A/C failure - shunt trip success.

[Christopher Hogue]

Hi folks,

Just wanted to say we had an experience last friday when the A/C unit
went down in our small cluster room with 108 rackmount dual CPU nodes.  

We had installed a temperature sensor on a shunt trip switch to the
power feed in the room when we set it up earlier this year.  With the
A/C off, the temperature quickly got up to 45C (~110F), while we were in
the middle of a 100-million protein structure run.  This temperature is
the set-point for the shunt trip - which then killed power to the
switches and cluster nodes.  With the A/C back up, (4 days later -
replacing a roof compressor), all nodes and switches recovered.

So this note, just to indicate, if you are building one of these
contraptions and putting a big cluster in, have the electricians & A/C
folks put in a shunt trip switch and temp sensor on the incoming power
feed.

Now if only I could find a way to put my cluster outside in the snow in
the winter...

Christopher Hogue
Samuel Lunenfeld Research Institute, Toronto
http://bioinfo.mshri.on.ca