> Tell me a little bit more if you have time, especially about the two
> network cards?!
Peter
following several questions about these "unexplained" and repeated crashes
of O2's for O2 controlling Bruker spectrometers, here is some info on
the problem I had and how I solved it.
The O2 controlling an AVANCE spectrometer (or a remote station with the
exact same configuration but not controlling a spectrometer, which
ruled out problems originating from the spectrometer) was crashing at
intervals from few hours to a few days.
Attempts to locate the problem with software were unsucessful but the
conclusions were that it resulted from an Irix 6.x bug appearing only
when two network cards were active with a larger amount of RAM (>=
256Mb). The problem requires *all* of the following to be simultaneously
true:
* mixed RAM sources (typically SGI + third party)
* more than 256 Mb RAM
* two activated ethernet cards
* Irix 6.x
As to a solution, removing any of these solves the problem.
A first solution is therefore to work at less than 256Mb of RAM. Not really
practical.
A second solution is to keep only one type of RAM within the box
(all SGI or all third party). It is surely more sound and cheaper to
buy a full set of RAMs (from a third party) and toss SGI's ones
than to loose a couple of experiments. My favorite solution.
A third solution is to get rid of the second ethernet card, which
leaves your O2 "offline". This can be done by hardware removal,
or by commenting the proper lines in /etc/config/netif.options.
Works for assuring proper acquisition on the short term,
but it is hardly practical.
By the way, SGI does not care: who on Earth has two ethernet cards
apart from a few Bruker customers ?? I have installed all versions of
irix patches existing for different versions of Irix without improvement...
All of this under SGI's advices. Which ended up saying it's the third
party's
RAM problem...
Bruker, although very helpful in localizing the source of the problem to the
external RAM
I had added, did not offer other alternatives (once I had explained
the sources of the problem to them...).
Good luck
Sebastien
Dr. Sebastien VINCENT
Research Scientist
Nestlé Research Center
PO Box 44, CH-1000 Lausanne 26
Phone: + 41 21 785 9165
Fax: + 41 21 785 8549
e-mail: sebastien.vincent_at_rdls.nestle.com
Received on Tue Nov 20 2001 - 09:34:16 MST