[Neolith-users] service-stop, removing all queued jobs 2008-07-01

[Pär Andersson]

Dear Neolith Users,

Short version:

The service stop started 2008-07-01 08:00 and ended 2008-07-02 01:00.

We upgraded lots of software, if you are not interested in the details then 
please skip the rest of this mail. Simply report any problems to 
support at nsc.liu.se as usual, especially problems that was not present before 
the service stop.

Long version:

I promised in my previous message that I would send out more detailed 
information about the upgrades we performed, here is a list.

SLURM 1.2.27 to 1.3.4
The update from 1.2 to 1.3 includes several changes and I mentioned a few in 
my initial announcement about this service stop. A complete list is in the 
release notes, and also the man pages for sbatch and other slurm commands is 
updated.
/usr/share/doc/slurm-1.3.4/RELEASE_NOTES

CentOS 5.1 to 5.2
CentOS is the Linux distribution used on Neolith. This upgraded the Linux 
kernel and several other packages, release notes is availabele online.
http://wiki.centos.org/Manuals/ReleaseNotes/CentOS5.2

OFED 1.2.5 to 1.3.1
The InfiniBand software stack including libraries and kernel modules. If you 
have compiled programs that use lower level infiniband libraries like 
libibverbs then this might affect your programs.
MPI-programs should continue to work as usual, we have tested that the 
installed versions of ScaliMPI and OpenMPI still works as expected on this 
stack.

Scali MPI 3.13.5 to 3.13.7
Minor bugfixes.

Other changes:
The amount of swap space on compute nodes was increased from 2 to 34 GiB. This 
was necessary to work around bugs in a few popular applications like Gaussian 
and VASP. The disk space was taken from /scratch/local so that file system is 
now 190 GiB on every node.


Best regards,

Pär Andersson
NSC