azurehpc/experimental/cc_glusterfs2

lding the infrastructure Here we will explain how to deploy a full GlusterFS system with a VNET, JUMPBOX, CYCLESERVER and Gluster by using building blocks. These blocks are stored into the experimental/blocks directory.

Step 1 - install azhpc

after cloning azhpc, source the install.sh script

Step 2 - Initialize the configuration files

Create a working directory from where you will do the deployment and configuration update. Don't work directly from the cloned repo.

$ mkdir cluster
$ cd cluster

Then copy the init.sh,variables.json, and scripts directory from examples/cc_glusterfs2 to your working directory.

$ cp $azhpc_dir/examples/cc_glusterfs2/init.sh .
$ cp $azhpc_dir/examples/cc_glusterfs2/variables.json .
$ cp -r $azhpc_dir/examples/cc_glusterfs2/scripts .

Edit the variables.json to match your environment. Leave the projectstore empty as it will be filled up with a random value by the init script. An existing keyvault should be referenced as it won't be created for you.

{
    "resource_group": "my resource group",
    "location": "location",
    "key_vault": "my key vault",
    "projectstore": ""
  }

Run the init.sh script which will copy all the config files of the building blocks and initialize the variables by using the variables.json updated above.

$ ./init.sh

Step 2 - Build the system

$ azhpc-build -c vnet.json
$ azhpc-build --no-vnet -c jumpbox.json
$ azhpc-build --no-vnet -c cycle-prereqs-managed-identity.json
$ azhpc-build --no-vnet -c cycle-install-server-managed-identity.json

Step 3 - Deploy the Cycle CLI

Deploy the Cycle CLI locally and on the jumpbox

$ azhpc-build --no-vnet -c cycle-cli-local.json
$ azhpc-build --no-vnet -c cycle-cli-jumpbox.json

Step 4 - Now deploy the GlusterFS cluster

$ azhpc-build --no-vnet -c gluster-cluster.json

Step 5 - Create the PBS cluster in CycleCloud

$ azhpc build -c pbscycle.json --no-vnet

Step 6 - Connect to CycleServer UI

Retrieve the CycleServer DNS name by connecting with azhpc

$ azhpc-connect -c cycle-install-server-managed-identity.json cycleserver
[2020-06-10 08:28:04] logging directly into cycleserver559036.westeurope.cloudapp.azure.com
$ [hpcadmin@cycleserver ~]$ exit

Retrieve the Cycle admin password from the logs

$ grep password azhpc_install_cycle-install-server-managed-identity/install/*.log

Connect to the Cycle UI with hpcadmin user and the password retrieved above. Check that you have a pbscycle cluster ready and start it

AzureHPC BeeGFS and CycleCloud Integration