Some months ago, the first virtual console on my home Linux box became “stuck” and wouldn’t respond. It wasn’t too much of an inconvenience so I didn’t bother much about it. A couple of days ago, I noticed that neither the CD drive nor the DVD burner was responding. Ok, I figured, I’ll have to schedule a reboot to fix that. Then last night’s backup failed when internal drive that my rack mount server backs up to suddenly decided it was read-only. I unmounted it, remounted it and it was ok. I restarted the backup and everything appeared to be running fine. But then a few hours later I noticed the CPU was pegged and two rsync processes were going crazy. Evidently the backup is screwing up again. I killed the backup, and decided I reboot it again as soon as I got home.
I was reluctant to take this step for a couple of reasons – first, it has 159 days of uptime. Second, it has always been a bitch to reboot. In the old days, it just wouldn’t recognize all the drives every time. Putting in a really really powerful power supply mostly cured that problem, but now it shows the BIOS screen showing all the disks, but then it just sits there not going any further. I can usually fix it with the reset button or power cycle.
I got home and found my office door had been closed, and so it was baking hot in the room with my computer. And when I rebooted, it wouldn’t come up. I’ve tried many of my old tricks, and it’s still not booting. I’m going to let it cool down, but if that doesn’t work, I guess I’ll be shopping for another server. Problem is, I don’t *need* a new server. The current one has plenty of processing power, memory, and most importantly, lots of disk space. I suppose I’d like something that’s more energy efficient, but otherwise I don’t know what to shop for in these days.
Sounds like a hard disk failure to me. You might might might be lucky when the drive cools down. But replace it ASAP.
Indeed, after a failure like this run “badblocks” on every single disk and see if any others have failures.
If you’re looking for something lower power, the little VIA C3/C7 boxes are pretty frugal (my current server setup is a pair of them; one is the main server, the other just holds off-machine backups. Total consumption is just shy of 2.4kw/day. I’m planning on replacing them with a single machine that holds a stack of 2.5″ drives, and I think that will drop the power consumption to about 700 w/day.)
The more-reliable server scheme I’m implementing is to put the root filesystem onto a flash disk, then wedge in a pack of immutable symlinks to put transient volatile things off on one of the data disks. I am planning to mount root r/o (toggled to r/w when I have to edit /etc/passwd and the like) so that ifwhen one of the data disks explodes I can reboot, hand-mount one of the other disks, and limp along until I can purchase a replacement disk.
Regarding energy efficiency, don’t count out the power supply. When my home server’s original Antec 300w power supply went blooey after years of 24/7 operation, I picked up an OCZ 700-watt unit – and noticed that the server’s load on my UPS had dropped 50%.
I picked up another big reduction in power consumption when I switched from my old 17″ CRT to a 22″ widescreen LCD.
I wouldn’t mind going to a C3 or similar low-power setup, if I could find one that had lots of SATA ports.