[Neolith-users] IPoIB and Gaussian problems fixed

[PärAndersson]

Hi,

During the last week the compute nodes on Neolith was upgraded from
CentOS 5.3 to 5.4. This caused a problem with IPoIB (IP over InfiniBand)
that was not discovered until yesterday. The problem was fixed yesterday
evening.

The only program that I know on Neolith that uses IPoIB is Gaussian,
both G03 and G09. Typicall errros caused by the problem is failure to
start, with error messages similar to this:

Rubpy1.out:ssh: connect to host ni545 port 22: No route to host
Ru_Ti_sep_optinmecn4.out:ssh: connect to host ni325 port 22: Connection timed out
Ru_Ti_sep_optinmecn4.out:ssh: connect to host ni401 port 22: Connection timed out

Gaussian jobs that was running yesterday while we did the diagnostics
and reconfiguration to fix the problem might also have crashed or
hanged, sorry about that.

Kind Regards,

Pär Andersson
NSC
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://www.nsc.liu.se/pipermail/neolith-users/attachments/20091111/2717631d/attachment.bin