Once a multipath device is down, I/O to this device is blocked. That can cause hang when collecting a sosreport or scanning/refreshing a disk. An example of a Dead Multipath device is shown below:
360060e80166bac0000016bac000000da dm-49 HITACHI,OPEN-V size=200G features='0' hwhandler='0' wp=rw '-+- policy='round-robin 0' prio=0 status=enabled |- 8:0:0:34 sdiq 135:160 failed faulty running |- 8:0:1:34 sdix 8:272 failed faulty running |- 9:0:1:34 sdjl 8:496 failed faulty running |- 10:0:0:34 sdjs 65:352 failed faulty running |- 10:0:1:34 sdjz 65:464 failed faulty running |- 9:0:0:34 sdje 8:384 failed faulty running |- 11:0:0:34 sdkg 66:320 failed faulty running '- 11:0:1:34 sdkn 66:432 failed faulty running
Removing Dead Multipath Device Online
The dead multipath device can be forcefully removed by deleting the mapping even when processes are operating on the device and those processes are in status “uninterruptible sleep“. Steps to forcefully remove a dead multipath device are as below.
1. remove the device map and replace the table with one that fails all I/O.
# dmsetup remove -f [map name]
For example:
# dmesetup remove -f 360060e80166bac0000016bac000000da
2. remove the multipath device.
# multipath -f [LUN name]
For example:
# multipath -f 360060e80166bac0000016bac000000da
3. Verify the deletion of the dead multipath device.
# multipath -ll