Boot disk mirroring on DL360 G2

[Al_56]

Hi Guru's,

I'm new to this board, and l have a question regarding disk mirroring on a DL360 G2 to pose to you.

I originally configured the server's RAID to have the 2 disks mirrored (RAID 1) which seems to be working fine.

I would like to create a backup boot disk in case the server ever gets hacked (and then back up the data regularely using other methods). What is the best way of doing this?

Could l attach an external SCSI disk to the external SCSI interface, and then use dd (this is a Linux 7.3 install) to mirror the drive (disconnecting the external device afterwards to avoid it getting hacked as well)?

Could l remove one of the 2 internal drives to emulate a disk failure, and then insert a new disk, keeping the removed one as my backup? Is this very risky?

Your thoughts would be greatly appreciated.

[Martin Breidenbach]

You can do this:

1. shut down system
2. remove one drive and put it away as backup
3. turn on system
4. insert new drive
5. automatic rebuild should start

The SMART RAID controllers store the array configuration on the drives and on the controller. So the controller can read the configuration from the drives.

[Randy Weis]

Part two:
To actually use the disk you removed, you would need to remove the corrupted or failed disks from your server, and put the saved disk back into the same slot you removed it from. I advise using Disk 0 as your saved disk. Follow the instructions above, removing Disk 0. If you ever need the disk, put it back into slot 0.

[Gert Luyten]

Part 2b): since the DL360G2 has a built-in SA-5i controller it doesn't matter where you put the disk in, the Cyborg chip will recognize it

Part 3) this is not supported by HP :-))

[Al_56]

Thanks for all of the input so far. I would like to go over my plan with you to see if l understood correctly:

1) Shut down server cleanly, remove a disk (let's say disk 0), ensure to note that it was disk 0, and then place it in safe storage (offsite location).

2) Boot system, and wait for RAID software to note that a disk is dead (l assume it will identify the error, and ask me for a new disk. Is this correct?).

3) Install new disk and wait for sync to complete successfully before returning server to active duty.

4) If servers disks ever become unusable (serious security breach, corrupted drives, etc), perform steps 2 and 3, installing the backup drive before restarting the server.

Does anyone know how long it would take to sync an 18 gig drive (step 3)?

I can imagine that this is not supported by HP. Is it a risky procedure? Has anyone done this? If not, what full mirroring/backup process to you use?

Thanks again for your input so far.

Alan.

[Al_56]

I wish to implement this procedure soon, and would like to know if anyone may have some answers to the following questions from their experience.

Does anyone know how long it would take to sync an 18 gig drive (from step 3)?

Is it a risky procedure? Has anyone done this? If not, what full mirroring/backup process to you use?

Thanks in advance,

Alan.

[Gert Luyten]

We have done this many times before as it's the easiest way to test something and not having to restore the lot. You have to assume that nothing can go wrong, trust me.
I've updated your plan of action with the following info:

1) Shut down server cleanly, remove a disk (let's say disk 1), ensure to note that it was disk 1, and then place it in safe storage (offsite location).

2) Boot system, and wait for RAID controller to detect that disk 1 is removed. It will show an error and you have to press F1 or F2 to continue with a broken logical drive or to disable it. Read the error message!!

3) Boot the server without disk 1 so the controller can update it's status and then insert the new one. Check with ACU the status of the rebuild.

An 18GB mirror should take less than 4 hours for the rebuild to complete depending on the load on the server.

[Mark Smith_20]

Al,

Gert has given you some good information. The only thing that I want to stress is that it IS important to label which drive you pulled out (ID0 or ID1). If you ever have an array failure that would cause you to put the "backup" disk back into the machine you WILL HAVE TO PUT IT IN ITS ORIGINAL ID!

Gert stated that since this machine has a 5i controller that you can pretty much mix up the drives and the controller won't care. This is partially correct. The controller can only handle a single failure. Putting the backup drive back into the machine in the NON-original slot will cause a DOUBLE failure and the array will NOT initialize. This is because the "backup" drive is in a failure state (you install it in the machine as 1 drive of a RAID1 set and it expects to see the other drive) AND you have performed a migration (moved drive IDs or machines).

With this in mind - this is a pretty safe process and you should not have any other problems.

Let me know if you have any other questions.

[Al_56]

Thanks very much for your help Gert and Mark. I will be very sure to follow your instructions to the letter.

Mark, your comments have raised a question regarding the spare disk l was going to use.

The spare disks l have are from the same server, and still have the OS on them (they were pulled because the server was hacked).

I'm concerned that if l use one of them as the replacement disk without wiping it first, it may confuse the RAID, as to which disk it will use as the master to resync from.

Is the RAID controller smart enough to know to ignore the contents of the replacement drive (disk 1) in step three of the instructions?

Is this an issue? If so, how can l wipe them before using them? (l don't have another SCSI box at that site to use, nor an external enclosure).

Thanks again for your help.

Al

[Gert Luyten]

If you install the spare disk AFTER you have booted the server with only one orginal disk, this is not a problem since the Array/LD is already initialized.

Mark, you're absolutely correct and thx for the correction.

Gert.