Hello Marcus,
> I'm running on a pi 3b (I think). I put up with intermittent sd card failures by having a spare lying around and rebuilding the system.
Heh... I do need to get the docker packaging finalized to make that
process simpler.
The series of cards that worked the best for me historically is
Samsung EVO Select - sure, I had a couple die on me, but that's out of
a sample of 20+, and those that died, I think, Home Assistant ran into
the ground by incessant writes. These cards survived being used in a
dashcam with 70C (160F) surface temperature, it's quite possible that
the dashcam itself was even hotter.
> Usually happens at an inconvenient moment. I send data to an oracle cloud instance (free tier)
I keep all the configurations version controlled with a GitHub private
repo being the authoritative source. One of the improvements I'm
considering in the long run is - have DZ talk to source control
itself, to make the task more seamless. Given their recent
improvements in account management, that'll be sufficiently secure.
> and use that to manage public network access via wireguard as I'm cgnat'ed. I have considered running proxmox on a more capable machine and using it for other tasks, but computers inevitably fail and it's nice to have things in modular bits that when something fails other things still work
My reasoning exactly. I still have a few spare Pis of different types
I can just repurpose in a blink of an eye - but I'm yet to have a Pi
hardware failure, usually, it's the SD card that goes.
> My sensor network is 1-wire and I use various esp 32's for other tasks, although I find them a little flakey (need resetting now and then)
This is my worst problem with ESPHome so far:
https://github.com/esphome/issues/issues/3415 - still unresolved. I
don't know if it is the ESP firmware, ESPHome, or the wireless
hardware problem, but it makes running actuators for expensive
hardware over WiFi a dubious proposition.
I wonder if anyone else experienced this problem? Other than that,
it's pretty stable. The only other thing that puzzles me is how could
DIGITAL I2C sensors report outrageous samples of 200C and above at
times - I've seen that with analog XBee sensors and it's
understandable, but digital???
--vt