Provisioning Services

Let’s provision our services

  4 minute read  

Infrastructure is finally all set. We only need to finalize the setup of a few services (from Kahuna’s perspective) and we’re done.

Storage Pool

Let’s update your TF configuration to simply declare the following:

locals {
  ceph_port          = 3300

  eu-west {
    pools = {
      "eu-west-ssd" = {
        desc    = "SSD"
        secret  = "YOUR_CEPH_FSID",
        cost    = 200.0,
        type    = "rbd",
        pool    = "rbd",
        address = "ceph",
        default = true,
        agents = [
          "kaktus-eu-west-a-1",
          "kaktus-eu-west-a-2",
          "kaktus-eu-west-a-3",
        ]
      },
    }
  }
}

resource "kowabunga_storage_pool" "eu-west" {
  for_each = local.eu-west.pools
  region   = kowabunga_region.eu-west.id
  name     = each.key
  desc     = "${local.eu-west.desc} - ${each.value.desc}"
  pool     = each.value.pool
  address  = each.value.address
  port     = try(each.value.port, local.ceph_port)
  secret   = try(each.value.secret, "")
  price    = try(each.value.cost, null)
  currency = local.currency
  default  = try(each.value.default, false)
  agents   = [for agent in try(each.value.agents, []) : kowabunga_agent.eu-west[agent].id]
}

What we’re doing here is instructing Kahuna that there’s a Ceph storage pool that can be used to provision RBD images. It will connect to ceph DNS record on port 3300, and use one of the 3 agents defined to connect to pool rbd. It’ll also arbitrary (as we did for Katkus instances) set the global storage pool price to 200 EUR / month, so virtual resource cost computing can happen.

And apply:

$ kobra tf apply

NFS Storage

Now if you previously created an NFS endpoint want to expose it through Kylo services, you’ll also need to setup the following TF resources:

locals {
  ganesha_port       = 54934

  eu-west {
      nfs = {
      "eu-west-nfs" = {
        desc     = "NFS Storage Volume",
        endpoint = "ceph.storage.eu-west.acme.local",
        fs       = "nfs",
        backends = [
          "10.50.102.11",
          "10.50.102.12",
          "10.50.102.13",
        ],
        default = true,
      }
    }
  }

}

resource "kowabunga_storage_nfs" "eu-west" {
  for_each = local.eu-west.nfs
  region   = kowabunga_region.eu-west.id
  name     = each.key
  desc     = "${local.eu-west.desc} - ${each.value.desc}"
  endpoint = each.value.endpoint
  fs       = each.value.fs
  backends = each.value.backends
  port     = try(each.value.port, local.ganesha_port)
  default  = try(each.value.default, false)
}

In a very same way, this simply instructs Kahuna how to access NFS resources and provide Kylo services. you must ensure that endpoint and backends values map to your local storage domain and associated Kaktus instances. They’ll be used further by Kylo instances to create NFS shares over Ceph.

And again, apply:

$ kobra tf apply

OS Image Templates

And finally, let’s declare OS image templates. Without those, you won’t be able to spin up any kind of Kompute virtual machines instances after all. Image templates must be ready-to-boot, cloud-init compatible and either in QCOW2 (smaller to download, prefered) or RAW format.

Up to you to use pre-built community images or host your own custom one on a public HTTP server.

locals {
  # WARNING: these must can be in either QCOW2 (recommended) or RAW format
  # Example usage for conversion, if needed:
  # $ qemu-img convert -f qcow2 -O raw ubuntu-22.04-server-cloudimg-amd64.img ubuntu-22.04-server-cloudimg-amd64.raw
  templates = {
    "ubuntu-cloudimg-generic-24.04" = {
      desc    = "Ubuntu 24.04 (Noble)",
      source  = "https://cloud-images.ubuntu.com/noble/20250805/noble-server-cloudimg-amd64.img"
      default = true
    }
  }
}

resource "kowabunga_template" "eu-west" {
  for_each = local.templates
  pool     = kowabunga_storage_pool.eu-brezel["eu-west-ssd"].id
  name     = each.key
  desc     = each.value.desc
  os       = try(each.value.os, "linux")
  source   = each.value.source
  default  = try(each.value.default, false)
}

At creation, declared images will be download by one of the Kaktus agent and stored into Ceph cluster. After that, one can simply reference them by their name when creating Kompute instances.

Congratulations, you’re now done with administration tasks and infrastructure provisionning. You now have a fully working Kowabunga setup, ready to be consumed by end users.

Let’s then provision our first project !

Last modified September 11, 2025: docs: add services provisioning (544ba82)