Uploaded image for project: 'RHEL'
  1. RHEL
  2. RHEL-8351

RFE: default to creating lvmdevices file on RPM install

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

    • Icon: Story Story
    • Resolution: Duplicate
    • Icon: Undefined Undefined
    • None
    • rhel-9.2.0
    • lvm2
    • None
    • rhel-storage-lvm
    • ssg_filesystems_storage_and_HA
    • None
    • False
    • False
    • Hide

      None

      Show
      None
    • None
    • None
    • None
    • None
    • If docs needed, set a value
    • None
    • 57,005

      Description of problem:

      In RHEL 9 an LVM devices file gets created by Anaconda if the install includes an LVM storage configuration [1]. This is an RFE to make the creation of the devices file in `/etc/` always happen (at least for new installs) via RPM scriptlets instead.

      [1] https://github.com/rhinstaller/anaconda/blob/6f604b55f6caa5d570764a1d236287fe7ad735ab/pyanaconda/modules/storage/installation.py#L334-L355

      Background:

      RHCOS (basis of OpenShift) doesn't use Anaconda to perform an installation. It does use a tool (`coreos-installer`), but the tool mostly just does a dd of a prebuilt image to the disk (basic partition+filesystems, no LVM). The pre-built image does include the LVM RPMs, though.

      We are currently encountering an issue [2] where Kubevirt guests (virtual machines) where the users choose to put LVM on attached block devices is conflicting with the host (both guest and host are interacting with the LVM devices). Here are a few details:

      • Running pods (containers) that start VirtualMachines.
      • These VirtualMachines can have block storage attached to them.
      • Users can choose to create LVM PVs out of this block storage.
      • Both the Host (RHCOS Running OpenShift) and Guest see LVM signatures on these devices.
      • Need a way to tell the host to completely ignore the devices attached to the guest.
      • lvmdevices is relatively new, but presents an option
      • RHCOS can implement a workaround/fix for this for now
      • but would like to not behave differently than RHEL in the long term

      For RHCOS we are planning to implement a workaround for this where upgrading nodes have a script that gets run to create an LVM devices file (either empty or with any detected LVM devices from the host), but in the long run we'd like to not differ from RHEL here, which means have a solution to this problem in the RPMs I believe.

      [2] https://issues.redhat.com/browse/OCPBUGS-5223

              lvm-team lvm-team
              jira-bugzilla-migration RH Bugzilla Integration
              lvm-team lvm-team
              Cluster QE Cluster QE
              Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

                Created:
                Updated:
                Resolved: