Topics created by Wugutech

https://access.redhat.com/solutions/5347091

System fails to mount filesystem on multipath partition during boot time in RHEL8
Solution Verified - Updated May 31 2021 at 2:25 PM - English
Environment

Red Hat Enterprise Linux 8 dracut-049-27.git20190906.el8.x86_64

Issue

System fails to mount filesystem created on multipath partition during boot time in RHEL8.

Raw

[ 235.191955] testsystem systemd[1]: Mounting /sysroot...
[ 235.398136] testsystem kernel: SGI XFS with ACLs, security attributes, no debug enabled
[ 235.398913] testsystem mount[2922]: mount: /sysroot: /dev/sdb4 already mounted or mount point busy.
[ 235.399207] testsystem systemd[1]: sysroot.mount: Mount process exited, code=exited status=32
[ 235.399259] testsystem systemd[1]: sysroot.mount: Failed with result 'exit-code'.
[ 235.399480] testsystem systemd[1]: Failed to mount /sysroot.

Resolution

Upgrade dracut package to dracut-049-135.git20210121.el8 as per RHBA-2021:1661 As a workaround, rebuild initramfs with /lib/udev/rules.d/11-dm-parts.rules.

Raw

dracut -f -v --include /lib/udev/rules.d/11-dm-parts.rules /lib/udev/rules.d/11-dm-parts.rules

Or create /etc/dracut.conf.d/part.conf with below lines and rebuild initramfs image.
#vi /etc/dracut.conf.d/part.conf
install_items+=/lib/udev/rules.d/11-dm-parts.rules

Root Cause

dracut doesn't install /lib/udev/rules.d/11-dm-parts.rules, which is what changes the link priority for kpartx partition devices, which should make /dev/disk/by-uuid point to multipath devices instead of the scsi path devices.

Diagnostic Steps

Check boot logs and verify whether root filesystem mounted using sdX device or mpath device.

Raw

[ 235.185721] testsystem systemd[1]: Starting File System Check on /dev/disk/by-uuid/3a9fc23b-e95d-44a2-b95c-9651799f8c77...
[ 235.190597] testsystem systemd-fsck[2920]: fsck.none doesn't exist, not checking file system on /dev/disk/by-uuid/3a9fc23b-e95d-44a2-b95c-9651799f8c77.
[ 235.191192] testsystem systemd[1]: Started File System Check on /dev/disk/by-uuid/3a9fc23b-e95d-44a2-b95c-9651799f8c77.
[ 235.191955] testsystem systemd[1]: Mounting /sysroot...
[ 235.398913] testsystem mount[2922]: mount: /sysroot: /dev/sdb4 already mounted or mount point busy.

Verify /lib/udev/rules.d/11-dm-parts.rules rule available in initramfs image.

Raw

lsinitrd initramfs-4.18.0-193.el8.x86_64.img |grep dm-parts |wc -l

0

Rebuild initramfs image with /usr/lib/udev/rules.d/11-dm-parts.rules.

Raw

#lsinitrd initramfs-4.18.0-193.el8.x86_64.img |grep dm-parts
Arguments: -f -v -a 'multipath' --include '/usr/lib/udev/rules.d/11-dm-parts.rules' '/usr/lib/udev/rules.d/11-dm-parts.rules'
-rw-r--r-- 1 root root 1442 Dec 17 2019 usr/lib/udev/rules.d/11-dm-parts.rules

Short answer: you can't.
Long answer: you can "cheat" with dynamic memory.

But in general, I would avoid it.

its working, it just took very long time to detect,

@TheRedHerring I have an HM90 that I use as a host and it worked with 8.2 out of the box except for the 2.5G NIC. I switched to 8.3 and everything worked out of the box.

Existing HA level is at VM level. It means, if the VM is down (or the host on the VM is running is down), it will be booted on a working host in that pool.

For hyperconvergence, it's far from being a miracle solution. It costs a lot of thinking, planning and maintenance (and careful usage of it).

Like HA in the end.

So before asking for advanced features, it's good to think about what you really need.