Hi All, I am stuck in post-draining stage in our upgrade from 1.7.1 to 1.8.2 the main cause is the unable to mount cos-state as read-only as mount point is busy and it fails on each and every retry as well. I have tried deleting the post-drain pod as well still it gives the same error. The node's boot disk (single-path local RAID volume, no real external storage) is unnecessarily getting wrapped in a device-mapper multipath map at every boot, because

multipath=off

isn't on the kernel command line (

multipath -ll

shows only one path;

/oem/harvester.config

shows

externalstorage.enabled: false

). This creates

COS_STATE

being reachable via both

/dev/sdaX

and

/dev/mapper/<wwid>-partX

simultaneously. Has anyone run into a similar problem.

Are you saying you need multipath=off for the kernel line?

This is what the dmesg is showing

could you run the

lsblk -o NAME,SIZE,FSTYPE,LABEL,MOUNTPOINT

command on the affected node?

See the findmt cmd resolves to a dev-mapper path not the disk itself

you mentioned the disk "is unnecessarily getting wrapped in a device-mapper multipath map at every boot, because

multipath=off

isn't on the kernel command line". Ordinarily,

multipath=off

should be there on the kernel command line. Was it removed for some reason? Can you add it back in and have the problem go away? What does

grub2-editenv /oem/grubenv list

show for

third_party_kernel_args

?