General operational results are here.

Tests of failure scenarios are here.

• satisfy prerequisites
• build the cloud infrastructure (with terraform)
• configure the servers (with ansible)
• setup grafana dashboard to monitor
• create test files on nfs-server
• run tests on nfs-client

The core requirements are:

• Terraform version v1.2.3
• Ansible version 2.12.5
• Wireguard client tools to generate keys

For AWS:

• follow the getting started guide here

For GCP:

• follow the getting started guide here

From the scripts directory, run this command to generate the keys:

./make-keys.sh

It will produce an output like this which you can copy and paste into the terraform vars file.

NOTE: these aren't valid keys.

server keys
    vpn_private_key = "YMAlCdUgLpVfRwXVI6RXR937YkHYAC2hrtAnheBWQ="
    vpn_public_key  = "GJXpV6tVJBR55CN9Hdv3BI6CE2wRfxc5mvTZ6aTlE="
client keys
    vpn_private_key = "kGNAGM61jLOoi0CDyegii6BrQMzsc0XBZNYIjBbVo="
    vpn_public_key  = "3zrNQ/MHP99CnLK+DFSceiysM/cM0i3F76pSl8fkw="

From within the terraform directory (terraform-aws or terraform-gcp), copy the file terraform.tfvars.examples to terraform.tfvars and customize to your environment.

For both environments:

• update the server-site variable:
  • set the region and zone to where you want the server to run
  • update the vpn_private_key and vpn_public_key values with the server keys generated by make-keys.sh
• update the client-site variable:
  • set the region and zone to where you want the client to run (needs to be different to the server)
  • update the vpn_private_key and vpn_public_key values with the client keys generated by make-keys.sh

For AWS:

• update the aws-profile variable if you're using a profile other than default
• make sure that the ami_for_region map has entries for the regions you are using
  • this should be for the Ubuntu 2022 LTS 64bit

For GCP:

• set the project_id
• set the credentials_file variable to point to the location of your GCP credentials file
• set your username within GCP

From within the terraform directory, run these commands:

terraform init
terraform apply

From within the terraform directory, run the ansible wrapper script:

./local/ansible/run-ansible

Setup port forwarding from your local machine to the metrics server, by running the below from within the terraform directory:

ssh -F local/ssh.cfg metrics-ports

Browse to http://localhost:3000 and you will be connected to the grafana instance running in the client VPC.

Create a datasource to the prometheus instance running on the same machine: http://localhost:9090.

Create a dashboard to monitor network and disk IO on nfs-server, nfs-cache and nfs-client. See the graphs in the test performance document for some examples.

From within the terraform directory, log into the nfs-server:

ssh -F local/ssh.cfg nfs-server

The helper tools were built as part of the ansible run and are available in the bin directory. Create the test files using:

mktestfiles -n 6 /shows/test 60

This uses 6 goroutines to create files in the /shows/test directory and fill it to 60%.

The files are named after the sha256 hash of their content. This is used by the read routines to verify the contents of the files when testing.

From within the terraform directory, log into the nfs-client:

ssh -F local/ssh.cfg nfs-client

To read all the files in the nfs mount point, run:

fsreadall -n 4 /shows/test

This will read all the files through the nfs-cache and save the contents in the local cache directory. It also verifies the sha256 content has of the file matches the file name.

Run it a second time and it will read the files from the cache, not the nfs-server. This will run in much less time than the first run.

Other tests you can run include:

fsreadrandom -n 6 /shows/test 30m

This will randomly read ranges from within randomly selected files within the test area for 30m. You can specify the time to run for as documented here.

fsreadwrite -n 6 /shows/test 30m

This randomly selects one of three actions to perform:

• read the full range of a file (and verify the sha256 content hash)
• read a random section from within the file
• create a new file on the server and delete the original