This message was deleted.
# general
s
This message was deleted.
đź‘€ 2
t
Do you know of any concerns I should have?
Nothing concrete comes to mind. It sounds like you are asking the right questions.
How transactional is pulumi?
It’s not really transactional in the sense that a deployment may fail on any resource (if a resource can’t be created) and it will stop at that point. The state file will be self-consistent but the a multi-resource deployment is not atomic. There are no built-in rollbacks. Providers do some number of retries for transient errors.
Hoping to mitigate them when temporal attempts to retry.
Retries may help for long-ish transient errors that a provider did not wait to resolve but retries wont help for a lot of errors in practice (say, a resource’s properties aren’t configured correctly, or a conflict with pre-existing infra).
b
@tall-librarian-49374 Thanks for the info. I've been tasked with cleaning up some Infrastructure deployment code / patterns. Probably going to attempt to create workflows / patterns where we limit the amount of resources and state in 1 single TF / Pulumi by wrapping it in workflow.
At the moment, I think we may just dump state before every execution. Not sure if there is a happy path to be able to take the current state file, the intended state file, and what is actually in the infrastructure and inline remediate a deployment with "import"