• This is our actual infrastructure.
• Changes made here can be damaging.
• Be careful.
• Terraform catches some mistakes but not all

This infrastructure repo will run in Jenkins on cron, to ensure our infrastructure (Cloud VMs etc ...) matches exactly what it should.

All changes should go through pull requests and never directly to the main branch, we will have the Jenkins bot comment on the PR with the terraform plan output, allowing the admin to decide whether or not to merge it.

We're using this to manager every cloud resource. If it is something you would do with the OpenStack API or UI, do not do it. Instead, use this repository for it.

Our DNS provider is Amazon AWS/Route53 since they have a mostly reliable service and a nice API.

All changes come in through PRs and are applied automatically by our build server. Sometimes jobs error there but it is normal.

All important variables like flavour names, AWS Route53 zones, groups of security groups for default things like webservices, etc. go in the variables file.

variable "vgcn_image" {
  default = "vggp-v31-j74-edc5aa3dc22c-master"
}

defines a variable and then you can use this with var.vgcn_image

All instances are stored in files named instance_<class>_<name>.tf. Class can be:

• core: needed by usegalaxy.eu to run
• dedicated: dedicated to some people or a project
• extra: all other cases Their structure is not too complex:

#        type of resource                 resource name
resource "openstack_compute_instance_v2" "apollo-usegalaxy" {
  # Server name in the OpenStack api. becomes the internal hostname with
  # .novalocal appended
  name            = "apollo.usegalaxy.eu"

  # We have several variables for you to choose from in the vars.tf file.
  image_name      = "var.centos_image"

  flavor_name     = "m1.large"
  key_pair        = "cloud2"

  # You can define this as a list or use the var.sg_webservice for all of the
  # default security groups required for a webservice (egress, ufr ssh, public
  # ICMP, public http(s))
  security_groups = "var.sg_webservice"

  # Here we attach two networks:
  network {
    name = "bioinf"
  }

  network {
    name = "public"
  }
}

# Here we define a DNS record for this VM
resource "aws_route53_record" "apollo-usegalaxy" {
  # The zone needs to be correct for the TLD you want.
  zone_id = "var.zone_usegalaxy_eu"
  # The actual record
  name    = "apollo.usegalaxy.eu"
  type    = "A"
  ttl     = "7200"

  # Here we use a computed value from the first resource, format is from above:
  # "type of resource"."resource name".access_ip_v4
  records = ["openstack_compute_instance_v2.apollo-usegalaxy.access_ip_v4"]
}

When you run make all terraform will sync, and if VMs need to be destroyed and re-created, they will and DNS records will update appropriately.

All infra IPs (without going to openstack.)

$ ./bin/tfinfo-to-json.sh | jq -r '.openstack_compute_instance_v2 | keys[] as $k | [$k, .[$k]."network.0.fixed_ip_v4"] | @tsv'