Hello! I’ve been looking into Pulumi for a greenfi...
# getting-started
Hello! I’ve been looking into Pulumi for a greenfield project, instead of Terraform. Looking to have state file in storage bucket, not in Pulumi service for this one, however I’m quite a bit put off by amount of issues and the fact they have been hanging there for a really time in the list of issues for self-managed backends. Is using self-managed backend common, or rare thing, to do in Pulumi, and is anyone using these for serious production use? In Terraform I’ve seen more self-managed backends than use of Terraform Cloud.
yes, usage of self managed backends is common and we’re aware of customers using self managed backends at a huge level of scale in production
I work for a company and we have hundreds of Pulumi projects across 81 AWS accounts, each with their own S3 backend. It is working quite well for us the last year.
@rich-stone-79955 Good to hear! Have you had any major issues? Do you execute Pulumi manually or in a pipeline? Do you keep project name in stack name, to work around https://github.com/pulumi/pulumi/issues/11320 ?
If you use only one backend for multiple projects then the workaround for now is creating your stack like: instead of
For one backend per project then there should not be any issue with the naming. I am using S3 as backend for all our IaC in pulumi
Hi @rich-stone-79955, would love the hear more about your setup. Happy to jump in a call if you are allowed to talk about it, ofc. 🙂
My company is using pulumi with Azure, for now really happy, but documentation I think is not so great, often trying to resolve issues with reading terraform docs 🙂 . A few times we had some issues with state file, we needed to remove manually resources, but mistake was on our side, not pulumi.
@ambitious-furniture-71289 - We’ve not had any major issues the last year. We do both. We execute in our
environment from both local and from a pipeline. This is a complaint I have with a self-hosted backend. The stack names and lack of segregation. I wish that the project name would be a “folder” (S3 doesn’t really have folders but part of the object name) and then the stack name. This would mimic the Pulumi Service. For now, what we do is duplicate the
so it would be something like
for the resources because both the project and stack names match. Not ideal but we deal with it. Because our
environment backend is separate from our
environment backend, this works fine for us.