Post

Kernel 6.19.0 has changed stuff with md devices. There's a new setting for logical_block_size (LBS) in /sys/devices/virtual/block/md*/ , and 6.19.0 will only mount md read-only where it's not configured. Also if you echo anything other than 4096 (the default) to that, you will destroy any PV you

danceswithcrows

@crow202.com

have on that md device. I'm just glad I have a routine of backing up every night, though the restore took forever because for some reason my backup disk showed up as USB2 instead of USB3. tl;dr: Do not use 6.19.0 if you use md and LVM, the default LBS is 4096, and you can't use a kernel < 6.19.0

15/2/26, 2:45 am0 reposts0 likes

danceswithcrows

@crow202.com

15 Feb

if you set logical_block_size on any md device. "Don't break userspace"? lvm and mdadm are AFAICT (mostly) userspace. I could fix it all eventually, but it wasted hours. The kernel message saying "echo current LBS to md/logical_block_size" should have a "(default 4096)" in it after "LBS".