#9681 Install Openshift on AWS for Fedora CI
Closed: Will Not/Can Not fix 2 years ago by phsmoura. Opened 3 years ago by mvadkert.

While we love our current EKS setup, we really lack the nice Openshift 4 features. Now that Openshift on AWS offering from Amazon is officially out, we would like to standardize our setup and use Openshift instead of our EKS setup.

I talked to @davdunc about this on the Public Cloud meetup on Devconf, and seems this would be possible to setup on the Fedora CI account.

So I am filing this request here, so we can move in this direction slowly.

CC @kevin @mobrien @astepano @msrb @bookwar


Metadata Update from @smooge:
- Issue priority set to: Waiting on Assignee (was: Needs Review)
- Issue tagged with: OpenShift, medium-gain, medium-trouble, ops

3 years ago

I would definitely like to work on this ticket but as it is a new offering I don't really know what would be involved in setting it up so may need some guidance @dkirwan may be able to give some input here

Please don't start setting this up yet. ;)

We are pondering on where and how best to do it.

@mvadkert some more information when you have time:

  • Do you need to run/manage the cluster yourself? Or would you be fine if someone else ran the cluster and you just did your work in it?
  • Do you have any idea of sizing you might need for a cluster to replace the current eks/your needs moving forward?

Thanks.

@kevin

Please don't start setting this up yet. ;)

Agreed!

We are pondering on where and how best to do it.

Ack, sure thing

Do you need to run/manage the cluster yourself? Or would you be fine if someone else ran the cluster and you just did your work in it?

Yes, we need cluster admin, we are fine with managing it ourselves as we do with EKS. We are developing our own operators for our services, so we need to be able to install our development operators for CI purposes for example.

Do you have any idea of sizing you might need for a cluster to replace the current eks/your needs moving forward?

Currently we use ~10 m1.xlarge instances for EKS

@kevin @mobrien @dkirwan hi, any update here? There is now a nice rosa installer available, we just used it to setup an internal RH openshift dedicated cluster, I wonder if we could use to easily install the cluster. It is pretty straight forward.

We are blocked on @davdunc making some changes to the aws account we want to run these clusters in. ;(
Apparently it's something that affects all the community accounts (or could), so he is being very carefull about making the change.

@davdunc any news? :)

@davdunc any news on that account setting?

So, turns out @davdunc cannot make those changes to our account because it would affect every account in our community org and break things. ;(

So, now we are asking Openshift folks if there's some workaround for not having those settings. ;(

Thanks for the update, we will see, still hoping :)

We are happy with AWS EKS now, just wanted to use openshift everywhere ...

Yeah, we really would like a dev openshift cluster too. ;(

So, a bit of progress finally on this. @dkirwan managed to get a cluster installed and working. :)

But talking to him and other folks I am not sure we are on the same page, so could you share some information?

Where does fedora-ci run currently? Just eks? Or eks and centos-ci?

What parts would you want to run in a aws openshift cluster? Just the eks bits? as much as you could?

How much capacity does the current setup take?

Would it be safe to run fedora-ci loads in a cluter with other applications? or is that a bad idea due to user modifiyable ci scripts?

Thanks for any info and hopefully we can move this forward and find a good place for it.

Testing Farm (Fedora CI backend) runs on EC2 and Fedora AWS EKS. Most of the testing happens on EC2 instances, running tests on the cluster is not very useful for testing the OS.

As for Fedora CI, I believe they are on AWS EKS also (another cluster).

Is this cluster something we can have dedicated for my team? Is the exprience as with AWS EKS? Can it be automated? We now having fun with terraform to manage our infra via git, and it would be a step back to go to something manually provisioned.

So, is this still required/needed?

Last time we talked about it is not a priority for us. Although, if we would have the ability to setup a openshift4 cluster from scratch ourselves sometime, it would be great.

@mobrien I fail to remember, did we say this is now not possible right?

@mvadkert From what I remember from the last conversation we had your team was going to continue using the resources you have now and look into creating an openshift cluster down the line with terraform and you would reach out to us when you are looking at that for further permission/help.

Its been a while so I could be a bit off there but that's what I thought was the status

Good :) so are we now permitted to try to run ROSA (https://aws.amazon.com/rosa/) on the Fedora AWS account?

The main fedora community account. No. There's way too many things running there to try a ROSA installer. :)

@dkirwan was able to install in a seperate CentOS community account, but I don't know if he used ROSA or not. David?

Discussed about this during the weekly and we are closing. Feel free to reopen if needed

Metadata Update from @phsmoura:
- Issue close_status updated to: Will Not/Can Not fix
- Issue status updated to: Closed (was: Open)

2 years ago

Login to comment on this ticket.

Metadata
Boards 1
ops Status: Backlog