revive ignition disks

[dustymabe]

currently ignition disks stage is missing from RHCOS because we pulled it out in the dracut modules. We need to re-evaluate this and put it back it and also make sure our "boot once detection" still works.

[dustymabe]

I believe from recent discussions this is a requirement for RHCOS so I've opened this card to start moving that work along. Will plan appropriately in JIRA

[ashcrow]

Is this stating that all of the disk work that ignition can do was dropped as part of the POC work? Or is this a more specific area of functionality that we didn't think was required?

[ashcrow]

/cc @bgilbert @ajeddeloh

[ashcrow]

This is the change you're referencing, correct? dustymabe/ignition-dracut@11d0ab5

[ajeddeloh]

Is this in reference to secondary disks or the primary (where RHCOS is installed) disk? The former is much simpler. It just needs to be run before the files stage but before any real filesystems are mounted.

Supporting the latter means ostree needs to evaluate if it needs to redeploy the rootfs (if the fs holding / has changed) and do that between the disks and files stage. We haven't built a POC but last I talked with @cgwalters he didn't seem to think there was a reason why it wouldn't work.

Regardless, doing the secondary disks first makes sense as a first step to handling the primary disk, so we should do that first.

[ashcrow]

/cc @crawford for his POV as well

[crawford]

I don't anticipate OpenShift directly needing this functionality, but I'm confident our customers will need it at launch.

[ashcrow]

Thanks all. So it does sound like we need to get this rolled back in.

Regardless, doing the secondary disks first makes sense as a first step to handling the primary disk, so we should do that first.

Does this sounds good to you as well @dustymabe and @crawford?

[dustymabe]

👍

[ashcrow]

I believe @lucab is working on this.

[lucab]

@ashcrow https://github.com/dustymabe/ignition-dracut/pull/17

[lucab]

I can't self-close this, but ignition-disks stage landed in 4.0.6138 via https://copr.fedorainfracloud.org/coprs/dustymabe/ignition/build/802355/ which includes the above PR. It has been manually smoke tested on libvirtd-qemu.

Things that RHCOS folks may still want to sort out after this:

• ostree-related ignition-files stage ordering is somehow unexpected: https://github.com/dustymabe/ignition-dracut/issues/20
• figuring out how to manipulate the primary disk in a ostree-aware way
• smoke-testing and cloud-integration specific test cases (e.g. NVMe on AWS)

[lucab]

/close

[openshift-ci-robot]

@lucab: Closing this issue.

In response to this:

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[dustymabe]

ignition-disks stage landed in 4.0.6138 via https://copr.fedorainfracloud.org/coprs/dustymabe/ignition/build/802355/ which includes the above PR. It has been manually smoke tested on libvirtd-qemu.

actually, that was the build that included https://github.com/dustymabe/ignition-dracut/pull/19. I'll build the one with ignition-disks in it today, sorry didn't get to that yesterday

[dustymabe]

https://src.fedoraproject.org/rpms/ignition/pull-request/11#

[dustymabe]

ignition-0.28.0-6.gitf707912 should be in 4.0.6174 and newer.