Hello Zack,

It seems there's a slight misunderstanding about the memory usage increase you're experiencing. The issue isn't related to Metal itself - Railway simply runs your application and reports on how much memory it uses. What you're describing sounds like a memory leak in your application.

Memory leaks occur when your application continuously allocates memory that isn't properly released over time. This would explain why your memory usage started lower after deployment but has been steadily increasing.

Switching back to non-Metal might temporarily mask the issue, but if there's a memory leak in your code, it will likely manifest there too, just perhaps more slowly (because of the slower hardware) or in different ways.

You might want to consider adding some kind of memory profiling to identify where the memory leak(s) might be occurring.

Best,

Brody

Hey Brody,

Thanks for the quick reply. Can you help me understand how the issue wouldn't have occurred in the previous hosting platform but would with Metal?

What does Metal do differently that the legacy hosting platform does?

Hello,

I would love to help with that enquiry but unfortunately, we have no observability into your application so I cannot explain why your application behavior is different on GCP Vs. on Metal, only you can answer that question with profiling.

As for our side, the control systems, metrics systems, and runtime all remain the exact same, the difference being that Metal is much newer hardware, Gen 5 Xeons on Metal Vs. Gen 2 Xeons on GCP, accompanied by faster memory and faster storage.

I'm really sorry we cannot be of more help here, it's just that how much memory any given application uses on the platform is out of our control, it is solely depended on how much the application or software decides to use.

Best,

Brody

Thanks for the quick replies. Appreciate your help