I'd like to reserve a minimum vcpu/memory allocation for a service. Scaling to zero is great, but some apps perform poorly on a single vcpu, id like to just pay to have 2 or 4 vcpus always reserved for some workload. Thanks

Take a look atsomeofthesestandalonenits

Also another one: There should be a comment preview feature here.
Edit: Also another one: Fix rendering of adjacent links

Deploying a template from the marketplace didn't ask which project to deploy to. It just blindly created a new random project against my intention.

Template search results on the dashboard canvas should have links to the actual marketplace page of that template. I don't want to blindly deploy whatever based on 5 words of description. Why can't I open the template page from the search bar?

In the project canvas, would be cool to be able to:

• duplicate whole groups of services,

• select multiple elements (either by Ctrl+click or rectangle select tool) and then perform bulk actions on them (like duplicate or delete),

• use Ctrl+C, Ctrl+V for the duplicate action,

• copy/paste elements between projects without the need of creating templates,

• create private templates.

Since you are on the Pro plan, every service has 32 vCPU and 32 GB of memory available at all times.

If you are maxing out at ~1 vCPU is likely because your workload is single threaded, and thus, it could benefit from horizontally scaling with our replicas.

I'd love if builds were more heavily cached. My service takes 5min to build without the cache, but 10sec if all dependency installation layers are cached. But despite only changing a few files, it often has to do a full build, including downloading mise packages. I'd love if each account got an allocation of cold cache space, so even if it's been a week since the last build it is still available rather than being purged instantly.

Mise packages also take a surprisingly long time to install, so I suspect they're being fetched from the source every time. It'd be great if Railway ran an internal mirror/cache to speed them up too.

Finally, as a stretch goal, it'd be awesome to get feedback in the build logs of why the cache isn't being used. Something like "Cache skipped because environment variable ABC_XYZ changed" or "Cache skipped because file package-lock.json changed" would be immensely helpful in knowing if the cache busting is intended or not.

Also shared volumes and ability to expose more than one port publicly would be great.

I have the following items on my wishlist.

Technical:

1. Roll out a basic WAF for which I can create basic rules by clicking on a http request in the logs.

2. Improve the GitHub integration when deploying a web app or website. It would be great if there was a way to automatically generate a preview URL and mention that URL in the PR comment on GitHub.

3. Making database deployments with backup as one-click as possible. Bonus if I can deploy a cluster this way.

Communication:

4. Add an RSS feed to https://railway.com/changelog or make it part of your blog.

5. Post to Bluesky.

I'm very much looking forward to the "Copy Volume Data" feature returning for when preview environments get spun up. This was a very handy feature.

Maybe this is a bug, but I cannot seem to get the "Base Environment" setting when set to "No Base Environment" to work properly. How I'd like it to work: I have two git branches, "main" and "develop", with each branch connected to services in a Railway environment ("production" and "develop", respectively). When PRs targeting "develop" branch are opened, I want the preview env to be based off the "develop" env. When PRs targeting "main" are opened, I want the preview env to be based off the "production" env. That isn't happening and all PR environments are getting based off "production".

• Sticky sessions for replicas. This would be helpful for websocket servers

• Read replicas across regions for databases

Many bounty questions are related to basic content, like if the private network isn't working or how to set up custom domains. Perhaps you guys (and girls) could add an FAQ section with a link to a YouTube channel where you explain these concepts... and the next time someone asks a question with an existing solution, we'll simply redirect them to that specific FAQ. (i know you have the docs, but apparently it is not enough, maybe a FAQ approach solves the problem)

Don't get me wrong, I love the idea of bounties rewards, but maybe this will help someone solve their problems faster.

When adding a custom domain to a service a window pops up with the value for the CNAME record, if you dismiss this window there is no way of checking what the record value was, if this is intended add a note to say "we'll show you this only once" or something like that.

You cannot view the value if your DNS record is correctly set. Unable to view the instructions = you set the value correctly. If you didn't set the record yet, or if it is incorrect, there should be a button that opens the modal again.

Though I agree there is no reason to hide this information from the UI. It is indeed confusing since there is no explanation as to WHY it refuses to let you view the instructions while also not showing any "success state" either.

I invited another member to my workspace, when they navigate to that workspace it takes 6 to 10 seconds to load that screen with the list of projects, it's really slow for some reason.

Command based health checks would be amazing! simply a bash command that can exit depending on the containers status.

It would be great to be able to set a default resource limit for all new services, something similar to the preferred deployment region, but for resource limits.

In my team, we almost always set the same vCPU and RAM limits (close to the minimum) for all projects, to prevent resource leaks and unexpected charges.

While it's great to have access to 32 vCPUs and 32 GB of RAM on the Pro plan, most of our simple projects — like APIs or microservices — don’t need that much power. Having a default limit would save us time and reduce the risk of overprovisioning.

These deploys are ordered incorrectly. The newest build should be first.

Also would be great to be able to set $ Usage limit per project.

I've deployed quite a few templates and have had to make many hacky workarounds to limitations. For example, we can't do the following:

1. Multiple volumes to a service

2. Sharing a volume with another service

3. IPv4 internal addresses (lots of issues with Redis over IPv6)

4. Specify multiple public addresses in the template. Some services require multiple ports (maybe 8000 for API, and 9000 for app). We could specify this after the template has been deployed with custom domains.

5. We can do pre-deploy commands. What about post-deploy commands? For example Apache Superset has this dumb requirement that you use their CLI to create the superuser after deploy. Would love to automate that in a non-hacky way in the template itself

6. Supplying configuration files in the template. Some images require you to pass a config file in a volume.

Also, US Central metal region pls

It would be great if state were preserved when navigating between Architecture, Observability, Logs. Essentially, proper tabs.

I would love to be able to set a "primary domain" for a service. That, or change the domain link within the service pane to a dropdown if there are multiple.

When changing a service variable it seems pointless having to rebuild the project all over again and having to wait 4+ minutes while it should have simply be a service restart to pickup the new variables

@kallikakis Railway cannot know if the variables are used at build time. They are available there and your application may use them at build time.

But your request is sound in principle. I would amend it to: Allow users to choose whether service should be fully rebuilt or just restarted when applying variable changes.

Being able to completely clone a PR environment directly from the Railway CLI would be great but the last remaining piece requires adding deploymentTrigger to the environment clone command. Added the feature request here too https://station.railway.com/feedback/add-deployment-trigger-option-per-servic-2f992b8f The need for this is that if you create an environment clone at the moment there's no way to connect it to a github branch...which is mostly the whole point of branching an environment. Right now we have to rely on a brittle github action which can fail due to Railway changing its API https://github.com/Faolain/railway-pr-deploy Thanks.

I hope to be able to configure the domain name of the TCP interface directly in the railway webpage, which is very necessary for us to connect smart hardware products to services deployed on Railway.

Additionally, I've encountered disk IO issues in the past two weeks. You can see the posts I've made on my account; the lack of timely responses has been very frustrating for me. I'm unsure if this is due to team members being on leave or other reasons, and I hope this situation can improve in the future. Some of our team's new services will be prioritized for deployment on Railway, and repeated periods of abnormality will cause the person in charge to lose trust in the platform, which will make my job very difficult.

What about choosing the region before deploying the template?
And I am really upset about the connection speed to db templates. I'm in South Korea and even though I moved database to the Singapore, I'm still getting at least 2s connection time from localhost and in production, I hit around 1s connection time even though I'm calling the database via private IP

The time displayed on a service card is the time of the commit, not the time of deployment (e.g. in the case of manually deploying a version) which can be quite confusing as it makes like things were deployed much earlier than they were

Notifications -- I'm a big proponent of Observability tooling, and general monitoring/alerting for all important lifecycle events - Deploys, Crashes, Errors (and error rates), etc.

Notifications via Email, Webhooks, etc. are critical, so I'd love to see Email added to your Alerting functionality.

1. Unwatched directories. Kinda the opposite of watched directories. I have a big typescript mono-repo where changes to my react-native app trigger builds for my server. I don’t want to explicitly add all the directories that get pulled into the server (since most do), just block the few that don’t.

2. Slackbot notifications for successful/failed deploys, crashes, and monitoring threshold alerts

As @matthew-magical, mentioned the time of deployment would be very helpful

Deploy without downtime when a service has a volume. I get that it's to prevent data corruption from multiple accesses but let that onus fall on the user's hands if they know they're using multi-access features that are safe (e.g., SQLite file access). Maybe put the feature behind a toggle with a big disclaimer or something.

I also really like the idea of shared volumes that others already mentioned.

I'd love to be able to run a CRON job not on the entire service so it triggers a rebuild, but like Coolify has it where you can target a script.

Railway official PostgreSQL template: Bad Defaults.

The Railway PostgreSQL template is configured to log SQL statements with argument values upon error by default. This can leak personally identifiable information or other privacy- / security-sensitive information into logs.

I think right now if you delete a volume, the name of that volume cannot be re-used, even if it doesn't exist anymore.

Would be nice for the name to be available again.

Maybe this is intended, but it's super annoying. Today I created a new environment and selected "empty environment." When it created the new environment it still had the groups that the first environment had (just empty, no services). I proceeded to delete these as I didn't need them. When I went back to the original environment all of the groups had been removed (services still existed). I was able to re-create the groups on the first environment without them showing up on the second.

Didnt search through here...but....the inability to simply copy your build log and your deploy log appears to be missing. At least I cant find it. Everyone is using AI now....any time you want it to analyze the logs, you have to click, drag....and hope that the whole log is there...or select all which grabs a bunch of stuff on the page and a lot of times, not the entire log itself. Can we simply get a "Copy Log" button at the top of Build and Deploy logs? That would be MASSIVELY helpful.

Groups are global across the environment. We understand this is a problem and want to fix it in the upcoming months

More reliable build and deployment timing and performance would be appreciated. It is often frustratingly inconsistent. Sometimes it takes 90 seconds, other times it takes 7 minutes. It seems like processes get stuck moving from one step to the next, such as from Build -> Pre-deploy, or Pre-deploy -> Deploy

I would love Dart support in Railpack. Right now I have to use Docker to deploy my Dart applications.

Would love an Australia region (ideally New Zealand), to meet requirements for systems that need to be hosted in countries with specific privacy standards.

I have another nit: preDeployCommand should have access to volumes.

According to the official docs, they explicitly don't have: https://docs.railway.com/guides/pre-deploy-command

Use case: SQLite database on a volume and need to run migrations as a predeploy command.

I would love to have horizontal autoscaling based on request queue time (or based on vCPU, RAM limit used by a service)

Environment variables groups. Often times there are services that have 15+ different environment variables that are needed. Maybe 4 of those are AWS creds. Maybe 4 of them are SMTP creds. Maybe 4 of them are creds to some other service. It would be nice to be able to group them and have a togglable header so they're easier to find/manage.

When a deploy doesn't happen due to "Waiting for CI" and CI fails, the deploy should be marked as failed with a proper error message, ideally linking to the CI job.

These skipped ones are also hidden below the fold so it just looks like nothing happened.

Linking issue from feedback section:
https://station.railway.com/feedback/connection-refused-when-service-sends-re-a0a97135

Disk I/O in Metal is abysmal. Totally unacceptable and rocked my confidence in Railway after I was sure I was gonna my 8 years old app here.

Also, logs are broken.

And to be honest, it doesn’t instill confidence that Railway’s support system is a public forum which blends community help for application-level errors with platform issues and bugs.

It would be great to have the ability to either clone the database volume when duplicating an environment, or be able to point at a database across environments

I get a 'deployment crashed' email every time I deploy a new app (standard express app) becuase the app gets a SIGTERM and exits with a nonzero code

Second on this one! Command based health checks, some way to add auth credentials to the healthcheck.

Configuring a service's dependency on other service(s) in order to deploy. Just like Docker compose "depends_on".

I discovered this issue when making a new environment in a complex project. The database service needs to be the first to successfully deploy. However, another dependent service deployed first and promptly crashed since it couldn't access the database service to make some startup migrations. This would be an incredibly useful feature that ensures higher successful deploy rate of templates. Looking forward to seeing this!

@benisenstein You can currently manipulate service startup order by using variable references: https://docs.railway.com/guides/deployment-actions#deployment-dependencies---startup-ordering

If A depends on B, just create a dummy variable in B and reference that variable from one of A's variables. Railway will then start B before A.

Thank you for the link! Surprised I never saw it. :)

Its not possible to create a new thread from mobile. Attaching screenshot.

I want a terminal for my deployment, if possible.

I'm not 100% sure if this doesnt exist somewhere, but if it does, its not readily noticeable . We'll deploy a new custom domain under networking, and then it validates it after DNS addition, great. But a week goes by and I need to validate on the DNS (Route53) side that someone didn't change the CNAME pointer...so I go back to networking and there is nowhere that you can click to show what the DNS is supposed to be to reference the proper container.

Bottom line, needs to be a way to look per domain at the "setup instructions" at any time so you can get back to validating the CNAME information. Once its validated, can't look again without deleting the custom domain and readding it, which we could never do if it was a production environment (which is the goal eventually).

Would love to have at least one region in Australia, preferably in Sydney, to comply with Australian Privacy Principles (APPs).

Have you tried the latest version of the railway CLI with the railway MCP server? You can ask it to find the latest deployment, read the logs and diagnose the problem

The ability to copy volumes between environments would be very helpful, both on environment creation (like a PR) or manually for refreshing data. Another nice thing to have is someway to download volume data or backups, either through the UI or an actual SSH service through the CLI so you can use something like rsync/scp.

Copying elements between projects is a great feature; hopefully, it will be available soon.

When I delete a service with an attached volume, the volume gets yeeted into the void and I have to scan around the project canvas to find it and delete it too.

It would be very useful to have an option to enable or disable Serverless mode in bulk, either for all services or just for selected ones in a project. This way we could quickly activate or deactivate environments like development or testing when needed, without having to do it individually.