I have what I guess is a tiny, NUC style PC, with Intel i3, 8GB RAM and a 128GB SSD.
It has two Ethernet ports and I initially bought it with the intention of using it as a router, running something like Untangle. I never got round to setting that up, so thought I’d re-purpose it as a HA box, and compare its performance to the pi4 I’m currently using (InfluxDB was using a lot of memory on it). I flashed the SSD with the NUC install via a USB adaptor. It booted fine and I started configuring it, but after an hour or two, it became unavailable, both via http and ssh. Luckily I had it connected to a monitor, showing these messages.
I’m not used to interpreting such errors, but I’m guessing the important stuff is the the “page_fault” line? Does this suggest a memory issue? It had previously been working OK with various other (Linux based) OS’s on it, but admittedly they’d not been used in anger. Can anyone suggest how to troubleshoot this? Maybe I should load an OS (most recently had Ubuntu server on it until I realised I could no longer install HA on top of this) and run some diagnostics?
Thanks very much, this has been very helpful. I downloaded the tool and started it running .It was looking like it was going to be a clean bill of health, but as soon as it got to Test 7 “moving inversions, 32bit pattern” the errors started appearing. 248 errors in total on the step, none on any of the others for some reason…
I’d ordered a new stick of memory. While I awaited it’s arrival I mentioned what had happened to a friend and he asked “did you check the memory settings in the BIOS?”. I’d given no thought to that, but had a look, and the memory voltage was set to “auto”. He said that’d probably be fine, but why not just check what voltage the memory was supposed to be run at? The stick said 1.35V. I wasn’t hopeful but set it to manual and 1.35V and ran the specific test again that had all the errors.
Not a single error. I ran the whole MemTest suite again, for the full 4 sweeps. Not a single error.
The new memory has arrived. Another single 8GB stick, luckily also designed for 1.35V. I now have both of them in and am running another full check with MemTest. It’s been running for a couple of hours and no errors so far. Once that’s finished I’ll boot up HA again and see how I get on. I can’t imagine anything on my HA install is going to need all that memory!