I spoke to support a while back about how we are expected to use Terraform in automation when we cannot specify options like the execution mode to be local vs remote in the backend config.
The answer was to have a separate terrform codebase that runs locally to set up the remote backend using terraform itself.
I suggested that maybe a change could be made to allow these config options in the backend config, so that we wouldn’t have a chicken and egg situation, but this idea was rejected.
I was confused by this as how am I supposed to have multiple operators utilise terraform when one of them has to have the local copy of the initial setup? It seems flawed as a concept to me.
Has anyone dealt with this problem before, and if so, how did you get around it?