In a CVM environment, adding a relabelled LUN to a shared disk group cause the I/O requests to fail until the LUN fails and disables the filesystem. (3979198)
In a CVM environment, when a relabelled LUN is added to a shared disk group, following error is logged in the
/var/adm/messages file for the same LUN until the LUN fails and disables the filesystem.
vxvm:vxconfigd: [ID 702911 daemon.notice] V-5-1-5249 ddl_detect_nonscsi :
find_physical_path() fails for /dev/rdsk/<disk_name>
When a new LUN is added to a Solaris system dynamically, the operating system detects the device and creates an entry for the LUN. After relabelling the disk on one node, it may happen that the other nodes do not refresh the physical paths corresponding to the relabelled LUN. As a result, DMP continues to use the stale paths for the LUN on that node and cause the I/O requests to fail until the LUN fails and disables the filesystem.
Related information
Workaround:
If the disk size is less than 2TB, format the disk as SMI on each node. For larger disk size, format disk as EFI on each node. Ensure that each node has the same disk label format. That is, all nodes either have only SMI labels or only EFI labels.