`volume_size` being ignored and runner starts with minimum storage

[gusfune]

I am running a setup with 0.40.0 which uses ephemeral runners but in this setup it seems the variable volume_size is being ignored. I have set it up for a value of 40, but it only deploys instances with 8Gb. Also running ARM instances in this setup.

I don't have any custom block configuration or any heavy customisation. Pasting my config from main.tf file below:

module "runners" {
  source                          = "philips-labs/github-runner/aws"
  version                         = "0.40.0"
  create_service_linked_role_spot = true
  aws_region                      = local.aws_region
  vpc_id                          = module.vpc.vpc_id
  subnet_ids                      = module.vpc.private_subnets

  environment = local.environment

  github_app = {
    key_base64     = nonsensitive(OMITTED)
    id             = nonsensitive(OMITTED)
    webhook_secret = nonsensitive(OMITTED)
  }

  # Grab zip files via lambda_download
  runner_binaries_syncer_lambda_zip = "lambdas-download/runner-binaries-syncer.zip"
  runners_lambda_zip                = "lambdas-download/runners.zip"
  webhook_lambda_zip                = "lambdas-download/webhook.zip"

  # Runner setup
  enable_organization_runners = true
  enable_ssm_on_runners       = true
  instance_types              = ["t4g.large", "c6g.large"]
  runner_architecture         = "arm64"
  runner_extra_labels         = "default,aws"
  runner_os                   = "linux"
  runners_maximum_count       = 8
  volume_size                 = 40

  # ephermeral instance rules
  delay_webhook_event             = 0
  enable_job_queued_check         = true
  fifo_build_queue                = false
  job_queue_retention_in_seconds  = 300
  minimum_running_time_in_minutes = 1
  redrive_build_queue = {
    enabled         = true
    maxReceiveCount = 3
  }

  # override scaling down
  scale_down_schedule_expression = "cron(* * * * ? *)"

  tags = {
    Project = "GitHubAction"
  }
}

resource "aws_iam_service_linked_role" "spot" {
  aws_service_name = "spot.amazonaws.com"
}

# Other parts omitted as they are unrelated.

[npalm]

Are you using ubuntu or amazon linux? For ubuntu you have to configure the right mount point.

[gusfune]

Are you using ubuntu or amazon linux? For ubuntu you have to configure the right mount point.

Amazon Linux

[npalm]

Maybe it is related to ARM64, we use the same config (volume_size) for our fleet of x64 runners.

[gusfune]

Another finding: If I try to set the block mappings manually the same way a default instance shows, no blocks will be mapped and the instance is "volumeless".

[npalm]

Actually we have the same issu

[npalm]

We planning to depcrecated the variable volume_size. You can set the volume size by provide a block_device_mappings, see example below for amazon linux. In the PR #2014 we revert the default behavior to create an EBS with 30GB.

  block_device_mappings = [{
    device_name           = "/dev/xvd"
    delete_on_termination = true
    volume_type           = "gp3"
    volume_size           = 30
    encrypted             = true
    iops                  = null
  }]

[forest-releaser]

🎉 This issue has been resolved in version 0.40.3 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀