Strange system lockups - kernel saying disk error
Michael Powell
nightrecon at hotmail.com
Sun Jun 5 20:49:00 UTC 2011
perryh at pluto.rain.com wrote:
[snip]
>
> Power supplies do fail occasionally, and not always in obvious
> ways such as failing to turn on at all. The output voltages may
> be a little too high or too low, or they may be correct but with
> excessive ripple or electrical noise; or the supply may be just
> fine until a disk draws a current spike to move the arm rapidly.
I've seen a fair number or power supplies degrade somewhere around the 5
year mark. Simple voltage checks with a VOM and its accuracy will usually
still show the voltages as being correct. To see the ripple you'll need an
oscilloscope. Excessive ripple can make a PC appear to have all kinds of
intermittent hardware failures with little or no rhyme or reason. A degraded
power supply will show large variations in ripple based on load. The largest
load from hard drives is when they are first spinning up. Servers are
commonly configured with the ability to spin up drives one at a time with a
short delay in between. You won't usually find this on a desktop.
Generally, this situation will develop more often on an old machine that had
a 'barely enough' capacity power supply when new. Add 3 more hard drives,
bigger video, etc and it was still just inside the envelope until enough
time went by and the power supply got old. Since the most amps pulled by the
hard drives occurs on power up you will see the ripple on a 'scope look
really ugly while this happens. The unseen danger here is that bits on the
drive(s) can get scrambled until things settle down. You will know this
happens when stuff goes wrong and fsck is needed to get the file system
clean, and after cleaning and working again will do the same thing again at
some future reboot.
Easiest way to look at this without a 'scope is to simply substitute a known
good PSU of sufficient rating from a machine with no troubles. If all the
random nonsense suddenly stops, you'll know. This is easiest for folks these
days as those without an analog electronics background are unlikely to have
an oscilloscope laying around.
> It might be worth checking the fan mounted on the CPU heatsink if
> there is one, and the fan in the power supply (which ventilates the
> case as well as the power supply itself).
Aside from the fans themselves, dust buildup plugs heat sinks eventually
drastically reducing their ability to get rid of heat. When you get to this
stage blowing them out with canned air can work wonders. My 2 servers at
home sit on the floor and need this about once a year.
-Mike
More information about the freebsd-questions
mailing list