I am planning to migrate my FC looped attached DMX to a brocade switch.
I came upon the attached migration guide stating that the hardware path to the attached target devices will change to reflect the presence of a fabric/switch in the route from the host initiator to the target device and new device file is created for this device.
Now my questions are, what is the best way to correct this change after moving to switch?
Can the change be done online? Can I simply rename the device file back to the old name how?
I am not a SAN administrator and do not perform this finction myself, so take my words with a grain of salt (a big grain at that).
Renaming the device drivers is something that you shuld not do at all costs. More than likely, a vgexport with the old device names before the upgrade and after the brocade switch insertion and reboot, vgimport with the new device names from the edited map file (replace old device files with the new names) is the least painful road you can take.
Also I can not say this enough times : BACKUP ! BACKUP ! BACKUP !....
________________________________ UNIX because I majored in cryptology...
No, you definitely do NOT want to rename any device files. Bad bad idea.
vgexport / vgimport is the way to go.
I would follow these steps:
1. BACKUP everything on the DMX if possible, just in case.
2. Unmount filesystems
3. Deactivate VGs
4. Make a note of the minor number for each /dev/vg*/group file for the effected VGs.
5. Export VGs and create the map files with the following command: vgexport -m /var/tmp/vg??.map -v -s /dev/vg?? (substitute your appropriate VG names for vg??)
6. Do your migration
7. Either reboot or run ioscan and 'insf -e' so you can see the DMX on the new paths.
8. Recreate your VG directory and group files. # mkdir /dev/vg?? # mknod /dev/vg??/group c 64 0x0?0000 (Use your appropriate VG names as above, and the appropriate minor number for the VG).
By using the '-s' option to vgexport and vgimport with the map file, you don't need to know the device names for all the disks in the VGs. The '-s' writes the VGID number to the mapfile when you run vgexport. Then when you run vgimport, the VGID number is read and all disks with a matching VGID number are imported into the VG.
Much much easier than having to keep track of device file names, especially in this situation.
- look at man ioconfig, You can edit the ioconfig file to reflect the old paths (test on a test system first) - vgexport -s -p -m -f ... vgscan / vgimport -s -m (the file from -f must be manually edited, or just use -s and be fine.)
If You're using BCVs, watch out to not mix up the real PVs and the BCVs!
yesterday I stood at the edge. Today I'm one step ahead.
