Not able to sign-up in public instance about bakalarka-general

[BorysekOndrej]

This issue is fixed for the public instance and has been running for 2 days without a problem. It seems ok.

Long-term solution for self-hosted instances would be a Python script which would generate the appropriate Docker Compose file based on the amount of RAM and CPU a user is willing to provide. I have PoC, but I'm not yet satisfied with it. The general solution might be docker swarms or Kubernetes, though I'm reluctant to introduce another technology to the stack. For now I've created issue from this.

[BorysekOndrej]

Hello, thanks for reaching out.
The primary issue is that the container for the API server got killed by something on the host system. When I correlate the timestamps of the last messages from server logs and the RAM utilization graph, it clearly points out a sudden spike in RAM utilization (from 35% to 100%) at that time that lasted less than 5 minutes. So it most likely got killed by the Out Of Memory killer. This is a known issue and mentioned near the end of the thesis as future improvement. I will have it implemented by Monday.

The secondary problem is that my uptime monitoring did not send an alert, because I was monitoring a cached endpoint. Even though the API server was down, cached responses were still being served from a reverse proxy. I'll also have this fixed by Monday.

[adamjanovsky]

Great, I'll wait with the review then until this issue is closed.