Eskimo Crash's

[Robert Dinse <nanook@eskimo.com>]

     Eskimo has crashed three times in as many days with memory errors, but the
errors have been in different SIMMS, though all of them were in the same bank
of memory in the same corner of a memory expansion board. 

     Tonight around midnight I took Eskimo down, disassembled it, and vacuumed
everything out.  The air intake screen was totally clogged so I suspect that
was a lot of the memory problems (heat from lack of any air flow).  There are
nine fans in this machine to keep it cool but that does no good if the intake
is clogged. 

     This is something that doesn't get done often because the only way to do
it is to take Eskimo off-line.  The physical construction of the machine
locates drive bays over the air intake, so one can't get to it without taking
it down and removing the drive bays.  And in this case I also pulled the cards
out of the VME cage and vacuumed them off, vacuumed out the cage, and the fans
and anything else I could gain access to with it disassembled since it had been
a long time since it was last cleaned out. 

     I also removed a dead drive which was still connected to the system bus
internally, that will eliminate 27 watts of heat generation since it was an old
style full height 5-1/4 inch 9gb drive (one of the old Micropolis drives that
had long since died but still was powered up and on the bus). 

     Eskimo is a particularly critical server since it houses all the user
files including web pages, and authentication server for the dial-up
connections, so when it goes down you can't dial-in on the low speed numbers,
and www can not serve pages since this is the machine upon which the files
reside.