Re: port-amd64/46833: NetBSD 6.0_BETA2 shutdowns under load

[Francois Tigeot <ftigeot%wolfpond.org@localhost>]

Hi,

On Sat, Aug 25, 2012 at 08:20:05AM +0000, Michael van Elst wrote:
> The following reply was made to PR port-amd64/46833; it has been noted by 
> GNATS.
> 
> From: mlelstv%serpens.de@localhost (Michael van Elst)
> To: gnats-bugs%netbsd.org@localhost
> Cc: 
> Subject: Re: port-amd64/46833: NetBSD 6.0_BETA2 shutdowns under load
> Date: Sat, 25 Aug 2012 08:16:28 +0000 (UTC)
> 
>  ftigeot%wolfpond.org@localhost (Francois Tigeot) writes:
>  
>  >                        Current  CritMax  WarnMax  WarnMin  CritMin  Unit
>  >               Temp6:    43.184   47.201   42.180    8.034    3.013 degC
>  >               Temp5:    48.205   47.201   42.180    8.034    3.013 degC
>  
>  Temp6 exceeds WarnMax
>  Temp5 exceeds CritMax
>  
>  powerd will shut down the machine when a sensor goes 'critical'.
>  
>  Maybe the sensors do not read out correctly or NetBSD assumes a wrong
>  conversion function. Can you verify that your your server really isn't
>  running too hot? Often you can see the sensor readouts in BIOS or
>  through IPMI.

Fan speeds vary automatically according to temperature; they where far
from running at fullspeed when powerd decided to shut down the system,
and believe me they're *loud*.
It's impossible to miss the sound when the machine gets hot and really
starts pumping air.

This Xeon box has been running without any issue under far heavier loads
with Linux and other *BSD systems. Never got a complaint, not even a
beep or a warning led.

BIOS setup doesn't show anything wrt environment sensors and I haven't
found a working ipmi client in pkgsrc yet.

>  N.B. the thresholds look like being tuned for operation in a real
>  air-conditioned and cooled computer center. Shutting down the machine
>  when the cooling fails seems to be reasonable to me.

The server is sitting on a test bench and not a regular machine room but
this shouldn't make too much difference. Ambient temperature is 24°C.
What are these Temp sensors supposed to monitor anyway ? Some report minus
30°C values, which seems about right for Siberia and not machine rooms...

-- 
Francois Tigeot