Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.

Bug 1809534

Summary: Enable multipath for boot lun on RHCOS
Product: OpenShift Container Platform Reporter: Ravi <rdave>
Component: RHCOSAssignee: Ben Howard <behoward>
Status: CLOSED NOTABUG QA Contact: Michael Nguyen <mnguyen>
Severity: medium Docs Contact:
Priority: medium    
Version: unspecifiedCC: aghadge, bbreard, behoward, dapark, dcain, dornelas, dustymabe, fiezzi, ibodunov, imcleod, jligon, kbell, kelly.brown1, kholtz, miabbott, nkurane, nstielau, oarribas, pchavan, rbost, smilner, umesh_sunnapu, walters
Target Milestone: ---Keywords: Reopened
Target Release: 4.6.0   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2020-06-08 18:24:35 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On:    
Bug Blocks: 1186913, 1903544    

Description Ravi 2020-03-03 11:02:22 UTC
Description of problem:

Enable multipath for boot lun on RHCOS

Version-Release number of selected component (if applicable):

OpenShift Container Platform 4.2
RHCOS 

How reproducible:

Often

Steps to Reproduce:
1. use multipath lun as boot lun for rhcos
2.
3.

Actual results:

It is using underlying subpaths as boot lun like /dev/sdX

Expected results:

It should use /dev/mpathx as boot device

Additional info:

We have a customer who wants to have a multipath device as a boot lun on RHCOS which is booting from SAN. Currently boot devices are getting mounted on one of the subpaths.

We checked that enabling multipath for other than boot lun works fine without any issue, however for boot lun it is not working and everytime it is taking underlying sd* device.

We also try checking early loading of dm-multipath module during boot but it did not help. Following are the steps used:

[1] Enable loading of multipathd module during boot

$ echo dm_multipath > /etc/modules-load.d/dm_multipath.conf

[2] Confirm if the multipathd service is enabled during core os boot.

[3] Reboot the server and check if boot LUN gets activated over the multipath device or not

Comment 1 Colin Walters 2020-03-17 22:28:45 UTC
Which type of SAN?  How exactly are you setting this up?  Are you using Ignition to do this configuration?

I don't think we've documented support for booting from any kind of SAN.

Some related discussion of multipath here https://github.com/coreos/ignition-dracut/issues/154

Comment 5 Ben Howard 2020-04-22 22:53:02 UTC
I've done a ton of digging on this, and really its a missing feature, not a bug.

RHCOS has its roots in the cloud, which have much simpler use cases. 

To date, there is no automatic method, I'm looking at how to do it on day one.
The best option now is:
- drop /etc/multipath.conf
- enable initramfs regeneration (rpm-ostree initramfs --enable)
- change the Kargs (rpm-ostree kargs --editor) to drop "ostree=..." and add in "root=/dev/mapper/<PATH>"
- reboot

I don't like this answer, but that is the state of play today.

For 4.6 we should be able to get most of the way there with:
https://github.com/dracutdevs/dracut/commit/b8a92b715677d52dbc2b27a710b9816fd8b9a63b

There's also discussion here:
https://github.com/coreos/fedora-coreos-tracker/issues/465

So we need to do a bit more work to make the UX better. Once we have all this, the CoreOS Installer might need to become aware too.

Comment 8 Micah Abbott 2020-04-23 15:43:21 UTC
Moving back to ASSIGNED and dropping from errata.  We haven't delivered anything for 4.5, but there is no 4.6 target, so we are in a BZ limbo state right now.

Comment 11 Ben Howard 2020-05-18 16:12:35 UTC
The work that I have been engaged in is trying to make multipath ergonomic and easy to use. However, in light of https://bugzilla.redhat.com/show_bug.cgi?id=1809534#c5, this is not a bug. 

The multipath feature work for automated multipath will likely land in 4.6.

Comment 15 Micah Abbott 2020-06-08 18:24:35 UTC
As per comment #5, this is not really a bug, but a feature request that needs to land in RHCOS.  We hope to have this available for OCP 4.6.

We don't officially support any kind of use of multipath devices for booting RHCOS; any kind of instructions or workarounds found here are officially *unsupported* and would require a Support Exception to be endorsed in any way by the support team or engineering team.

Please use traditional support channels to request additional information about this feature.