Let me start with, that I am running Raspberry Pi servers since the first Raspberry Pi was released more than a decade ago. Only problems I ever had until now, where dying SD-Cards with the first generation of RPIs. Since them I only buy really big high quality SD-Cards and I have RPI(4) servers running 24/7 for years w/o any troubles.

For a new project, I am running a web service on a Raspberry Pi ZeroW2 with an Apache reverse proxy on the same machine. Memory usage, even under load, is a maximum of 100 MB. This RPIZW2 simply dies after a few days, and I have no idea how debug this problem.

More details of the RPIZW2:

  • Uses Raspbian configured via Ansible to be an exact replica of my RPI(4), only Apache and a webservice were added
  • Quality power supply (original RPI hardware) and literally plugged to the same electricity circuit as the RPI(4)
  • The webapp is just a ‘hello, world’ with the current time and my internet connection is not fast enough to be DOSed
  • Monitored memory usage etc. for several hours and found nothing out of the ordinary
  • fail2bann is active and running
  • SD-card has several unused GIGs of free space and is same brand/quality as the one in the RPI(4)

Anyone experienced something similar? Has anyone an idea how to approach debugging this problem?

I am not sure that there is a better place at Lemmy for this kind of question than here. I’ll happily move this post to another place, if it is not appropriate here.

  • Eggymatrix@sh.itjust.works
    2·
    3 days ago

    The issue is that it is simply not built with reliability as a high priority so probably some hardware component shits itself too much after a while.There is a reason every reasonable company that needs a server to run reliably in production uses something orders of magnitude more expensive than a rpi.

    You lucked out with your previous experiences, but many others did not, or the industry would not pay the price of a rpi a month to run a machine with the specs of a rpi.

    That said, if you don’t need the reliability some easy hacks like a reboot cronjob or systemd timer, or trying to turn off unneeded services or peripherals could give you 90% of an industrial server’s reliability