[linux-lvm] Does cmirror can tolerate one faulty PV?

[linux-lvm] Does cmirror can tolerate one faulty PV?

[Eric Ren]

Hi all,

I performed a fault tolerance testing on cmirrored LV in cluster with 
lvm2-2.02.98.

The result really surprises me: a cmirrored LV cannot continue to work 
after disabling
its one leg PV. Is this a known issue? Or am I doing something wrong?

The steps follows:

# clvmd and cmirrord started...
# vgcreate vg1 /dev/sda /dev/sdb
# lvcreate -L1G -n lv1 --mirrorlog mirrored vg1 /dev/sda /dev/sdb
# dd if=/dev/zero of=/dev/vg1/lv1 bs=1k count=1000        ===> OK
# lvchange -an vg1/lv1
# echo 1 > /sys/block/sdb/device/delete   ==> disable one leg

# lvchange -ay --partial vg1/lv1

# dd if=/dev/zero of=/dev/vg1/lv1 bs=1k count=1000   ===> stuck forever
^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C

The following log repeats while dd command is stuck:

"""
Oct 31 13:39:37 sa kernel: [316278.368081] device-mapper: 
dm-log-userspace: [NZ8pEil9] Request timed out: [9/178486] - retrying
Oct 31 13:39:37 sa kernel: [316278.448069] device-mapper: 
dm-log-userspace: [NZ8pEil9] Request timed out: [15/178497] - retrying
Oct 31 13:39:52 sa kernel: [316293.368081] device-mapper: 
dm-log-userspace: [NZ8pEil9] Request timed out: [9/178498] - retrying
....
"""

Thanks,
Eric