azurehpc/examples/beegfs_local_ssd/readme.md

BeeGFS Cluster built with local ssd's

Visualisation: config.json

This will deploy a BeeGFS PFS using local SSD's (D16_v3), a headnode, an NFS server running on the headnode (User accounts shared home directories will be stored here), compute cluster and PBS will be deployed.

NOTE:

• MAKE SURE YOU HAVE FOLLOWED THE STEPS IN prerequisite before proceeding here

First initialise a new project. AZHPC provides the azhpc-init command that will help here. Running with the -s parameter will show all the variables that need to be set, e.g.

$ azhpc-init -c $azhpc_dir/examples/beegfs_local_ssd -d beegfs_local_ssd -s
Fri Jun 28 08:50:25 UTC 2019 : variables to set: "-v location=,resource_group="

The variables can be set with the -v option where variables are comma separated. The -d option is required and will create a new directory name for you.

azhpc-init -c $azhpc_dir/examples/beegfs_local_ssd -d beegfs_local_ssd -v location=westus2,resource_group=azhpc-cluster

Create the cluster

cd beegfs_local_ssd
azhpc-build

Allow ~15 minutes for deployment.

To check the status of the VMs run

azhpc-status

Connect to the headnode and check PBS and BeeGFS (it will be mounted at /beegfs)

$ azhpc-connect -u hpcuser headnode

Resources:

* Head node (headnode)
* Compute nodes (compute)
* BeeGFS
  * Management server (beegfsm)
  * Object storage servers and metadata servers(beegfssm)

> Note: The Hb nodes are used for the cluster.  To get best performance nodes with accelerated networking should be used.