Drive Path renumbering

[C R S]

running 11.23. I'm looking to a step by step guide to recover after SAN maintenance that will cause the Device paths attached to the SAN to be renumbered.

by recover, i mean, reconfiguring the lvmconf's to use the new Physical paths and alternate paths.

-CRS

[James R. Ferguson · Accepted solution]

Hi:

Since this is after-the-fact, you could recreate the '/etc/lvmtab' from scratch with:

# mv /etc/lvmtab /etc/lvmtab.old
# vgscan -av
# vgchange -a y
# lvlnboot -R

Then, re-adjust your primary and alternate links. You can ther demote ('vgreduce') primary links to secondary ones; and promote ('vgextend') demoted primary links back in as secondary ones.

Another way which is faster with ar large number of devices in a volume group is to 'vgexport' and 'vgimport' the volume group using a file containing the path information:

# vgexport -m vgNN.map -v -f vgNN.paths vgNN

...then [reorder the device files in the 'vgnn.paths' file to distribute your primary and alternate links the way you want and use that editted file to do:

# vgimport -m vgNN.map -v -f vgNN.paths vgNN

Of course, the first method does not require denying access to the volume group, whereas the second method, using 'vgexport' and 'vgimport' does, since a 'vgchange' must be performed to deactivate and then re-activate the volume group.

Regards!

...JRF...

[C R S]

Thank you,
That is exactly what I was looking for. I had done it in the past a long time ago and that how I had done it... using vgscan and all that.


Thanks,

CRS

[C R S]

one more question,
I found a procedure for 10.20 that is supposed to change the controller numbers back to what they originally were... would this still be effective with 11.23?

> 1. ioscan -kf | grep -e INTERFACE -e DEVICE | grep -v target | awk
> '{printf"%s %s %s\n",$3,$1,$2}' > /tmp/infile
>
> 2. vi the /tmp/infile and change the "ext_bus 4" (as an
> example, number
> 4 in this case to 7 if you now want c7t0d0 instead of c4t0d0)
Continue
> to do this for other "ext_bus X" to arrange the controller numbers as
> you see fit.
>
> 3. mv /tmp/infile to /stand/infile
>
> 4. mv /stand/ioconfig to /stand/ioconfig.sav
>
> 5. mv /etc/ioconfig to /etc/ioconfig.sav
>
> 6. shutdown -ry 0
>
> 7. hpux -is (single user mode)
>
> 8. cd /stand
>
> 9. /sbin/ioinit -c
>
> 10. exit and continue booting by pressing Ctrl-D
>
> 11. init s
>
> 12. /sbin/ioinit -f /stand/infile -r

[Torsten.]

The solution provided by JRF is much, much better than the steps listed in your second post. Please don't consider these steps as a solution.

Keep it simple ...

[James R. Ferguson]

Hi (again):

I've used the procedure you cited to change tape instance numbers, but as Torsten notes, "why?" go to this length.

At the least, changing instance numbers requires reboots.

The procedure I first cited can be done while to volume group is active or, at most, requires a short period of deactivation. It's also very "safe".

Regards!

...JRF...

[Torsten.]

Back to the question.

"SAN maintenance" means you changed something (e.g. connect the storage to another port) and get new hardware pathes?

I think so.

With the vgexport/import procedure you will adjust the systems setup to the current hardware/wire layout, but with the other procedure you will force a match between the existing layout and the old config (in other words, you will "fake" the current layout to look just like the previous). Some background is needed to understand all this (especially all the cons), but I would not work on the wrong end of this line.


BTW, all this is history with 11.31 ...