In some cases the Volume Manager Dynamic Multi-Pathing device tree may become corrupted or underlying device information may change causing a DMP node to become inaccessible or undiscoverable by VM.

The following are some possible symptom conditions:
 

First, invalid devices should be removed from Volume Manager. For all devices in error state do the following:

# vxdisk rm
# rm /etc/vx/disk.info
# rm /dev/vx/*dmp/*       // omit any s# element from this command

Example: device named c2t0d1s2 is the issue device:

# vxdisk rmc2t0d1s2
# rm /etc/vx/disk.info
# rm /dev/vx/*dmp/c2t0d1*         // note: s2 is omitted here

Once this has been done, restart vxconfigd.

Please note that while typically production safe, if another pre-existing condition exists on the system that prevents vxconfigd from starting up all current VM configurations will be frozen (no import, deport,grow, shrink, or mounting).

Additionally vxconfigd startup issues maybe come system invasive (to the point of requiring a reboot).  If you have further questions regarding the system impact of restarting vxconfigd in your environment please contact Veritas Support before proceeding.

For systems without any Cluster Server services:
#vxconfigd -k -r reset


For systems running VERITAS Volume Manager in a VERITAS Cluster Server environment

1. Freeze all service groups
# hagrp -freeze

2. Restar tvxconfigd:
# vxconfigd -k -r reset

3. Unfreeze service groups
# hagrp -unfreeze


For systems running Cluster Volume Manager

1. Stop the cluster on the local node:
# hastop -local

2.Restart vxconfigd:
# vxconfigd -k -r reset

3. Start the cluster backup
# hastart