It seems that checkerloop isn't just checking the your multipathed devices. It's checking all your devices. Wouldn't it be more sensible to find out which devices make up part of a multipathed device, and only check those? Currently, when I start up multipathd, I getting this printed out: Unknown switch: (null) multipath v0.4.4 (16/04, 2005) Usage: /sbin/multipath [-v level] [-d] [-l] [-S] [-p failover|multibus|group_by_serial|group_by_prio] [device] -v level verbosty level 0 no output 1 print created devmap names only 2 default verbosity 3 print debug information -d dry run, do not create or update devmaps -l list the current multipath topology -F flush all multipath device maps -p policy force all maps to specified policy : failover 1 path per priority group multibus all paths in 1 priority group group_by_serial 1 priority group per serial group_by_prio 1 priority group per priority lvl group_by_node_name 1 priority group per target node device limit scope to the device's multipath (udev-style $DEVNAME reference, eg /dev/sdb or major:minor or a device map name) because checkerloop is trying to check the status of /dev/hda, which is not multipathed. -Ben