Formating disk and deleting everything , for decomission.??

[rveri]

Hi all ,

I wanted to format the drive(Contains OS) and delete the filesytems (vg00) everything that contatins in the system, as the server is de-comissioning.

Can any one tell , how to do that.

1. Any thing from LVM maintenance mode can be done.
2. Mediainit. pvcreate .. ? how ?

Please suggest.

Thanks in advacne ,

Regards,
Veri.

[Pete Randall]

I would not suggest mediainit. Chances are that the disk will end up being unusable. Here is a thread with a rather thorough and interesting discussion on the subject:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x24b53a7b3682d611abdb0090277a778c,00.html


Pete

[Pete Randall]

Oh, now I see! Since it's vg00, you could just run "rm -rf /*". That should do a pretty good job of trashing it. If you have any data LVs in vg00 (which you really shouldn't). I would suggest running lvremove against them first.


Pete

[Geoff Wild]

Well - simple way - re-install HP-UX...

rm -rf * from root is generally sufficient...

Boot a HP-UX install cd - then dd /dev/null over your root disk...

Many others...

Rgds...Geoff

[A. Clay Stephenson]

pvcreate will only write over a small portion of the data. I would do this instead:
(I'll assume that you have blasted everything but vg00 at this point; if not do this same procedure on all the non-vg00 disks as well if all you have done is pvcreate them.)

vgdisplay -v /dev/vg00 and note the disks listed. e.g. /dev/dsk/c0t6d0

If you have mirrors, then break the mirrors using lvreduce and blast that disk first.

To blast a disk, e.g. /dev/dsk/c0t6d0 use the character (rdsk) device, /dev/rdsk/c0t6d0

dd if=/dev/zero bs=1024k of=/dev/rdsk/c0t6d0

[Pete Randall]

If you choose to use the dd suggestion, be sure to add a block size to the command so it doesn't take forever to run:

dd if=/dev/zero bs=1024k of=/dev/rdsk/c0t6d0 bs=1024k


Pete

[Rita C Workman]

You can do what folks have explained to totally hose up that box.

But, you could simply do a cold install of the O/S (as was mentioned) and that will clear up all that was there before, AND leave a box that can readily be turned on by the next person getting the box.
Hey...most old boxes get re-used for testing, training or somebody who just want to play and learn.

Besides....was your data was on external disks...?....then vg00 doesn't have that critical information.

Why destroy....just set it to a clean O/S.
Hey, on Ebay it would be worth more with a clean O/S.

Just a thought,
Rita

[Steven E. Protter]

I do an OS install, include every disk. This will completely wipe out anything that was there.

I consider this superior to mediainit. pvcreate doesn't erase anything.

SEP

[Cheryl Griffin · Accepted solution]

HP will do this for you as well:
http://h20219.www2.hp.com/services/cache/66370-0-0-225-121.aspx
Data Sanitization

[A. Clay Stephenson]

Even an OS install will not overwrite everything though it will make it difficult to retrieve meaningful data. You really need something that writes to every disk sector preferably with random data and multiple passes but for anything except the most rigorous criteria, I would stick with writing NUL's (using /dev/zero, or random data using /dev/urandom) via dd.

[Nguyen Anh Tien]

I suggest you do like these:
1, vgexport
2, pvcreate.
All data will be delete.

[Peyman Javaheri]

Hi,

I would write a script to loop in a combination of mediainit AND dd, where dd writes RANDOM numbers.
Check the following doc, it has an example:

Date: 12/22/00
Document description: wipe classified, sensitive or proprietary data from disks.
Document id: KBRC00005001


regards,
peyman;

[Q4you]

Depending on how much lazy one can be..

If you have multple disks in vg00, remove all labels. Take all disks out, mix them up, put them back in and vanish.

[D Block 2]

veri,

I agree w/ Clay above.. nothing like "dd".
command.

[Ralph Grothe]

Though I think it already has been addressed by the other, more experienced repliers
I would like to add that (sufficient paranoia on your behalf given) I think that a mere rm -rf wouldn't suffice.
I'm sure that people who know the guts of VxFS or HFS and who are furnished with appropiate filesystem debugging tools can well recover stale date, at least partially.

To really be on the safe side you would have to write random bit patterns on the disks in several sequential runs.

There are I guess plenty such tools available.
For instance with linux the following is quite often used.

http://www.llnl.gov/linux/scrub/scrub.html

However, to be dead sure that no data whatsoever is recoverable you would have to physically atomize your hard disks
(usually shredding and burning should suffice).

For an average degree of paranoia
I think booting into maintenance mode
(or from CD) where your rootfs is only in a ramdisk, and then creating new filesystems or running dd 1-3 times over your character devices should should be sufficient.