db16487214 | ||
---|---|---|
.. | ||
install_prolb.sh | ||
prolb.sh | ||
prolb_beeond.sh | ||
readme.md | ||
runtime_prolb.sh | ||
tune_prolb.sh |
readme.md
Install and run PROLB
Note : This version has been tested on HC44rs and HB60rs. When running on other SKUs, please update the
prolb.sh
script to adapt the memory per core to be used.
Prerequisites
Cluster is built with the desired configuration for networking, storage, compute etc. You can see the tutorial or examples folder in this repo for how to set this up.
Dependencies for binary version:
- HPCX with C++ bindings
- OpenMPI shared libraries symbolic links (see
runtime_prolb.sh
)
Installation
First upload the install packages and cases in your favourite blob storage account.
NOTE: Provide the
INSTALL_TAR
,TAR_SAS_URL
,LICENSE_PORT_IP
andAPP_VERSION
as parameters to the $azhpc_dir/apps/prolb/install_prolb.sh
Then copy the apps directory to the cluster. The azhpc-scp
can be used to do this:
azhpc-scp -r $azhpc_dir/apps hpcuser@headnode:.
Alternatively you can checkout the azurehpc repository but you will need to update the paths according to where you put it.
Then run the installer
azhpc-run -u hpcuser apps/prolb/install_prolb.sh
Note: This will install into
/apps
.
Finally, if the runtime_prolb.sh
is not part of your compute node installation, run that script on all compute nodes as follows
azhpc-run -n compute -u hpcuser apps/prolb/runtime_prolb.sh
Connect to the headnode
azhpc-connect -u hpcuser headnode
Running
Copy the case file to /data/prolb
Now, for example on 8 HC44 nodes, you can run as follows:
CASE=mycasename
case_dir=/data/prolb/working
mkdir -p $case_dir
qsub -f -k oe -j oe -l select=8:ncpus=44:mpiprocs=44,place=scatter:excl -N prolb $azhpc_dir/apps/prolb/prolb $CASE $case_dir [version]