Existing deployment migration to 1.0.0 fails

[rberger]

When updating an existing deployment that had been done with a version before 1.0.0, plan and apply fail with the error message:

│ Error: Cycle: module.superblocks_agent.module.dns[0].aws_acm_certificate.superblocks (destroy), module.superblocks_agent.module.dns[0].aws_route53_record.validation_record (destroy), module.superblocks_agent.module.lb[0].aws_lb_listener.superblocks (destroy deposed 35ec1523), module.superblocks_agent.module.dns[0].aws_acm_certificate_validation.superblocks (destroy), module.superblocks_agent.module.lb[0].aws_lb_listener.superblocks
│

My usage of the module:

  source  = "superblocksteam/superblocks/aws"
  version = ">=0.1.0"

  domain                        = var.domain
  ecs_subnet_ids                = local.ecs_subnet_ids
  lb_internal                   = var.lb_internal
  lb_subnet_ids                 = local.alb_subnet_ids
  subdomain                     = var.subdomain
  superblocks_agent_data_domain = var.superblocks_agent_data_domain
  superblocks_agent_environment = var.superblocks_agent_environment
  superblocks_agent_key         = local.superblocks_agent_key
  vpc_id                        = local.vpc_id
}

[bruce-y]

Hi @rberger! This may indicate that depends_on was not specified correctly in that older module which I believe has been fixed in the latest version. The DNS creation and the cert creation used to be created in the same submodule which then caused a cyclical dependency with the loadbalancer submodule. The CNAME target would not be ready before the loadbalancer was up and the loadbalancer would not be up unless the cert was created prior.

I think if you do targeted destroy commands by removing the Route53 record first in the dns module, it should end this cyclical error.