FreeBSD shutting down unexpectedly
soralx at cydem.org
soralx at cydem.org
Sun Mar 5 23:41:21 PST 2006
> >> I'm having a problem with my FreeBSD server shutting down for no
> >> reason. This started happening recently (within the last month) so
> Or, sorry to say, you may be yet another victim of the "bad capacitors"
> that have plagued many motherboards, even from quality makers, during
> the past few years. Do a google search, or try this site that google
> just showed me, for what to look for: http://www.badcaps.net/
or even better: http://en.wikipedia.org/wiki/Capacitor_plague
capacitors recently confirmed (by me personally) to be bad (found in
Thermaltake PSU and a Gigabyte mainboard): G-Luxon, Choyo, Tayeh.
As I mentioned before, bad/failing/questionable power supplies is not
something I advise to use in any systemm, if the system and it's data
are of any value.
Failing capacitor may (though seldom) cause a MOSFET failure, which
may possibly be a Source-Drain short circuit. This means that voltage
way too high will be applied to components before the transistor lets
out some magic smoke, or (current limiting)|(overcurrent shutoff) kicks in.
There is another way in which a failing PSU may destroy your data. My old
AT server became unstable a month ago (halts, shuts off, etc). So I
measured it's power supply's output: +5V was at ~12.5Vdc level, and +12V
was as high as 20Vdc. The server couldn't boot, becase somehow the boot
record on HDD became corrupted (the rest of data was ok). Turns out
the PSU's fan stopped, so it worked at high temperatures for months,
which immensely shortened capacitors' life (for electrolytics, the service
life roughly halves for every 10*C temperature increase), ending
in parametric failure.
As you can see, it's always a good idea to replace all power capacitors
on mainboard and power supply with quality ones before use (even if new).
It cost me ~$30 to get all capacitors > 470uF for a socket478 mainboard
and 2 ATX PSUs (I got high-quality, low ESR Nichicon UPL-series caps).
So, ensure good cooling, test components before use (especially the PSU),
watch for symptoms of hardware failure, avoid poor quality parts, and don't
forget about maintenance; these simple rules become automatic very quickly.
BTW, by 'good quality parts' I mean not just some brand parts, but devices
made for optimum performance/efficiency, not for optimum profit to the brand.
Since such components are nearly impossible to find nowadays (for reasonable
price), it's a good idea to buy some more-or-less good (well engineered) parts
(such as Thermaltake PSU, for example -- not some no-name genius of a Chinese
engineer), throw out low quality components (such as power capacitors and
fans), and install good quality parts. This will _greatly_ increase stability,
service life, and sometimes efficiency.
If more people refused to buy crappy devices (many don't even imagine how
bad, horribly bad, most non-server grade products are), we wouldn't have
manufacturers who try to cut cost by deliberately utilizing
(flawed)|(low quality) components. Would you consider spending U$15 (cost
of good capacitors) more on a $320 mainboard and get a reliable, stable
(better overclocking, too) board that will serve you well for many years
(istead of just 0.5 - 3y max)? Would you try to save a few bucks on a $100
power supply to get an unstable, high-ripple PSU which may kill
your HDD one day?
This applies not only to computer, of course.
> 5 years ago, we bought 25 APC and more than 7 had some problems. Some of
> them turned machine off, while monitor meter did not show losing power.
Possibly, the UPSes had dead/discharged batteries, so when doing a self-test,
would lose power to the load.
> Do you know which ports specifically can show running temps?
'ports/sysutils/mbmon' works for me
More information about the freebsd-hardware