Ignite Post install script

[catastro]

Hi all. I just faced a new issue for me. I had to recover from ignite an Itanium server, and all the device files for the disks were missed (as ignite does not recover /dev).
I have several FC disks, also the booting disk, and we have been making changes in all the disks adding and removving several times. When I recovered the ignite image, I got that all the disk fikes were changed.
I know I can fix this up with mknod.
The question is, is there any way to tell ignite to recover /dev (in order not to have all the disks configuration messed up) or I have to use a post install script?
If I have to make a post install script, could you please provide me with an example for doing it?
Thanks in advance

[Matti_Kurkela · Accepted solution]

Depending on the complexity of your FC environment, trying to keep the device files identical may be a lot of effort for a very limited benefit. You might want to learn to live with the fact that the device files may change. Unless you have raw databases on /dev/rdsk/* devices, it might actually be easier that way.

Managing disk devices with mknod/rm instead of mksf/insf/rmsf is not recommended, because it will cause the /dev directory to get out of sync with /etc/ioconfig and /stand/ioconfig.

This may cause problems with future configuration changes: when adding a new disk, insf will look at kernel configuration and ioconfig files to create the correct device entry, and then will create a name for it using the standard scheme. But if you already had used the standard name for your mknod-created custom device files, the new standard device file may overwrite your custom one... making your existing device name to point to a different device. Confusion ensues.

A LVM volume group can be easily adapted to the new device names by exporting it (IMPORTANT: using the -s option of vgexport), then re-importing it:

The traditional (11.23 or older) procedure:

ls -l /dev/vg/group
vgexport -v -s -m vg.map vg
mkdir /dev/vg
mknod /dev/vg/group c 64 0xHH0000
(replace HH with the appropriate hex digits seen with the above "ls -l" command)
vgimport -v -s -m vg.map vg

In the recent 11.31 releases, the mkdir/mknod commands can be omitted, as the vgimport will be able to create the directory and the group device file automatically if it does not exist. In 11.31, you will also want to use the -N option with vgimport, to get the full benefit of the new Agile I/O system.

When you vgimport a volume group using the -s option, the vgimport command will scan all the disks visible to the system, and automatically find the disks belonging to the VG using the VGID stored in the map file. If you have multiple paths to your FC disks, this will automatically configure the first path for each PV as the primary one, and all the other paths leading to the same PV as alternate paths.

For raw databases, you might want to use /dev/vg/rlvol* devices to manage them using the LVM too, or perhaps something like Oracle ASM to deal with the device name changes in an automated way.

MK

[catastro]

OK, thanks Matti.
We will think in what to do.