User Tools

Site Tools


cluster:160

Warning: Undefined array key -1 in /usr/share/dokuwiki/inc/html.php on line 1458

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
cluster:160 [2017/04/05 08:33]
hmeij07 [OpenHPC page 4]
cluster:160 [2017/05/31 11:07] (current)
hmeij07 [OpenHPC page 4]
Line 5: Line 5:
  
 ** ib0 ** ** ib0 **
 +
 +Using Infiniband for MPI traffic involves somewhat more configurations. So from the ground up (v1.3 documentation) we start with installing packages needed on CHROOT.  Be sure to follow recipe and install these on SMS too.
 +
 +<code>
 +
 +# Add IB support and enable
 +yum -y --installroot=$CHROOT groupinstall "InfiniBand Support"
 +yum -y --installroot=$CHROOT install infinipath-psm
 +chroot $CHROOT systemctl enable rdma
 +# User environment
 +yum -y --installroot=$CHROOT install lmod-ohpc
 +
 +</code>
 +
 +Next we import a template file in which the IPADDR and NETMASK values of the ''ib0'' interface will be replaced with values from the database. Add to your deploy scripts lines like
 +
 +<code>
 +
 +wwsh file import /opt/ohpc/pub/examples/network/centos/ifcfg-ib0.ww
 +wwsh -y file set ifcfg-ib0.ww --path=/etc/sysconfig/network-scripts/ifcfg-ib0
 +
 +
 + wwsh node set $node --netdev=ib0 \
 + --ipaddr=$ipaddri \
 + --netmask=255.255.0.0  --network=255.255.0.0 -y
 +
 + wwsh provision set $node --fileadd ifcfg-ib0.ww -y
 +
 +</code>
 +
 +Reassemble the VNFS and reimage nodes. Now you can follow IpoIB instructions [[cluster:145|Infiniband]]
 +
 +Then add these lines to ~test/.bashrc file and resubmit job.mpi and you'll notice we now run MPI over Infiniband.
 +
 +<code>
 +
 +# User specific aliases and functions
 +module load gnu/5.4.0
 +module load openmpi/1.10.4
 +module load prun/1.1
 +module list
 +
 +# job.102.out
 +/opt/ohpc/pub/prun/1.1/prun
 +[prun] Master compute host = n29
 +[prun] Resource manager = slurm
 +[prun] Launch cmd = mpirun ./a.out
 +
 + Hello, world (8 procs total)
 +    --> Process #   0 of   8 is alive. -> n29.localdomain
 +    --> Process #   1 of   8 is alive. -> n29.localdomain
 +    --> Process #   2 of   8 is alive. -> n29.localdomain
 +    --> Process #   3 of   8 is alive. -> n29.localdomain
 +    --> Process #   4 of   8 is alive. -> n31.localdomain
 +    --> Process #   5 of   8 is alive. -> n31.localdomain
 +    --> Process #   6 of   8 is alive. -> n31.localdomain
 +    --> Process #   7 of   8 is alive. -> n31.localdomain
 +
 +</code>
  
  
Line 11: Line 70:
  
   * http://www.admin-magazine.com/HPC/Articles/Lmod-Alternative-Environment-Modules   * http://www.admin-magazine.com/HPC/Articles/Lmod-Alternative-Environment-Modules
 +  * http://lmod.readthedocs.io/en/latest/015_writing_modules.html
 +
 +
 +[[cluster:154|OpenHPC page 1]] - [[cluster:155|OpenHPC page 2]] - [[cluster:156|OpenHPC page 3]] - page 4
  
 + --- //[[hmeij@wesleyan.edu|Henk]] 2017/05/30 11:03//
  
-[[cluster:154|OpenHPC page 1]]  [[cluster:155|OpenHPC page 2]] [[cluster:156|OpenHPC page 3]] page 4+**POSTFIX NOTE**
  
 +In order to have users be able to send email from the jobs (from inside their jobs, like job progress reports), install mailx on the nodes. in ''/etc/ssmtp/ssmtp.conf'' define the relayhost (like sms-eth0-private) as the "mailhub". No other changes needed.
 + 
 \\ \\
 **[[cluster:0|Back]]** **[[cluster:0|Back]]**
cluster/160.1491395586.txt.gz ยท Last modified: 2017/04/05 08:33 by hmeij07