Hi misc@, I have two Supermicro X9SBAA servers in a rack. These are Atom S1260 systems with 8GB of ECC memory.
I noticed that one of them was beeping while flashing an alarm light. There were no symptoms otherwise, but obviously this is concerning. It was running 7.8 and I ended up upgrading to 7.9 trying to diagnose it. There was no impact. Triggering it was very random. It doesn't seem to be thermally related. The temperature, according to sysctl, is maybe 3C higher than the (basically) identical server next to it. I can peg all four threads and it won't trigger it. It will start/stop fairly randomly, but I know at one point during fw_get it started as soon as ftp was going. It seems to be potentially network related, but downing the links and/or unplugging ethernet cables did not seem to reliably stop the alarm. I tried with spread spectrum on and off in the BIOS. PSU voltages looked normal. I eventually lowered the clock with setperf=1 and did not hear the alarm after that, but I'm not confident if it's going off or not. I wish I knew if it was going off because I don't want to annoy anyone else in the datacenter. It's a very loud alarm. Are there any interfaces exposed to OpenBSD (and maybe accessible through sysctl) that can tell me if an alarm is going off, or that might help me diagnose it? The system has never crashed or done anything weird otherwise. Maybe some way to see if there are ECC errors that are being fixed on the fly? Very perplexing. I would appreciate any tips you can suggest. Please include me in replies as this account is not on misc@. Thank you! -Slow Servers
