I have a FastAPI based service that is borking on OOM when I try and deploy it. However, it doesn’t tell me how much memory I am actually using, which is hard for sizing. Also, when I run the same service locally it’s well under the 512mb threshold for the basic instance (this is a staging env so keeping costs down would be nice!).
Is there any way to get a bit more detail on what’s happening, and why Render is introducing about 300mb of memory overhead over and above what runs locally?
I am new to Render so apologies in advance if I missed something obvious here.
The ‘metrics’ tab can provide you with an estimate of how much memory your service is using at any given moment, though we acknowledge some limitations of the graph, such as its potential to overlook brief spikes in usage.
Render does not add any overhead to the instance. Thus, the memory usage displayed reflects the actual amount of memory your application is consuming within the instance, as we do not run any concurrent operations. Without the ability to run anything concurrently, pinpointing the exact cause of memory consumption is tough. However, processing files and the use of ML models are the usual culprits.