Is it possible to provision a CI/CD pipeline from within the Pulumi code, like it's done in AWS CDK Pipelines?

Is this what you are looking for? https://www.pulumi.com/registry/packages/aws/api-docs/codepipeline/pipeline/

No. CDK pipelines make it easy and abstract to deploy self mutating pipelines that deploy cdk stacks to different environment.

ah ok, ive never used cdk myself but at pulumi up they announced the interop between pulumi and cdk, you might find something useful in here: https://github.com/pulumi/pulumi-cdk

I've asked whether CDK pipelines were supported over at #pulumi-cdk, but didn't get a response. I don't think it can be supported, though, that's why I asked here for an alternative.

I'll repeat myself. If you can AVOID all AWS Code** solutions because they are not developer friendly at all. You'll get done everything 3x faster with other CI solutions even Jenkins

I've been using CodePipeline/CodeBuild for a while and they work fine. I'm not married to them, that's why my original question doesn't mention them. I'm asking if it's possible to provision a self-mutating CI/CD process in the same Pulumi project that deploys the actual apps.

That's the beauty of all other CI/CD pipelines, you do not need anything self mutating garbage. But you can achieve that with pulumi I'm sure.

Why don't I? For example, if I want to add another stage in the pipeline, I want to be able to do it with a simple

git push

of my IaC code. The pipeline fetches the code and updates itself.

because pipeline definition lives together with code.. eg GitlabCI - just add 1 section in yaml and that's it. For implementation, you'll need to write quite a bit of code and end up with a similar code CDK provides

I don't want to write yaml, I want to write my IaC in TS/Python/etc, because the CI/CD definition is just as dynamic as the other infra.

the point of gitlab/github is that you do not care about infra at all

CI/CD is infra too. As mentioned, the pipeline definitions are dynamic and depend on the rest of the infra code. For example, I want to be able to automatically provision/tear down pipelines based on my infra code, and to modify them as well. Does this have to involve manual changes to the yaml config, or can Pulumi handle this?

pulumi gives you full programming language to do it. I'm just saying it's not worth it. I have pipelines with 10+ stages/jobs in Github/Gitlab and if I need to disable anything it's as simple as commenting out a few lines of YAML file. And I did not have to touch any infra at all. The experience you are trying to achieve with CodePipeline is exactly that but at the cost of your time developing workarounds for crippled AWS Service.

if I need to disable anything it's as simple as commenting out a few lines of YAML file.

You're missing the point - I'm saying I want to change a line in my Pulumi code, and that should not only mutate the infra, but the pipelines as well - for example, provision new ones. I don't want to make any manual changes to yaml.

Here are very few users of AWS Code* tools, usually, the main reason to use it is silly compliance. So the chance this is solved is very low. Writing Pulumi is rather simple however you need application architecture skills. For example to add actions to the pipeline, you can read the directory that contains your projects.

infra/
app/1
app/2

so every time you add something to

app/*

the infra code will look there and creates/removes action/job/stage - but I think with CodePipeline it's going to be a lot of code

@freezing-umbrella-5483 Hi. I'm a Pulumi employee and have also used CDK Pipelines in a previous job as a consultant. I think the short answer is "no". Right now you'd have to create this from the raw resources in the AWS provider. I'm aware firsthand of how difficult this is. I would probably try creating the pipeline with regular CDK constructs and using the CDK-Pulumi integration to deploy. While they aren't self-mutating and it's slightly more code compared to CDK Pipelines, they will take care of what can be some very tricky IAM perms. I've created a feature request in AWSX. Feel free to upvote if you would prefer a native Pulumi component (although it would not be self-mutating): https://github.com/pulumi/pulumi-awsx/issues/835

@stocky-restaurant-98004 thank you for the response! Unfortunately, it's far from "slightly" more code, as CDK pipelines abstracts so much away - especially with deploying to multiple environments (accounts/regions) and with creating actions for building/uploading assets automatically. I'm not married to AWS's CI/CD tools, as mentioned, so I'd welcome an abstraction like CDK Pipelines that works with other providers as well. The ultimate goal is to have

pulumi up

provision CI/CD pipelines that will in turn use

pulumi up

to deploy the app itself, and have it all dynamic and within a single app. Not sure I explained it well, but you know what I'm talking about, having worked with CDK pipelines yourself. Appreciate the feature request!