You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
and it would be really helpful to be able to run the same remote script, or a different one, when = update
My script is written to be safe to re-run under all circumstances, and doesn't actually need to be removed or cleaned up, it just needs to be reliably run every time there's any sort of local change.
Attempted Solutions
right now, what seems to happen is DESTROY is always run before CREATE, and I'd prefer if i could run an UPDATE instead.
Proposal
terraform_data supports a lifecycle element replace_triggered_by=[ stuff ]
it should also support a lifecycle element of update_triggered_by=[ stuff ]
eg:
this doesnt work but work be great:
in this hypothetical, "replace" would run destroy+create, but update doesn't run either of them
Thanks for this feature request! If you are viewing this issue and would like to indicate your interest, please use the 👍 reaction on the issue description to upvote this issue. We also welcome additional use case descriptions.
A well-written provisioner with current 'create' and 'destroy' capabilities is specifically hamstrung by not having "update" even made available in the ecosystem.
I have found that my current best-practive is to make my 'create' provisioner shell scripts arbitrarily re-runnable without concern [ ie enough basic if logic that they dont re-perform an action that's already been done to completion ] AND a 'destroy' provisioner that does almost nothing, but only because that's the only way to prevent execssive tear-downs and re-builds in the current no-update world.
Or to put it another way, real software engineers expect CRUD capabilities, ie Create,Read,Update,and Delete , which when mapped to TF terminology is "Create,Read,Destroy" ( whats missing )? -> Update! TF "fakes" a "update" by running "destroy+create" to mean the same thing , but in what world is that the same thing.? What that means is it artificially tears-down ( the destroy) a bunch of stuff we almost certainly didnt need to, to present a state that's more-likely to "create" friendly. but its really not a "create", its a update.
There are many reasons why supporting "when = update" on a provisioner can make more sense, and it could totally be done in a totally backward compatible way ... ie in the absence of a "when=update', it could just do the same behavior as it does now.
Terraform Version
Use Cases
i have a :
resource "terraform_data" "ssh_target" { ... }
and I "push" a shell script to the remote server:
and I currently run 2x provisioners :
and it would be really helpful to be able to run the same remote script, or a different one, when = update
My script is written to be safe to re-run under all circumstances, and doesn't actually need to be removed or cleaned up, it just needs to be reliably run every time there's any sort of local change.
Attempted Solutions
right now, what seems to happen is DESTROY is always run before CREATE, and I'd prefer if i could run an UPDATE instead.
Proposal
terraform_data supports a lifecycle element replace_triggered_by=[ stuff ]
it should also support a lifecycle element of update_triggered_by=[ stuff ]
eg:
this doesnt work but work be great:
in this hypothetical, "replace" would run destroy+create, but update doesn't run either of them
References
No response
The text was updated successfully, but these errors were encountered: