Wrong Physical Devices on server... How to fix?

[TWBSupport]

OK - here's a dilly...
We have 2 rp2470's in a cabinet, each with 2 - 18 G drives & 12 - external 72G hard drives. When I first set these systems up, I had a portion of the hard drives allocated to 1 rp2470 cpu (primary)& then had the remaining hard drives allocated to the other rp2470 (secondary). Everything was great & then I went on vacation... What I'm being told is the secondary rp2470 was all of sudden thinking it was the primary rp2470. No one knows why, it just started happening.. yeah, right.. My replacement reset the hostname & ip address to the correct hostname & ip address after he was notified aprox 4 days after this must have started happening. Everyone thought it was fixed & never bothered letting me know, until now.. 1 1/2 month later. It looks as though someone copied over the /etc directory FROM the primary server to the secondary server. Everything on the secondary server is now an exact duplicate of the primary. I can run a bdf & the results are exactly the same as the primary server, if I do a ls -l the results are the original layout of filesystems. Can I just comment out the incorrect mnttab file entries from the primary server & re-allocate the correct physical disks to the secondary server? These are new servers & aren't in production yet, but a ton of work has been done on the primary server, with the secondary just starting to be worked on now, hence the problem was discovered. By the way, no one & I mean no one NOW has root pwd on these systems, but myself! I have tried setmnt & works up until I reboot & then everything gets put back to the wrong way (primary server duplicate).

[Paulo A G Fessel · Accepted solution]

If somebody copied /etc from primary host to secondary host (YIIKES!), then /etc/lvmtab of the secondary host is the same of primary's - and so, the system keeps seeing the wrong disks.

The solution would be recreating /etc/lvmtab from scratch on the secondary server. This can be done with vgscan after renaming the original /etc/lvmtab in two steps:

1) issue "vgscan -pav". This will tell you the exact disks the secondary server is able to see. If you think that incorrect disks are detected then you'll have to vgimport the VG's manually, specifying each disk for each VG.

2) if you're really sure that the result of "vgscan -pav" is actually correct, then issue a "vgscan -av". This will recreate /etc/lvmtab, hopefully with the correct disks.

3) for file system entries, the file you must change is /etc/fstab, not /etc/mnttab. /etc/mnttab only stores temporary information about the server's file systems. This can also be accomplished with SAM.

HTH
Paulo Fessel

[Tim Sanko]

You must have my old junior admin working with/against you!!!

I have a preference for doing a lvremove/vgremove from the secondary server and ignoring the problem as you can make it "GO AWAY" really easily.

Then you can either remove the /dev/vgXX of the offending filesystem, and remake the directory,/dev/vgXX and the node with mknod, and/or pvcreate and then the vgcreate and then lvcreate the appropriate volumes and run a newfs on the drive...

Tim

[Tim Sanko]

I hope that you also have ignited the server originally. if you did that is the fastest way to ground Zero.

HP IGNITE will backup/recover VGXX for you and you can change the size of the vg00 volumes and allow the rest to be recovered. If you have an IGNITE TAPE then this is a short answer. Stop reading here.

Otherwise download HP IGNITE for your 11.X server and install it (using swinstall utility)
Hopefully you have an attatched tape drive.

if you do have a working tape drive (DDS'2/3/4') then run "make_tape_recovery -Av"

Go out to dinner as this process will run for quite a while.

This will create the archive for you of the vg00 and all VGs you include.

I do this before any significant system changes to any of my servers. (almost any change can be significant.)


Tim

[TWBSupport]

Thanks to both of you. I ended up doing what Paul suggested. Because that was the only way I could figure out how to do it also.

Even though I now know that it would have been much easier by following Tim's suggestions...

No ignite, since this was still a development server & I knew that I would be applying more patches before it became production.

Ignite is going to be done next week. Lesson learned.

Thanks again

[Paulo A G Fessel]

I'm glad it worked. Don't forget to assign the points both to my answer and Tim's.

[]'s
Paulo Fessel

[TWBSupport]

I keep trying to award you both 10 points, but the site will not update. I'll keep trying..
:o)