I installed v1.7.3 amd64 on decent hardware (32GB RAM, 4TB SSD, 6 physical CPU cores). Every ~11 minutes, the node stop being able to receive inbound requests, and my browser tabs report ERR_CONNECTION_TIMED_OUT errors. This occurs even before setup (on the onboarding screen). As a constant workaround, sudo systemctl restart NetworkManager will instantly resolve the problem, but only for ~11 minutes until the issue reoccurs.
I tried reinstalling from scratch a couple of times, but the issue continues to occur. This same hardware previously hosted a v1.4.2 amd64 installation without issue, so I believe this is an Umbrel issue, not a problem with the hardware, network, or clients.
Any ideas what is wrong or how to resolve it?
It looks like this was being provoked by a misbehaving local network situation. Running sudo journalctl -u NetworkManager -b --no-pager | tail -200 just after a disconnect yielded this error:
May 25 17:45:58 umbrel NetworkManager[63198]: <info> [1779731158.0209] manager: startup complete
May 25 17:50:15 umbrel NetworkManager[63198]: <info> [1779731415.9590] device (enp1s0): conflict detected for IP address <REDACTED IP ADDRESS FOR UMBREL DEVICE> with host <REDACTED MAC ADDRESS FOR GOVEE H5051 WiFi Thermometer Hygrometer>
This seemed to cause the Umbrel networking stack to back off until sudo systemctl restart NetworkManager was run. This was very surprising because the Umbrel’s IP address had been reserved for almost a year. Unsure whether this represented a router bug or a bug on the Govee ESP32 WiFi thermometer/hygrometer, I rebooted the Govee device, but the issue reoccurred after that 10 minute Govee web sync. At that point, I used the Govee app to manually reconfigure the WiFi configuration for the Govee device, which finally caused it to choose a new IP address. It seems the IOT device is persisting an IP address and it was stomping my Umbrel.