Crossplane – Open Source Multicloud Control Plane | Hacker News Reader

Crossplane – Open Source Multicloud Control Plane

blog.upbound.io

185 pointssytse36 comments

At GitLab we're really excited about the vision of Crossplane. Multi-cloud seems to be the future [0]. This is the missing technology to make that convenient. Our first step is to work with Bassam to become the first complex application deployed in this way. We already have a popular helm chart. But that doesn't set you up with things like object storage and a cloud managed databases (PostgreSQL and Redis).

After we make that work we will integrate Crossplane in Auto DevOps [1] so you achieve application portability for the apps you make in GitLab.

The most interesting thing to me is that Crossplane works outside Kubernetes since k8s is another service it will provision. Bassam his plan is to make crossplane + Kube-api more tightly integrated, maybe even a crossplaned at somepoint.

0. https://medium.com/gitlab-magazine/multi-cloud-maturity-mode...

1. https://docs.gitlab.com/ee/topics/autodevops/

What is the business model of upbound/Crossplane? Product seems to be free.

hi bassam here (CEO @ Upbound). Our business model is still firming up, but we're thinking we would be hosting crossplanes among other things. We want the crossplane project to be community driven (and not vendor driven) and welcome other contributors (see our governance model).

Thanks for CrossPlane, this will work if you guys stay true to your mission. A frustrated cloud user. Personally I will support you guys and GitLab any day.

Thanks but aren't you worried about someone else hosting your product and competing with you.

Crossplane could be one package that customers would not want to use directly from a major cloud provider. You'd never trust AWS/Google/etc to really give equal footing to a competitor. Separately, I'd prefer to trust the company behind the open source software as opposed to some other upstart. So this model may really work.

One observation: Crossplane and Upbound seem a little too unrelated of names. Github has benefitted from being almost synonymous with git (admittedly, Github also created the opening for Gitlab at the same time).

Great response! Regarding naming we did not want to tie the company to a single project, not did we want the project to feel like a vendor driven project. So we went with different names completely :-)

I asked Bassam if it would be completely open source (and not open core like GitLab) in the architecture document [0]. He said it would be completely open source [1].

0. https://docs.google.com/document/d/1whncqdUeU2cATGEJhHvzXWC9...

1. https://www.dropbox.com/s/bypp7unw8802gkr/Screenshot%202018-...

This looks really exciting and congratulations on launching! We will definitely be taking a look at this as it looks like a great fit to recommend to our customers.

My startup Meshcloud [0] is helping companies manage the "organizational" side of multi-cloud like integrating account provisioning, SSO, policies, chargeback, quotas etc. using a declarative orchestration approach we call "org as code" (basically infrastructure as code but for org stuff). Our product is e.g. used by a large german car manufacturer to orchestrate hundreds of tenants across AWS, Azure and a mix of private Clouds based on OpenStack, Kubernetes and Cloud Foundry. Workload orchestration is obviously the next challenge down the road for these customers, so we're currently looking at providing good integration to tools like Open Service Broker API, Terraform, Bosh, Spinnaker etc. Each of those solves a slightly different set of problems but at the scale our customers are operating at they all have their place.

0. https://www.meshcloud.io/en/meshcloud/

Your comment here did a better job explaining your product than the website.

This looks fairly similar to the new Kuberentes KEP for portable service definitions.... https://github.com/kubernetes/enhancements/blob/master/keps/...

Very similar. Can someone more familiar with Crossplane/KEP comment?

Hi, this is ichekrygin (crossplane maintainer). Yes, indeed. There appears to be an overlap in general principles (even in use-case example). From what it appears, KEP sig was "published/released" (not sure what is the right term here) on the GitHub 3 days ago, we did not see this. However, now since it out there, we will definitely reach out to discuss.

Yes, indeed. There appears to be an overlap in general principles (even in use-case example). From what it appears, KEP sig was "published/released" (not sure what is the right term here) on the GitHub 3 days ago, we did not see this. We will definitely reach out for further discussion.

This looks awesome. Do you know what the latest status on this is?

I thought it was going to be a cross-cloud Terraform, but it looks like it's a bunch of presets operating on k8s? So when he asks for the DB, is he getting an RDS/CloudSQL/whatever or is it spinning up a DB on an EKS/GKE/whatever?

I really don't like the latter in comparison to the former, tbh.

scarejunba, this bassam (one of the maintainers on crossplane).

when you ask for a DB you get an instance of RDS/CloudSQL/etc., in other words you are using the managed services of your cloud provider.

we're just getting started with cross-cloud scenarios. would love your feedback on https://github.com/crossplaneio/crossplane

Oh, I get it. Reading the design document[0] helped greatly. Should've done more than skim the blogpost. The definition of what you want is specified as a custom resource that is cloud-agnostic. Then a set of cloud-specific custom controllers are installed on a K8s instance and they control which cloud is affected by your `kubectl apply`.

You have to set up your own k8s manually (there's no way to bootstrap), but then it's all fun and games.

Nice. Very clever.

0: https://docs.google.com/document/d/1whncqdUeU2cATGEJhHvzXWC9...

> I thought it was going to be a cross-cloud Terraform

Isn't Terraform already cross-cloud?

From the architecture document:

Terraform is a popular tool for provisioning infrastructure across cloud providers. It offers a declarative configuration language with support for templating, composability, referential integrity and dependency management. Terraform can dynamically provision infrastructure and perform changes when the tool is run by a human. Unlike Crossplane, Terraform does not support workload portability across cloud providers, and does not have any active controllers that can react to failures, or make changes to running infrastructure without human intervention. Terraform attempts to solve multicloud at the tool level, while Crossplane is at the API and control plane level. Terraform is open source under a MPL license, and follows an open core business model, with a number of its features closed source. We are evaluating whether we can use Terraform to accelerate the development of resource controllers in Crossplane.

It's a missed opportunity to link your blog back to blog.upboundio instead of to www.upbound.io

Blog posts often lead me to wonder what the product is, so it should be convenient to find out.

What are the benefits of using Crossplane as opposed to Terraform?

It seems like you would still use both. Terraform to provision your Kubernetes cluster and other non-application specific infrastructure.

It seems like, and I have only briefly read the article, Crossplane is meant to bring all that application specific cloud infrastructure into the same configuration files you would use for defining you Kubernetes application.

So along with defining your ingress you define your mysql database and it gets provisioned. This keeps all your application specific provisioning together instead of having to use Terraform to provision your database ahead of time and keep it in sync with the application.

See https://news.ycombinator.com/item?id=18603051

I really like idea of separating a service’s infra requirements from it’s infra implementation details.

> Crossplane supports a clean separation of concerns between developers and administrators.

If I’ve understood correctly im not sure this is the best way to sell the idea though.

How does this work in a world where developers administer their own services?

> How does this work in a world where developer administer their own services?

in that world the developer plays admin too, and can configure crossplane as an admin.

Why do i need to configure a mysql database for a wordpress if that wordpress always needs a mysql database?

While i really like the idea, i'm looking forward to see if we moved our problems down to the operator code or if those configuration problems are gone.

Wordpress supports MySQL or MariahDB.

Other software supports multiple database systems.

There are opinionated packages of the sort you are looking for, I think, perhaps you would be interested in sandstorm.io?

So I could use this to setup a local dev env? Instead of using something like https://github.com/localstack/localstack

Anyone care to comment on whether this complements or competes with Rancher 2.0?

Seems to me you would run this on a Rancher 2.0 cluster, and then get automagically provisioned databases. It extends the default kube system, and Rancher just provides an interface to that.

It looks perfectly parallel as far as I can see. Very interested in answers here too.

So how is this different from AWS Operator?

Hi, this bassam (one of the maintainers on crossplane). We describe the relationship in our arch doc here https://docs.google.com/document/d/1whncqdUeU2cATGEJhHvzXWC9.... The following snippet is from that doc:

The AWS Service Operator is a recent project that implements a set of Kubernetes controllers that are able to provision managed services in AWS. It defines a set of CRDs for managed services like DynamoDB, and controllers that can provision them via AWS CloudFormation. It is similar to Crossplane in that it can provision managed services in AWS. Crossplane goes a lot further by offering workload portability across cloud multiple cloud providers, separation of concern, and a scheduler for workload and resources.

This looks to work similarly to the Persistent Volumes in Kubernetes. To put it into OOP language Kubernetes persistent storage is modeled as an abstract class and a set of concrete types that implement a class.

The advantage of this model is that an application requests the "Large and Slow Disk Class" not knowing how it works on each environment and a cluster owner then describes how to configure a spinning disk EC2 Storage to fulfill this class.

So, the AWS Operator lets people use the Kubernetes APIs to configure AWS resources. But, if you use those APIs directly in say your Helm chart then your app won't be able to be configured and deployed on other cloud providers without editing those objects.

Is it important to your application to be loosely coupled to the cloud provider? Then you need this sort of abstraction to enable portability. It sort of continues the Kubernetes vision beyond computer/network/storage into other cloud services like databases/serverless/AI/etc.

The AWS operator targets AWS resources. You can't take them, in a general sense, to minikube, Azure, or elsewhere. It's declarative to AWS services.

Some would call the AWS operator as a form of vendor lock-in.

I think the idea here is about using services with portable apps across cloud provider.

Reply on news.ycombinator.com