how to repair out of sync lv mirror with raidintegrity=y ?

[devZer0]

hello,

i'm playing around with lvmraid with raidintegrity

i created mirrord lv via:

lvcreate --type raid1 --mirrors 1 --size 4G --name lv_mirror9_4G --raidintegrity y --raidintegritymode journal vg_mirror9

i attached that to a virtual machine, and did some testing via break-raid-odirect.c mentioned at https://bugzilla.kernel.org/show_bug.cgi?id=99171

indeed this introduced inconsistency into the mirror.

indeed, after

# lvchange --syncaction check /dev/mapper/vg_mirror9-lv_mirror9_4G

i see mismatches as expected:

# lvs -a -o +lv_health_status,raid_mismatch_count /dev/mapper/vg_mirror9-lv_mirror9_4G LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert Health Mismatches lv_mirror9_4G vg_mirror9 rwi-aor-m- 4.00g 100.00 mismatches exist 81536

now, if i run --syncaction check again, i would expect those to be repaired, but it doesn't happen.

syncaction repair does not seem to work with integrity feature

# lvchange --syncaction repair /dev/mapper/vg_mirror9-lv_mirror9_4G
Use syncaction check to detect and correct integrity checksum mismatches

in journalctl, i cannot see any checksum mismatches, so my question is what can be done to repair the mirror and how i know, which side of the mirror does contain the correct data.

[teigland]

Without looking at the specific test case you're running...

now, if i run --syncaction check again, i would expect those to be repaired, but it doesn't happen.

If one of the two images is corrupted, then the first "check" will repair it. The raid layer automatically does repair if integrity reports a checksum error on read, so "syncaction check" does repair, and no "syncation repair" command is needed. When one image reports an integrity checksum error, the raid layer reschedules the read to the other image, and if that's successful it writes the good data back to the bad image.