only service that does this. We have our password managers, guys. Let us use them.

Passwords are inherently insecure, best we could do would be additional SSO providers. What would you like to see there?

There is a change that requires me to manually approve deployments instead of automatically deploying updates that hit my repository.

I don't see the logic of insisting that everybody who might contribute to my repository necessarily should also be a user on my railway account. That is extremely irritating and feels like a very underhanded way to try to increase revenue and it stinks.

That's the kind of move that would make me want to look elswhere.

@sinscrit, why not just have them create a PR and you approve the PR + merge into your main branch?

• Build/deploy timers almost always have stale values when starting a build or switching to a tab, i.e. for a couple of seconds it might show that a build has been going for 20 hours, before showing the correct value.

• I would love to be able to deploy by tag. This way we could have all commits on main go out to a staging environment, and whenever we push a tagged commit it goes out to a production environment. This is really easy to set up when using build triggers in GCP, but from what I've found cannot be done in Railway which forces us to do some minor but annoying branch juggling that could be avoided.

@angelo - on passwords - at the moment with "magic links", Railway assumes that my email is protected by something other than passwords? But if my email is also password protected then how are magic links any more secure than just using passwords? It's a more convoluted and annoying way to get the same level of security as my email provider - so the current solution doesn't provide any additional benefits... only additional annoyances.

Personally I'd be happy with passwords + TOTP - that would allow me to just be able to access Railway without the gymnastics currently required.

If railway does want to use something else though then perhaps create an additional option to use Passkeys?

Maybe also ask for email OTP verification IF someone tries to change security settings or critical settings on an account, instead of doing it every time I try to log in?

@artvilla The issue here is that it still requires me to approve things. I have different people contributing and different people giving feedback on the contributions. I don't see why I need to be involved on every minor update.

I want to be able to customize the databases template before deploying

Currently when I choose to add a new postgres database, it's instantly deployed. I want to be able to, for example, remove public proxy, change postgres version ... etc

It is very frustrating to be forced to use email verification for every single Railway sign-in, rather than logging in with a password and TOTP like every other service I use. This is especially tiresome because I need to have two separate Railway accounts, one for personal use and one for business use. (They need to have separate payment instruments.) I checked my email inboxes and I have clicked through 61 of those verification emails.

I submitted this feature request previously at https://feedback.railway.app/feature-requests/p/password-login, where it seems to have been ignored and then deleted (the page now 404s).

raxod502, Is GitHub login not working for you?

raxod502, Is GitHub login not working for you?

I do not voluntarily use third-party OAuth login for any website, including Railway. This is for a variety of reasons, including the fact that it makes keeping track of which accounts I have on which platforms more complex. Logging in with GitHub is not a suitable solution for me.

You can see Angelo's answer above, but we don't have any interest in doing password based logins.

Passwords are inherently insecure

That seems like an overgeneralization. Passwords and TOTP are good enough for GitHub, PayPal, Gmail, Amazon, Cloudflare, Namecheap, and all my other VPS / DNS / PaaS providers. Is there a particular threat model you have in mind that has led to the decision to diverge from standard industry practice in Railway's case?

(Also, I can log in to my GitHub account with a password. If passwords are inherently insecure, why am I allowed to log in to my Railway account with my GitHub account?)

@brody - "Not interested"... sure... Seems like a popular suggestion which several users have brought up - care to elaborate / consider it?

Or are you saying that this feature request is being dismissed because it isn't "interesting" enough for the railway team?

I don’t think we use “interesting” as a category for what we consider for prioritization. This actually sparked a conversation within the team on what we do try to prioritize on for each quarter.

If I had to roughly summarize:

1. UX pain

2. Business impact

3. Use-case enablement

4. Urgency (“Will it kill the company if we don’t do this?”)

I wanna disambiguate Railway’s stance on authentication. I think it’s okay to complain about the passwordless login flow with our vendor Magic. We aren’t too happy with it as it stands. However, we frankly don’t wanna store a bunch of password hashes. We never want to get into a data breach, but staying password-less as an org reduces the appeal of a attacker to go after your data. We don’t ask for much, heck, let alone care for your data (within reason) - adding passwords would need to demonstrate a critical blocker from you or others to consider adding to our threat model. Someone mentioned that we deviate from practice- I think imo that’s okay if it shrinks our threat model.

However, if the ask is, “Your current login flow sucks” - I can see that and we can work from there, maybe it’s native passkey support, maybe TOTP hardware support, additional SSO providers- and if passwords are super important, you can use GH. However- we would need to learn more about your use-case.

As it stands- I would need more data to make a informed decision for us to change our stance on our roadmap.

I’ll motion to move the “Add Passwords” discussion over here

https://help.railway.app/feedback/support-password-based-logins-5ba3d487

@angelo - thank you for elaborating on the previous rather dismissive response that railway wasn't interested in doing passwords...

The information you shared helps clarify that the concern isn't about end-users misusing passwords but that storing hashes would make railway a target for hackers.

I'm not sure how Passkeys work, but if implementing that is something Railway could support then that might be enough - most password managers now support passkeys.

The conern I have with using SSO is that it makes me dependent on the SSO provider.

As an example, several years back I deleted my facebook account, and then realized that I had used facebook to SSO into just a couple of sites which I could no longer access...

In general, I find SSO makes it difficult to close accounts / maintain a level of control over my information on the internet - and personally I'm uncomfortable doing that.

(However the same concern is with magic links too at the moment - sigh, so I guess in that sense it's a moot point - but personally those are the reasons I would prefer passwords + TOTP, or even passkeys).

@jake - noted, won't comment on the password topic here after this

Builds show out of order when multiple building at once.

Here you can see the newest one is shown second:

Double-clicking the service pane highlights the entire app

• Git LFS support for larger files like ML models. Possibly directly connected to existing volumes.

• Ability to browse and modify data in a volume.

• US East region in Hobby play.

Re: -ok-, I will be sure to move my response in that thread, but super duper appreciated. Forgive me for the initial blaze response, trying to balance spewing walls of text and being curt- you can tell I am not calibrated enough there.
Re: @gschier - Lineared, really wish we have comment to Linear, would be nice and fun to do.
Re: @pietz- noted!

1. We are getting hit with those large image sizes thanks to CUDA + SciKitLearn/TF, really kills our build times. Heard here.

2. Would basic file navigation work for now, what's the most critical thing that we could provide that would unblock you?

3. Soon, we will have Railway Metal East available for testing within this quarter.

Update for everyone on this thread, we just solidified our provisional roadmap for Q4, a lot of feedback from this thread is going in, esp. on the DB experience front. I won't spoil it but will reply in here once we have all of that in place.

Edit: Roadmap thread posted and boy do we need to add pagination to Forum posts like this. Check it.

I realize this is a super old thread, but wanted to add a minor nit: tailing logs via the UI is annoyingly flaky.

Ideal: logs are spewing out, I scroll to the bottom, as new logs emit my view stays pinned to the bottom, and old logs flow out of my view

What actually happens: it works like above for a bit, but then sometimes a new log being emitted causes my view to unpin from the bottom, and my view stays static from that point forward; as new logs are emitted I don't see them at all, aside from the scrollbar on the side indicating new content.

I would love it if the dark/light theme supported a "system" version.

Having ``[skip ci]`` would be amazing.

re: Don't worry Sina- we love talking about whats in here in our weekly calls. With that said, logs are much better than what they initially were on the platform but still can be much better.
re: Vimtor - honestly pretty good idea!

Mrxbox - can you explain more on what happens, I don't think we force CI checks, can you expand more on your usecase?

My common use case is "change some code + change env var". Currently this results in 2 deployments and adds a bit of confusion

I love the "Architecture" visual with services and arrows between them, but it's hard to understand what environment variables are responsible for a particular arrow. It'd be awesome if I could click on the arrow and Railway showed that it represents DATABASE_URL

w1nt3r-eth, RE: double deploys - Assuming the variable(s) you change / add are directly related to the code you had just pushed -- You can first make the needed variable changes, then save them without deploying via the alt + shift + enter shortcut, then when you push your code shortly after, the variables will be applied to that deploy, where they are actually needed.

This also goes for many other config changes so you can minimize the amount of deploys, want to add a health check but you are going to push some code in 2 minutes anyway? save that change without deploying and then your next deploy triggered by the new code will have that health check!

Now you may be thinking, why isn't this talked about / documented, and that would be a great question -- we want to keep the platform simple, and having the user choose if they need to apply without deploying or apply with a deployment is not simple, open to feedback on this though!

When are you planning to expand your servers to ASIA?

Can we at least have access to customer support. The customer support here is very bad. Actually, it is horrible. I have been trying to get someone from railway to help me with the issue of mail queuing using supervisor but to no avail. There isn't even a reliable way to talk to a team member and all my threads are being ignored

iwatrouve, I have searched the entire help station, but the only message you have ever sent is this very message I'm replying to -- how have you tried to get in contact with us?

I created a thread when I just got a pro account explaining that I wanted to know how to modify the nginx configuration file for my vps server and it never got answered

I sent you emails even before I choose a pro subscription and it never got answered to

I'm sorry but I'm not seeing any threads that you created on this help station, can you link me to them please?

Here in the image I even sent you guys an email and got an automated response

Following that, did you open a help station post?

The only option I have is creating a new thread which I did after purchasing the pro version

Currently I have a problem with setting up a problem with supervisor and setting up a background process for email queuing on the vps and there are no resources to help or customer support. I am actually considering redeploying the application on a vps and leaving railway because of the lack of support. I have clients currently using my application and I cannot afford no customer support or unclear documentation.

And it seems my thread even disappeared I cannot seem to find it anymore

We soft delete threads, so my search would have returned deleted threads too, but there have been so threads from you at all, deleted or otherwise, at least from the account you are currently commenting with.

In the interest of keeping this channel on topic, please open a new thread to clearly explain the issue and what you need help with.

Would you also mind if I cleaned up some of these messages to keep this thread tidy and on topic, since I don't think the language used was particularly pleasant, and we couldn't have helped if you didn't open a thread when our systems asked you to.

I understand it might be upsetting what I am saying. But I think focusing on the problem is more important than protecting your image. I am not trying to be rude I am just being honest with my experience here. Maybe have someone technical who replies to mails and threads. Because I am sure I sent a thread except if I am mistaken. I do have another account, but it is a free account, and I am sure I cannot post threads using that account

Because it is a free account with no customer support

You can post threads with trial accounts, if that is the case, it indeed could have been missed, we are a small team and need to prioritize answering threads from paid plans, I hope you can understand that.

I understand your issue as a team, but I also hope that in the future your customer support improves at least if you want to attract enterprise clients. Because I actually wanted to improve to the enterprise plan but I got scared off because I assumed if I am currently getting this kind of support it will only get worse as we scale

I think we have a pretty good support system for Pro users, unfortunately we aren't able to reply to a pro user's help thread if one was never created.

Pro user's have a support SLA of 12 hours, meaning you can expect a response within 12 hours, and even faster if its a critical issue.

We just can't offer similar to trial users, they are generally limited to community support.

I just sent you a new thread again

Hopefully you got it

Hello I just spoke with one of the railway team members and they assisted with my problem. I wanted to just mention this and appreciate them publicly for handling my issue and giving me a solution

I have an issue with static outbound IP's.

We have one service running typesense which is only accessible publicly (typesense doesn't support ipv6). In another service we are connecting to the typesense public URL without issue. As soon as we enable static outbound IPs on the other service, the connection becomes unstable and very slow with most requests timing out.

dane-stevens, I wouldn't call that a nit, please open your own help post for that issue.

Cron reliability

I've had a cron job running daily for months... Had a few skips earlier but couldn't reproduce the issue so nothing was done earlier, but the railway team said they've fixed cron jobs / improved reliability...

Well, just this morning it skipped again.
No changes / deployments on the project for a couple of months now, it was running smoothly.
No logs from this morning either - the cron just didn't even attempt to execute.

Cron reliability really needs to be a priority - and perhaps add some metrics to monitor how frequently cron jobs are skipped so the team can monitor cron reliability over time.

It will be great to have a more detailed costs breakdown in the projects setting ("Usage" tab). When opening account usage you can see breakdown by individual services in the project, but it's missing in the projects setting.

Also having a breakdown by environment will help. Something like this:
- Production:
- Express - 4$/m
- Postgres - 2$/m

- Development:

- Express - 2$/m
- Postgres - 2$/m

Hey @-ok-! We've definitely pushed a ton of work into cron reliability. It's actually at a point where we're pretty satisfied with it internally (finally), but we're not satisfied if you have issues with it!

Can you please add what your issues are inside the "New Cron Experience" thread?

File uploading in the dashboard would be a big plus, im tired of getting random URLs from the internet and putting in the platform for service icons or my profile

Update to the Hobby folks on this thread, we just shipped US-East to Hobby now since we have the computers for them. Deploy away! (Stateful/DBs coming soon there.)

Loving Railway!
One small thing I find myself constantly concerned over is egress fees.
For context I'm running a service that generates images (and potentially videos in future).

I need to upload these generated images/videos to a S3(R2) bucket for user access and I find myself worried about egress when I add more complex functionality that may generate lots of media.

PLEASE add a confirmation button of some kind to the railway browser login experience. Maybe I'm missing something, but it seems like I could generate a login link via the CLI, get someone who's already logged in to railway to click it, and have immediate access to their account until they figure out how to delete my access. Hopefully you have some safeguards in place, but I would feel better with a button to confirm, like every OAuth implementation uses.

If I'm right, the only way to secure your account is for everyone to not be logged in to the railway UI. And regardless, the current experience made me seriously question the security of railway as a whole.

sometimes when deploying new templates for myself I forget that I'm deploying on my company's account (massive skill issue).
Maybe showing the current team on the navbar would help, at least some indication.

Biggest papercuts at the moment on Railway are...

• Private Networking not being available during build time? Super odd, shouldn't require a custom nixpack just to get builds from one provide to another to work. Also since private network is being pushed so hard, builds connecting to DATABASE_URL should absolutely work

• Let me remove my personal account entirely? I just use this for my business, but I can't remove my personal "workspace"

• Let me set a logo per project or a branding color or some sort for differentiation

• HUGE: please provide automatic networking configuration if a www and a root/apex domain are routed into a dyno/project. Vercel does this automatically, I don't want to have to configure this on Cloudflare, I want www to automatically route to root and vice versa and this be controlled on railway.

• Potentially offer up an analytics package like Vercel, this feels like a stretch though unless aligned with longer term mission

1. make it so that this input gets picked up by password managers as a valid OTP input box

2. for help threads can you put a reply button at the top (on or right below the starting message of the thread)

• Better support for uv (e.g. https://help.railway.com/questions/can-t-use-cache-mounts-on-shared-dockerf-e578f8ca)

• Ability to move services between projects (have a DB that I created in a different project and want to move to new one after closing down old project)

• Better UX + more details on metrics dashboard (e.g. I can see recent commits <1 day on the charts, but not older commits)

I would love to be able to set a different default region for my Hobby account (being located in the EU is not really convenient to have us-west-1 as default).

We are going to be introducing a EU region for Hobby users soon!

It would be great if each template included the name of the region that it's deployed, see screenshot with proposed change it would have saved me a good 2 hours of debugging my redis latency issues today!

Introducing more pricing tiers between Pro and Enterprise could be a game-changer for Railway. There’s a significant opportunity to cater to users with pricing points like $100, $250, and $500 tiers. These options could help bridge the gap between hobbyists and large-scale businesses, offering more flexibility for scaling projects without jumping directly to Enterprise-level commitments.

Personally, it took me quite some time to decide to migrate to Railway because I was unsure how well it would scale. This hesitation might be common among potential users who are evaluating their options but aren’t ready for a big leap into Enterprise. Mid-tier pricing could provide a sense of security and a smoother growth path.

Why mid-tier pricing matters:

• Attract Growing Teams: Many startups and small businesses outgrow Pro but aren’t ready for Enterprise. Mid-tier plans can capture this audience.

• Build Trust with Scaling Users: Offering flexible options fosters confidence in the platform’s scalability.

• Encourage Adoption: Lower barriers to entry for scaling teams could reduce churn and bring more users into the Railway ecosystem.

Thanks for the comment, it's something we've been taking a look at.

What would be your expectations for this class of service?

I believe these mid-tier prices should primarily enable access to higher RAM and vCPU allocations per service, can't think of anything else that would be needed at those tiers.

Edit 6. 5. 2025: As of today 1Password recognizes this input. Don't know if it got fixed on Railway side or 1Password side.

As alexng353 wrote: Let the OTP reauthentication popup be recognized by password managers. Reposting sice his image upload doesn't render.

Not recognized by: 1Password

When searching for a template on canvas (or anywhere, really), let us filter out community templates. I would like the search results to only include templates created by Railway.

Can't attach an illustrative screenshot, because the file upload request ( POST https://station-server.railway.com/api/upload-attachment ) currently returns HTTP 500

Can't attach an illustrative screenshot

Sorry bout this, it's been fixed

Thats is really great to hear