Description of problem: When upgrading from rhvh-4.4.10.3 to rhvh-4.5.0.5, the latest layer will switch to use system.devices file as lvm filter. During upgrade, the generated 'system.devices' by `vdsm-tool config-lvm-filter -y` will ignore a gluster VG on a VDO volume. imgbased.log during upgrade: ~~~ 2022-07-06 13:26:40,690 [DEBUG] (MainThread) Executing: ['nsenter', '--root=/tmp/mnt.QvpGq//', '--wd=/tmp/mnt.QvpGq//', 'vdsm-tool', 'config-lvm-filter', '-y'] 2022-07-06 13:26:41,708 [DEBUG] (MainThread) Result: b'Analyzing host...\nFound these mounted logical volumes on this host:\n\n logical volume: /dev/mapper/rhvh-rhvh--4.5.0.5--0.20220529.0+1\n mountpoint: /\n devices: /dev/vda2\n\n logical volume: /dev/mapper/rhvh-swap\n mountpoint: [SWAP]\n devices: /dev/vda2\n\nConfiguring LVM system.devices.\nDevices for following VGs will be imported:\n \n rhvh\n\nConfiguration completed successfully!\n\nPlease reboot to verify the configuration.\n \n' ~~~ After booting into the new 4.5.0 layer, gluster services can't be started because its VG (on a vdo volume) is filter out by etc/lvm/devices/system.devices file. This is impacting RHHI host upgrades. A quick workaround is to run `vgimportdevices` and the gluster VG will be generated correctly to system.devices. This steps seems to be skipped during the upgrade. How reproducible: 100% Steps to Reproduce: 1. install rhvh-4.4.10.3 2. add a new disk to rhvh-4.4.10.3 and setup vdo: $ vdo create --name=vdo1 --device=/dev/vdb $ pvcreate /dev/mapper/vdo1 $ vgcreate vdo_vg /dev/mapper/vdo1 $ lvcreate -L1G -n vdo_lv vdo_vg vdb 252:16 0 6G 0 disk `-vdo1 253:13 0 2G 0 vdo `-vdo_vg-vdo_lv 253:14 0 1G 0 lvm 3. format and add it to fstab: $ mkfs.xfs /dev/mapper/vdo_vg-vdo_lv 4. Install rhvh-4.5.0.5 and reboot Actual results: Booting up 4.5.0, the 'vdo_vg' is not activated because the generated system.devices file only contains 'rhvh'. Expected results: During upgrade any VG on top of a VDO volume is generated for 'system.devices' as well.
Moved to storage team on vdsm for investigation. The change doesn't seem related to RHV-H only.
Albert, duplicate of bz 2095588?
Yes, matches exactly what is observed in bz 2095588 .
*** This bug has been marked as a duplicate of bug 2095588 ***