IIRC 100GB hard limit

But what is 10GB limit tho? Which limit actually gets applied? It is confusing

There is no 10GB limit, it's 100GB

was this a typo in documentation then? I see that you made a correction and docs got updated. But there is still one of the 10GB mentions remaining

@pepper - pretty please.

It's 100GB

Whoops. Sorry. I'll fix that real quick.

so as long as my server file system is under 100gb, i can cache as many files as i can in /tmp folder, is that right?

Correct

Though, that will likely change with runtime v3, as in ephemeral will be reduced.

what will be a new limit?

Undetermined at this time.

It might be best to use a volume.

@Brody

question 1:

• is it also possible to attach volumes to railway sandboxes for running ai generated code on files that are in my volume?
• what about attachinhg the same ephemeral storage to railway sandbox that my server is using?

.

question 2: 100 gb volume costs $0.0156 per month if running nonstop, is that right?

neither is possible, but you can use the sandboxes files API to inject all the files into sandbox, or just download it via a command from within the sandbox, read more here:

no, by my calculations, 100gb per month in a volume would cost: $15.5

-# 0.00000006 * 60 * 60 * 24 * 30 * 100

object storage is much chaper though

(also pls don't ping team members)

sorry about it.

can i have persistent file system or volume attached to any future sandbox that gets created rather than fetching needed files inside sandbox every single time?

You can fork sandboxes, that way you initialize one sandbox filesystem and fork it into a new sandbox.

https://docs.railway.com/sandboxes#forking

sorry but does it mean that if i wanna keep my file system for sandbox, i have to always have one sandbox running so that new ones can fork from it and have the same file system?

if all sandboxes stop running and i lose file system, right? So I have to pay for long running sandbox just to keep file system? If this is the case, it is gonna cost a lot I guess.

I couldn't find any info on sandbox pricing. I assume you still don't have since it is a new product.

Also out of curiosity, why not add support to be able to attach volumes to sandboxes?

If your goal is having a long running sandbox with a persistent file system, then I wouldn't recommend relying on forking. If something goes down, say an incident that causes sandboxes to go offline, then your filesystem disappears with it. It's not reliable

You could maybe look into checkpoints though, it's basically the same as forks but it persists even if the sandbox dies

-#

sandbox pricing:

• Memory: $50 GB / month
• vCPU: $50 vCPU / month
• Egress: $0.05 GB -#

I'm not sure why Railway doesn't support mountable files in sandboxes, sandboxes are a very new feature, so additional features might show up later

I don't wanna have long running sandbox. I was asking whether it is only way for having shared file system with other sandboxes: parent sandbox lives forever so that other sandboxes can fork from it??

you cant have a shared filesystem but yes you can inherit the files of other sandboxes via forking and checkpoints

and I'd recommend using checkpoints over forking for your usecase

so it means that i dont need long running sandbox, right?

if you use checkpoints then no you dont

that's nice to know. thank you very much for your help 🙂

of course, happy to help