Documentation on GPU support? Some threads show it's maybe possible? Unclear if so.

I'm a newbie, video tutorials on how to use railway like a pro would be fire; or on-hover explanations for every project setting ...if not...I guess I'll read the docs

I have two related wishes regarding volumes:

1. Support multiple volumes for the same service.

This would provide more flexibility for organizing and managing data within a single service.

2. Enable sharing the same volume across multiple services.

This can be very useful for workflows where different services handle steps involving the same group of files, or simply to change config files without requiring change the volume to a file server service.

Railway does not currently support GPUs, but it's being considered - see https://help.railway.com/questions/gpu-support-56d19c42.

a great feature would be to integrate a file browser directly on a volume. just like the database browser. i don't want to host the file browser template every time and unmount and mount the volumes to look at them

Advanced level load balancing strategies.

Many of us don't use Github

So I'd love to see built-in support to connect into BitBucket repos!

I'd love to see an sdk for accessing log drains, project usage statistics, etc. Even more so, being able to embed these components in another dashboard would add huge value to me and my clients.

Alerts and notifications on specific log filters

Do GitLab first, GitLab first!

We have a GraphQL API (https://docs.railway.com/reference/public-api) that you can generate an SDK against (using gql codegen tools). Is this what you're after?

Can you tell me more about your use of BitBucket, is this for a company?

Would love more information, what vendors are you using that you would like to export the data to?

App Sleeping using the CLI or Public API

App Sleeping is truly a great feature! It would be useful to be able to put to bed:
- A service (or group of services) during expected downtime or low usage.
- Apps that aren't actively used outside a given timezone or country, or on weekends.
- Staging apps that can be dormant outside of developer work hours.

Using the API or CLI, one could set up a script or chron job or what-have-you and do all this effortlessly. Win!

Come to think of it, is there truly ever a use case where one would want an actively used production app sleeping full-time?

If nothing else, perhaps a simple time range input next to the toggle in the GUI.

It's been asked about in the Help Station, but I thought I'd mention it here as well. Thanks, and kudos to the Railway team and community!

- https://help.railway.com/questions/how-to-set-app-sleeping-via-api-07a12afb

Bit more informtion/clarity about the followings:
1. Celery/Redis/Cron related information, how to deploy them
2.Delpoy any specific commit (Really useful if we need to rollback commit for any faliure)
3.Pricing clarity, what are the services can I use on given credit on a specific plan
4. Access Shell directly from dashboard
5. Download Execution logs irrespective of any specific build ( All logs at once , on given date range

1. We have an entire guide for that! - https://docs.railway.com/guides/django#set-up-database-migrations-celery-beat-and-celery

2. That can be done via the API.

3. It's super simple, you can deploy anything that uses under 8GB of RAM and 8 vCPU, no restrictions on what kind of services (besides TOS related stuff).

4. Planned!

5. Good idea!

Please add more locations outside the USA. Europe, Asia. Really need this as I have too much latency for my European projects

Ok, I see you have it in $20 plan

Allow for custom domains on TCP proxies

Prepaid Pro subscriptions. I have PTSD from
overages but want a higher level

You can set a usage limit on the Pro plan too!

Would you mind sharing your usecase?

Yes, we are running an IoT solution that receives sensor data via MQTTS. I am now using a CNAME record with a custom domain that links to the tcp proxy domain (which I think is fine for now and not a blocker). This means that all envs will have a different port to which our gateway devices will be talking, but the same domain. In our current deployment on AWS we have seperate domains for each environment and we expose the standard MQTTS port.

And another possible feature request: I have an API that is publicly exposed. I am not sure about the security features Railway applies to its HTTP proxies. However, I found that quite a few bot requests came through and so I put it behind a CloudFront distribution. It would be great if this step wouldn't be necessary and firewall settings could be configured directly in Railway. I had a look at Traefik and others, but didn't want to move the potential load into my deployment.

Hey, thanks for the response.

At the moment, we provide our clients with a heavily modified Strapi instance to manage their online services. I'd love to include actual server-use metrics to the panel.

If the beautiful dashboards you provide could be embedded in our panel, that would be very valuable to us. Ideally, these dashboards would have an API to inject custom styles into them a la Plausible.

Pre-build command step where you can run scripts like "nx ignore" (monorepos) https://www.npmjs.com/package/nx-ignore in order to skip builds if a project hasn't been affected

Frankfurt region (eu-central-1)

I haven't really followed what's exactly blocking this now (I've heard volumes that multiple services can access, but I don't understand why), but a Supabase template would be amazing for what I'm working on. Right now, I have Supabase hosted on their service, but then a few misc microservices hosted on Railway. They're able to talk to the Supabase database, but 1) I wish everything were in one spot, 2) It's not nicely separated with a private network, and 3) because of that, it's a bit slower to read/write to the db. It'd be so nice if everything could be contained in Railway.

Also, a super selfish one. I'm a weirdo that deploys Dart web services (mostly using Dart Frog), and it's a supported language in Nixpacks. But, for whatever reason, when I deploy a Dart server it runs on Dart 3.0.6. We're all the way at 3.6.0 (which is a couple years of feature development). I'm able to work-around quite easily with a custom Dockerfile, but it'd be nice not to.

Have you checked out the supabase template? -

https://railway.com/template/supabase

SSH connection to the docker containers. Must have.

SSH is planned for this quarter.

remove egress.

1. Australia servers.

2. To build docker images that use private base images

3. To can change the version of the deployed docker image from cli, to be able to do automated upgrades

This is a small gripe, but a fix would go a long way towards improving my UX:

https://help.railway.com/feedback/request-tell-us-about-your-nits-on-the-a3c5878e#5bmf

Hey, thanks for the response. I'll give that a shot and see if it meets my use case.

I'd still love the ability to simply embed the fancy dashboards you've already made, rather than create my own based on the gql data.

Can you direct me towards any of the GQL SDK codegen tools you're referring to?

Better logging support. The logs for the service I am running are out of order on timestamps when there's difference in milliseconds. Makes debugging a hassle.

TLDR: "Expand ALL" button while browsing logs

We are logging pretty big data structures in json format. Now to see the whole thing we either need to manually expand tree of nested properties or copy it as json and paste somewhere else both are suboptimal. A simple solution would be to have a "expand all" button next to "Copy as JSON". Thanks!

To fix this: https://help.railway.com/questions/issue-with-git-hub-repository-loading-tim-85e127de

that's my only wish

I would like to have below features:
1) Shared volume between service

2) Multiple volumes for a service

3) Browse volume content and allow download/upload or via ftp

4) Better Docker Compose support (related to above issue of volume sharing between services)

5) Backup project by downloading it as a single config file. This is helpful if you want to reuse the same project setup.

5) Allow username/password login to the railway

• Frankfurt region

• PITR for PostgreSQL

• Service dependency when deploying (Meaning, when I deploy changes to my frontend and backend, the frontend gets build & deployed after the backend)

• Create Environment From Backup (see: https://help.railway.com/feedback/create-environment-from-backup-90e52d3f)

  • This feature would be incredibly useful for testing with data that closely mirrors production (e.g., copying production data into a staging environment). However: how do you handle the differing environment variables between the environments? (I know the ,,Copy Volume Data" feature exists, but that solves another problem in my opinion)

Default regions when deploying templates. A bit tedious to spin something up and then migrate it.

You can update your teams preferred region via the command pallet.

It would be nice to have "placeholder" service for visually representing other interfaces (e.g. SaaS services like MongoDB Atlas). Background: For my dev and PR environments I am using a Railway MongoDB, but for my prod environment I am using Atlas. It would be nice to have a visual anchor that represents this (which would also hold the variables needed for access).

You definitely can!

Right click on the canvas -> Empty Service

Then name it, give it an icon, and set your variables.

When a service is at sleeping mode, any request comes will be lost and user will get 500 response.

The service wake up takes few seconds, Can you delay request delivery till all services wake up to get it successful?

Do you mean 502?

Internal server error

Not sure is it 500 or 502, but it feedback as internal server error, then second request is successful

Alright, please open your own help thread for that!

1. Better control for Backups. The backup feature is great and easy but still needs more features.

   1. Download DB backup directly.

   2. Ability to save backups in other locations (Ex. S3 or Drive). Since backups are very important it is recommended to have them in multiple locations.

   3. API to manage backups (Trigger backup, delete, restore, download/copy).

2. SSH into container (I see you have already planned ).

3. When duplicating environments have the option to include data as well (Ex. Create a staging env based on production env with all the records of DB)

   1. Implement the API to be able to this programmatically

   2. Add manage environments (create/duplicate, delete) in CLI

   3. Guide how to do this from GH actions

this!

Just starting to use Railway...cant tell if this is missing or I just dont see how to do it (Chatgpt doesnt think its available)...but direct integration into your AWS VPCs would be stellar.

Adding postgres extensions to the default postgres instance

Unlikely we may do that again, but I agree- our DBs need to be easier to configure.

This has been asked by a lot of our corporate customers- we have been using Tailscale for this for now, but I think it needs to be built into the platform.

First of all I love what you guys are doing. The biggest improvement to me would be the ability to view the full environment file-tree in the observability dashboard. Also read above that you'll be implementing SSH so that should help a ton.

Can we please add AU region as well.

Just deployed to railway. Really enjoyed the experience. Some things I'd like to see:
1) ability to reference private network url directly via a variable (apologies if this is already supported)
2) The ability to access the container directly via SSH (or at least some direct interface on the service in the Railway UI). I feel like this is a MUST HAVE.
3) docker compose support (the docs say this exists, but I could not figure it out!)
4) service dependency on deploy (ie, depends-on...)

Hello!

1. Yep very possible, use a reference variable - ${{<service name>.RAILWAY_PRIVATE_DOMAIN}} to reference another services private domain, or ${{RAILWAY_PRIVATE_DOMAIN}} to reference the same services private domain.
Useful docs -

https://docs.railway.com/guides/variables#reference-variables

https://docs.railway.com/reference/variables#railway-provided-variables

2. Soon, very soon (this Q) - Upvote this thread in order to be notified when we ship it! - https://help.railway.com/feedback/allow-ssh-access-to-containers-via-cli-e4985f6f

3. You can just drag and drop your docker-compose file into the project canvas!

4. Thats resolved automatically when you use reference variables!

One more thing to add...if it hasnt been...we have a lot of load volatility. Rule based horizontal scaling for a container environment is pretty much a must for our environment team. There's some specific workload that happens a couple of days a month that really requires a lot of dynamic compute. Having the ability to define rules like.... 70% capacity, spool up another instance...ducking below 40% spools one down...would be really helpful!! Enjoying getting to know the product...thanks for the efforts!

Thank you for your kind words willo2kk!

By 2025 Q1, I wish Brody’s ticket replies would slow down just enough for me to process what went wrong before he resolves it. At this rate, by 2025 Q2, I fully expect his responses to arrive before I even hit ‘send’ on my ticket!

I wish that in 2025, I would be able to pay for railway in my local currency

I want to see a fully fledged API for resellers..... If you want $1,000+ a month from me that would be one way to do it.

Can you tell us what is missing from our current public API?

I originally thought it was meant only for managing personal accounts rather than catering to resellers - but if I’m mistaken, then come take my money! I will go over the API docs now!

The tokens are bound to individual accounts or teams, limiting the flexibility required for a reseller managing several clients. With a cap of 1000 requests per hour, the API’s rate limiting may not be sufficient for the higher demands typical in a reseller setup. The API’s endpoints and schema are optimizzed for a single tenant’s use, rather than providing a multi-tenant framework that resellers require. What’s also needed is an endpoint that lets each user retrieve real-time usage data or check the current status of their environment. This capability would enable resellers to accurately monitor consumption and implement their own pay-per-use models. (While I could use APIs from Google or Amazon etc, that's beside the point. Railway already lays a solid foundation. Replicating everything from scratch would take over 6-12 months, whereas you lot already have endpoints for deploying templates and GitHub repositories. With just a few tweaks, I could develop Railway V2. ;) ) ...... I run a small provider and sold another in 2015 that's still thriving with hundreds and hundreds of clients. However, services like Railway show that legacy WHMCS providers are on the path to obsolescence, they simply can’t keep up with the pace of change. After using Railway, it’s disheartening to think that even with over 400 servers worldwide, I can’t match what they offer in any aspect like I simply refuse to start from scratch it feels worthless however if railway had an reseller API I would jump all over it. There's tremendous potential in a future feature that would allow users to integrate their own hardware, essentially turning the platform into a full-fledged SaaS solution. This would blend the benefits of dedicated hardware with cloud-based scalability, offering resellers a powerful tool to implement a pay-per-use model without having to reinvent the wheel. I currently utilise VirtFusion but even that being slightly different to railway it would still have major effects on it if railway was to adapt to resellers. Some top players in the industry are using VF but if they had a sniff at Railway even with losing some control it wouldn't be the case of crossing the mind it would be a instant switch or even providing both services.

Support python web app deployments better by using pyproject.toml files to define dependencies. Currently it seems only requirements.txt works

We support pyproject.toml files - https://nixpacks.com/docs/providers/python#install

Would be awesome to have a file explorer for volumes! or at least a SFTP connection

An official terraform provider that have full coverage of Railway API.

Little late to the party but:
- The ability to join the railway private network via a VPN / Proxy for testing and validation of railway environments and cloud/hybrid environments (Ideally via the Railway CLI)
- Shell access to containers / applications (I know this is planned but please, soon, please!). This would be super useful for debugging.
- Lazy loading / performance improvements to template search in the UI. It hurts to wait 3-4 seconds on a $2500 machine to search for a template. I'm willing to bet react-window or some paging would solve the lag without significant work on y'alls end.

2. It will be soon, before the end of this Q!

UPI payments please!!

Best we can do is cutting the egress bill by half.

You’re welcome/we’re sorry.

someone should update the templates page. its laggy to search. but worst of all. its bad at searching. like if i want to find "file sharing" it won't even find "flare" even though it's tagline literally says "file sharing"

imho this guide is still incomplete, not suitable for beginners and only details "how" to deploy and doesn't go into the "why" these services are needed.

For instance, REDIS is used but for what and why is barely mentioned. Imho it shouldn't be presumed devs know all the services offered on Railway for allot of devs come here because they are not dev-ops experts/new to it. Some tips & guides on how to deploy with e.g. FastAPI // SQLAlchemy would also be a nice addition.

Some improvements for the platform that would be great:
- Code as Infrastructure methods: Terraform or a new approach
- More teachings / tutorials for best utilizing the platform, common pitfalls, SaaS stacks, API stacks, How to test etc.
- Would love to see some love for API/token managment simplification with e.g. LiteLLM.

I would love to be able to deploy docker containers from a Azure Container Registry (mycompany.azurecr.io).

I would love to be able to do deploys from github via Tags and a regex expression so a tag like v1.0.2-rc.0 would go to stage, and v1.0.2 would go to prod and then be able to rapidly change deployed version from railway UI using the tags (with some age or number related limit). I don't like branch based deployments.

Don't delete the last working deployment, regardless of how many failed deployments happen after.

For vibe coders please implement a "copy to clipboard" button for the build and deploy logs.

GPU!

aumomate scaling base on metrics? not just hardcore the replica number
more RBAC, ex: member can edit in environment dev, stage... but cannot in production
sync selection services from env to env, not entire the env

This one should be live now!

SSH is live! Update your cli and connect using railway ssh

SSH is live!

Very super late reply, but my money is on us picking this up for next quarter.

Prosper’s been doubling down on our YouTube content as of late. Let us know what you think!

https://www.youtube.com/@railwayapp

I think this is going to be touched finally in our upcoming quality sprint. We can’t wait to make a bunch of the product materially faster.

I would like to:

Be able to configure the delay for services to enter sleeping mode.

The default is 10 minutes (according to the documentation), but I would like the option to modify it and have them wait 1 hour or 30 minutes before going to sleep.

In my case, we have it configured so that in the dev/qa environment, which is only used during the workday, it doesn't consume unnecessary resources overtime. However, when I develop, it often takes me more than 10 minutes to make a new request, and it's a bit annoying having to restart the service every 15-20 minutes.

Overall, it's still a great feature that I use and would continue to use because its benefits outweigh the drawbacks. However, I think the benefits could be improved with this.

chaloem khompitoon

Let's please get GPU Support, Advanced Load Balancing strategies, and better log exploration

Can you support distributed PSQL? I would like to install CITUS — but with no visibility into the underlying psql infrastructure I am not confident I will be successful.

2. will you help incoproate AI features to help with debugging why deploys fail? (edited)

3. [5:55 AM]

   would be mega useful

4. [5:55 AM]

I would love to have a template for an app that is private by default: for example a rails service that is connected via Cloudflare tunnel and uses Cloudflare 0 trust.

We already have a template for that! -

https://railway.com/template/cf-tunnel

It even comes with a guide!

I would like an agency plan or feature. You guys mention that you can invite someone to a workspace, and have them be them the payment method. I would like you guys to streamline this feature for agencies. I think it would be cool.
I would also like the volume browser. That is a very key service, and I cannot wait for it to be complete.

I would love to be able to send a request to a webhook whenever a health check fails or my service goes down. I use Discord for a lot of my observability and would really like to get notifications there via a webhook

It would be great to have a file explorer for volumes!

How about GPU support to begin with?

I'm on a free tier, and I like Railway's performance and capacity for MySQL database deployment, as well as their remote support. I think it would be nice if you guys bring back the 5$ refresh token per month on free tier!

And also would be convenient if i could see other schemas on the web ui. I would appreciate it <3

A button to force a repo to check github for a new merged PR. The automatic doesn't always work so it would good to have a fallback.

Let me download the "config as code" for my current settings

more transparency and stability, REALLY sucks waits 25min without any notice, status page shows nothing, all operating, how?

Thank you for the feedback, the status for the incident can be found here -

https://status.railway.com/cmbsswe9t001jtnuetts1sh1q

Please allow us to download DB backups!

Database backups are disk snapshots; they wouldn't be useful outside of the platform, so I would recommend pg_dump if you would like an off-platform backup.