How to erase data on disks

[Danesh Qureshi]

I have 2 Alpha server ES45 systems running Tru64 UNIX v5.1A on them. I wish to erase or all data from all the disks including the operating system disk.

I have attached the output the of df -k, showfdmn and hwmgr.

Can anyone tell me how I would erase all the data on each system?

[Steven Schweda]

> [...] I wish to erase [...]

Why? Erase how well? (So that the next
normal user can't see the data, or so that
the government can't see the data, or ...?)

[Danesh Qureshi]

The disks hold sensitive data on them therefore I require to safely erase all data.

Any advice appreciated.

[Hein van den Heuvel · Accepted solution]

I wouldn't bother with Tru64 to do this.
Don't send in a boy to do a mans job!

Go straight for the data source... the HSG80.

Do not trust ANYTHING, not any drive.
It is way to easy for a drive to have been in a failed-set and back, or for there to have been a clone or mirror or whatever.

Find yourself an (online) userguide, or bravely run DILX from a terminal emulator connected to the serial port.
Use a most awk/vi generates list of commands to hit all disks.

For the (few) local disks, just:
dd if=/dev/zero /of=c bs=1024k

I'd recommend to hit the system disk last :-)

Good luck,
Hein.

[Rob Leadbeater]

Hi,

Do you have a ProLiant server into which the hard drives from the ES45 and HSG80 can be installed ?

If so, then take a look at DBAN (http://www.dban.org)

You can use this to create a bootable CD, which will wipe hard drives to various military standards. Unfortunately there isn't an Alpha version, hence why you'd need to move the drives into an Intel box...

Hope this helps,

Regards,

Rob

[Khairy]

hi danesh,

get the dec-office serial cable which comes with with hsg storage and hook it to serial port of your pc/notebook and attach it to the hsg80 controller.

1) List disks in storage
hsg_80> show disk
eg:
name type port targ lun used by
-------------------------------------------
disk10000 disk 1 0 0 M1
disk10100 disk 1 1 0 R1
disk10200 disk 1 2 0 R1
disk10300 disk 1 3 0 M1
disk10400 disk 1 4 0 R1

2) the above will also show you 'used by'. Usually it indicate by mirrorset (M) or raidset (R). Type the following to list out units.
hsg_80> show unit
eg:
Lun uses
===========
D0 M1
D1 R2

3) Now delete every unit. Eg:
hsg_80> delete D0
hsg_80> delete D1

4) Now delete the mirror and raidset
hsg_80> delete M1
hsg_80> delete R2

5) create a stripeset. Create a new storageset and initialise will destroy metadata structure in all disks.
hsg_80> add stripeset S1 disk10000 disk10100 disk10200 disk10300 disk10400
hsg_80> init S1

For your OS disk, you can dd them. You can do this by boot the server with OS cd and go to single user mode.

Don't forget to assign some points if this answer your questions, thanks!

Rgds

[Liviu I.]

Hi Danesh ,

I'd boot up a cd-rom with Debian-Linux (alpha port) and then use several times :

dd if=/dev/zero /of=c bs=1024k

or

dd if=/dev/urandom /of=c bs=1024k

HTH,
Liviu

[Hein van den Heuvel]

Liviu,

That' a fine first step, but probabyl; not good enough.
1) how will you reach the drives which once were in the configuration, but no longer are. This may have happened transparently in the case of a raid-5 drive failing and being bumped out if the production set into the failed set.

It's slowish because the bits have to go over the wire.

I suspect that the 'writing multiple times' as a securtiry precaution against extreme recovery techniques only works when different values, typically multiple patterns, are written really shaking the magneto bits on the platters. If that is a concern then the drives should be physically destoyed (drillpress ? :-).


2) Admittedly I have not tried under debian but under on server Unixen /dev/urandom is very slow. Not practical. You would need to fill a block-range first, and then clone
that.

Khairy,

Good expansion on my quick description.
You'd have to make sure to grab all disks to re-present them to the OS.
I believe that INIT as a strie is quick, and thus not good enough. Only meta data is involved. Init as RAID-1 or RAID-5 would write to all data blocks.

All,

I knwo from practical experience that clearing the metadata alone is NOT enough. All too often I have found back a a useable disklabel from some random prior usage.

fwiw,
Hein.

[Danesh Qureshi]

I am familiae with connecting my PC to the HSG80 controllers using a serial cable.

I've read the HSG80 controller user guide to familiarise how I use DILX.

So once I get a CLI prompt all I need to is type "run dilx" which you take me through a series of questions?

[Hein van den Heuvel]

Danesh,

I haven't actually done this myself.
No access to an HSG80 now.
But yes, that's what I think should happen.
You may want to use the autoconfigure to let it find all.
Or you can do a SHOW DISK first,
capture on screen.
Reformat as needed to become input for the device list.

Let us know how it goes !?
Hein.