Uploaded image for project: 'RHEL'
  1. RHEL
  2. RHEL-96106

dracut needs to support populating /etc/multipath/wwids from the kernel commandline

Linking RHIVOS CVEs to...Migration: Automation ...RHELPRIO AssignedTeam ...SWIFT: POC ConversionSync from "Extern...XMLWordPrintable

    • dracut-057-101.git20250808.el9
    • No
    • Moderate
    • ZStream
    • rhel-bootloader
    • ssg_core_services
    • 26
    • 24
    • 3
    • Dev ack
    • False
    • False
    • Hide

      None

      Show
      None
    • None
    • None
    • Regression Exception
    • Unspecified
    • Unspecified
    • Unspecified
    • All
    • None

      What were you trying to do that didn't work?

      https://issues.redhat.com/browse/OCPBUGS-56597 needs dracut to support adding multipath WWIDs to /etc/multipath/wwids from the kernel commandline.

      OCP does not rebuild the initramfs to pull in /etc/multipath/wwids from the rootfs. Because of this, it needs some way to determine that a device should be multipathed in the initramfs, even if that device only has one path.  OCPBUGS-56597 wants to solve this issue by reading the WWIDs from the kernel commandline. The Red Hat versions of the multipathd.service systemd unit file include the command

      ExecStartPre=-/sbin/multipath -A
      

      in the Service section. This adds WWIDs from the kernel commandline (identified by "mpath.wwids=<WWID>") to the /etc/multipath/wwids file. Upstream dracut-ng used its own copy of multipathd.service until 5e87b68c ("refactor(multipath): remove custom multipathd.service"), which is included in dracut-107.  Dracut's version of mutipathd.service doesn't include this command ("-A" is a Red Hat only multipath option). In rhel-9, it still makes sense for dracut to use its own version of mulitpathd.service, but it needs to include the call the ExecStartPre command from above. In rhel-10, it is also safe to simply backport the dracut-ng commit that pulls in and modifies the rootfs version of multipathd.service.

      What is the impact of this issue to you?

      OpenShift Container Platform fails to boot when there is only one path of a multipathed root device available.

      Please provide the package NVR for which the bug is seen:

      All versions. I believe OCP needs this fixed starting in rhel-9.4

      How reproducible is this bug?:

      Always when the conditions are met.

      Steps to reproduce in RHEL

      1. install a mutipathed root system.
      2. back up /etc/multipath/wwids
      3. remove all the wwids from /etc/multipath/wwids
      4. rebuild the initramfs
      5. restore the backed up /etc/mutipath/wwids
      6. fail all of the root multipath device's paths except one.
      7. reboot the system

      Expected results

      The system will drop into an emergency shell during bootup, since the root device will not be multipathed (depending on the how the /etc/fstab is setup, you might instead boot up with the root filesystem mounted directly on the remaining path device, instead of using a multipath device).

      Actual results

      The system will reboot correctly, with a multipathed root.

              pvalena@redhat.com Pavel Valena
              rhn-engineering-bmarzins Benjamin Marzinski
              dracut maint mailing list dracut maint mailing list
              Frantisek Sumsal Frantisek Sumsal
              Votes:
              0 Vote for this issue
              Watchers:
              9 Start watching this issue

                Created:
                Updated:
                Resolved: