API calls to my server fail randomly [FastAPI]

Question

My usual response time is <50ms. Here's my tech:

• Web Server: FastAPI+uvicorn
• Proxy: nginx
• DB: Postgres on VM
• DB API: sqlalchemy+pg8000

There are two docker containers (~1GB) running on the VM.

However, the request wouldn't reach the server, and would timeout with 404 with the following exception (not reproducible):

sqlalchemy.exc.InterfaceError: (pg8000.exceptions.InterfaceError) network error

All servers are Azure D series VMs and in the same VNET.

I've had this issue previously, which I was able to solve by having only one docker on that VM. However, in this case both images are lightweight.

Answer 1

There was a recent bug fix in pg8000 version 1.29.5 that may fix the problem you're seeing. Looking at the release notes the relevant bit is:

Fixed bug where pg8000 didn’t handle the case when the number of bytes received from a socket was fewer than requested. This was being interpreted as a network error, but in fact we just needed to wait until more bytes were available.

and the commit itself may be helpful.