nomad

Installation

Requirements

allow_privileged must be enabled in Nomad config.

Components

To make vultr csi work you need to run two components:

• csi-controller
• csi-node

csi-controller and csi-node uses the same binary, but different schedulers. csi-controller should be run as service and csi-node should be run as system (run on every host).

See more at Nomad documentation on CSI here.

You will need to run a separate deployment for each Vultr region.

API key

In order for the csi to work properly, you will need to provide an API key to csi-controller.

To obtain a API key, please visit API settings.

API key can be passed to csi-controller securely via Vault integration.

Example snippet:

      task "csi-controller" {
        driver = "docker"

        vault {
          policies = ["vultr"]
        }

        config {
          image = "vultr/vultr-csi:v0.5.0"

          args = [
            "-endpoint=unix:///csi/csi.sock",
            "-token=${VULTR_API_KEY}",
          ]
        }

        template {
          data = <<-EOF
          VULTR_API_KEY={{ with secret "secret/vultr/csi" }}{{ .Data.data.key }}{{ end }}
          EOF

          destination = "secrets/api.env"
          change_mode = "restart"
          env         = true
        }

Run Vultr CSI

Adapt and run example jobs definitions:

• controller
• node

In Nomad UI in Storage tab make sure plugin is healthy.

Create and register example volume

Nomad will not create volume on demand. You need to create a volume yourself either by hand or with Terraform and then register it in Nomad, again by hand with nomad volume create command or with Terraform.

Adapt and use this config to test.

Validate

To validate run example job and the following commands:

nomad exec -job example touch /data/example
nomad stop -purge example
nomad system gc
nomad run example.nomad.hcl
nomad exec -job example ls -alh /data

Examples

Examples of Nomad jobs and Terraform configs can be found here.