Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

5207

December 17th, 2009 03:00

Can I wipe a Symmetrix device without presenting it to a host?

We'vre had a whole batch of LUNs de-allocated fromm some servers, and they've been unmapped and unmasked. We're wanting to wipe the data on them, prior to reuse. I'm aware EMC support can wipe a device, but I'm ... not entirely sure if I can do it, using Solutions Enabler.

I can map all these devices to a proxy host, format them, and move on (formatting is ok, we don't need secure wipes or anything), but before starting doing this with a couple of hundred devices, thought I'd see if there was a better way?

April 20th, 2010 03:00

Thanks for the comments here, appreciate the activity. I created an "EMC internal" request a few months back, which didn't go far, but just added that the competition can do it, CLARiiON can do it, and that customers are requesting this via enhancement request. I'll champion that request as the Lead Specialsit for EMEA on Symmetrix, will try to update this thread as we go on.....

Alasdair, EMC Lead Specialist, CSS EMEA

2 Intern

 • 

20.4K Posts

December 17th, 2009 05:00

not better way i am afraid,  symcli does not offer any erase/vtoc options.

53 Posts

December 18th, 2009 02:00

I thought that might be the case. I was just hoping. Mostly having vague inconveniences with disk signatures - am wondering if a recent problem with disk signatures was related to reusing a decommissioned BCV copy of the same LUN.

I'll have to figure out if I'm better off having a 'template' BCV (that's been 'dd'ed with zeros) to 'restore' onto the LUNs I want to clear, or whether I'm stuck with the 'format on a proxy, relabel' approach.

2 Intern

 • 

20.4K Posts

December 18th, 2009 03:00

create a brand new device so you know it does not have any information and clone (with copy/pre-copy options) it to devices you are trying to wipe. You can define 15 clone sessions (1 source device and 15 target devices) and run 8 copy sessions concurrently. I think that would be faster then presenting each and every device and running DD against it.

53 Posts

December 18th, 2009 05:00

I was under the impression that even brand new devices don't explicitly get 'zeroed' though? I really want to avoid 'accidentally' ending up with disk signature clashing, as that causes us all manner of grief it seems.

2 Intern

 • 

1.3K Posts

December 20th, 2009 01:00

what is the operating system on the planned proxy host? Sounds like Windows as you scared of disk signatures..

53 Posts

December 21st, 2009 01:00

We are indeed entirely windows. Coming from a Unix shop, I was pretty laid back, but having done some digging have found some potential nastiness with signature clashes and whatnot. Mosty though, we would like to avoid having files showing up on disks we've re-used (we've had a few incidences of rebuilding a meta in the same 'shape' as the original, and a bit of confusion when the disk appear to already have data on it when allocated to a host)    

2 Posts

December 21st, 2009 09:00

We have a similar issue. UNIX servers with Veritas Volume Manager get decommissioned, but no one does a vxdiskunsetup. Then the disk gets presented to a Windows box running VVM and that server gets really confused. We're looking for a way to erase/format the disk prior to giving it to "any" server to prevent this confusion. (We are also trying to improve the decommission process, but thought it would be easier to fix the problem on the storage side). Any ideas or suggestions would be helpful.

108 Posts

December 29th, 2009 01:00

I heard of a symmetrix licensed product symerasure.. I am not sure if that can help on this.

66 Posts

December 30th, 2009 08:00

You can use the devs to create a meta and then disolve it. The process of creating a new meta deletes the old data. I just used this to clean up some clone targets.

2 Intern

 • 

20.4K Posts

December 31st, 2009 12:00

not sure about that. I just dissolved a meta presented to a Sun box, build another meta (same meta head) and presented it to a Linux box. This Linux box could not format that disk, fdisk was still showing sun label. Had to DD the volume and then format worked.

45 Posts

January 4th, 2010 06:00

This is correct.  If you destroy a meta, then create a new meta of the same size with the same members, the data will all still be there.  The only way I've gotten this to work is to change the meta head or present the original meta members to create different metas.

The best way to keep it clean (assuming we don't eventually get a symcli erasure tool) prior to reallocating it is to have the SA's clean it up as part of their deallocation/decommission process.

2 Posts

January 4th, 2010 06:00

you can delete the hyper and recreate new hypers that might clear all existing data on decommisioned hyper luns.

2 Posts

January 4th, 2010 07:00

I understand that deleting and recreating a meta "might" resolve the problem if I change the meta head around. However, that doesn't help with individual devices. It sounds like I can either clone a known empty device to the used device, or delete and recreate the device. We are currently in change freeze mode, so I might not get a chance to try anything until Feb. 1. In the meantime, I'll also look at getting decommission processes changed.

53 Posts

January 5th, 2010 05:00

Dissolving and reforming a meta doesn't 'clear' the data on it. Neither does deleting/recreating a hyper.It just 'seems' to, if you have a different geometry next time around - if the OS can't find the file allocation table/MBR or superblock, the file system might as well be 'blank'.

'rotating' the meta so a different meta member is 'head' means you've got a volume full of incomprehensible data for the OS, and that's _almost_ as good.

Seems there's no magic bullet to do what I want though, which is re-initialise a particular volume 'storage side'.

I think I'll knock together some script that does almost that though - should be fairly easy to create a 'template' BCV and splat it over a defined list of volumes, and only slightly more complicated to accomplish this with a meta device (e.g. dissolve meta, trash data, reassemble meta).

Thanks for all the feedback so far though.

No Events found!

Top