As per discussion with hashicorp, we had finalized two workspace per environment per project strategy in our organization.
But while working on projects, we faced issue of operation dependancy. We have network components in network workspace and compute resources in compute workspace. Now we have created all required configuration file for both network and compute and we are using output and remote data block for dynamic resource dependancy. But we want task dependancy so that compute workspace will wait for all components in network to be created and then it will trigger an run.
Ex. In azure,
Resource group will be created in compute workspace
Then Vnet subnet nsg routetable nic will be created in Network workspace
And then finally vm or aks will be created in compute again.so it will wait to complete resource deployment in network workspace.
If your dependencies are this complicated, you need all this stuff in one workspace.
What is the reason for trying to divide it up across multiple workspaces?
EDIT: I should add, that although Terraform Cloud/Enterprise has a feature for running one workspace after another, Run Triggers - Workspaces - Terraform Cloud | Terraform | HashiCorp Developer, I don’t think it would be a good idea to set up a cycle of workspaces which trigger each other - if that’s even allowed.
EDIT: Just tested in Terraform Cloud…
First, the Run Triggers feature carries a note which makes it kind of useless:
Runs initiated as the result of a run trigger connection will not auto-apply, regardless of your auto-apply setting selection. You will need to manually apply these runs.
Second, it actively prevents creating cyclic dependencies: