my system is RH EL 5.8 with kernel 2.6.18-308.8.2.el5.
I'm doing some tests for online moving of several multipath LUNs from
a storage array to another one, to understand pros&cons with other
I'm encountering a strange hang I think could be a bug itself, apart
from the approach itself that can be considered risky.
But know my question is only related to pvmove command itself in a
scenario where LVM activation filters are enabled.
I don't know if in my problem some components form rhcs take place or
if it would happen independently.
But in logs I don't see anything from cluster daemons so I presume
they are not related.
- storage arrays SA1 and SA2
- VolGroup00 on internal disk
- PV1 is multipath device on SA1
- PV2 is multipath device on SA2
- VG_ORACS_APPL volume group composed by PV1 and PV2 and all occupied
PEs are on PV1
- VG_ORACS_APPL has LV_ORACS_APPL logical volume and file system
mounted on it (all on PV1)
- I want to pvmove from PV1 to PV2
- LVM is configured with activation filters enabled
volume_list = [ "VolGroup00", "@iclnacs1" ]
I want to temporarily bypass activation filter adding my VG, but I
mistype (I miss the final L in vg name)
I try the pvmove command
time pvmove -i 30 /dev/mapper/mpacsappl /dev/mapper/mpnacsappl
and due to my error I get this
Not activating VG_ORACS_APPL/pvmove0 since it does not pass activation filter.
Temporary pvmove mirror activation failed.
Unable to deactivate open VG_ORACS_APPL-pvmove0 (253:20)
Unable to deactivate open VG_ORACS_APPL-LV_ORACS_APPL (253:9)
libdevmapper exiting with 1 device(s) still suspended.
and the command exits
But now as a side effect, all LVM operations hang; also the ones
related to other Volume Groups.
I cannot run successfully neither a "sync" command or "shutdown -h" command.
I reproduced two times with different VGs
for example commands such as
strace -p of a command related to LVM hangs
relevant lines in messages when this occurring are "only" related to