command to manually break a raid1 disk array for patches?

[Peter Berger, Jr.]

Is there a command-line utility that I can run to break a 2 disk RAID1 array (operating system partition)?
The reason I ask is that for monthly windows server 2003 patching outages we: shutdown the server; pull disk0 out of the machine; power the machine on; patch the machine; wait a few days to make sure patches don't crash server; and then push disk0 back into array to rebuild raid array.
this helps us in-case of an event where we cannot boot disk1 (the OS) because of a patch issue -- we can simply pull disk1 out and only boot from disk0 (pre-patch environment).

I was wondering if I can run a array utility command (via batch script) to break the mirror without physically powering off and pulling disk #0? my plan is to then run another command to "add" disk0 back into the raid1 array when all is OK.

any disk guru's out there have a better tried and true method of patch-problem-prevention? or do you think this will work? any 411/help would be great. thanks...

[Steven Clementi]

You are not actually breaking the mirror. The removal of the drive is read as a disk failure, not a mirror break.

What you are asking would actually... convert the the array to RAID 0... and then to RAID 1 when you are done. You would effectively lose the data on that disk.

Removing a drive IS an effective way of being able to restore an environment to working order. perhaps you can keep a set of spare drives around... that you just.. shut the server, pull a drive, boot it up and insert a new drive after the OS is up... at least you do not have to re-insert the drive days/weeks later.

Or perhaps you can have a QA / Dev environment where you actually "test" the patches first?

Steven

[Peter Berger, Jr.]

thanks for the quick reply. although I'd love to get management to pony up the $ for a dev/test -- with this recession it just won't happen. :(

but you are right -- i would want to initiate a disk failure on disk0, not break the mirror. I guess there's no way to manually initiate a failure on disk0 via command-line?

if not -- how about a command to quiesce the data in buffers to all disks in a particular array -- so I can simply pull disk0 out while the system is still running and live? the maroon/hot drives should be able to handle this, right? i just don't want the os to be writing data at the same time i pull the drive and I corrupt the OS.

my ultimate goal is to utilize WSUS to automate the patches/process, but be safe too by being able to go back to the pre-patch environment.

[Peter Berger, Jr.]

by pulling out Drive0 (RAID1) when the system is running (Windows 2003 SP2 fully patched), will it corrupt data on Drive0? My colleagues state that pulling a drive out while the system is running will corrupt data. they'd rather shut the machines down, pull the drive, power it on, patch it, reboot it.

I am of the thought that the Smart SCSI Array controller is "smart" enough in the way that it works so that it won't cause problems as it will continue to write to Drive1 just fine and that anything that was written to Drive0 was already write-committed by the SCSI controller. There are no applications installed on C:\ -- just the OS.

Is pulling the drive OK, or should we shut machine down, pull drive, power on?

thanks...

[Uwe Zessin · Accepted solution]

Many array controllers provide a "writeback cache" that is independed from the operating system. If you just pull a disk drive while the system is running, the WBC is not flushed and the data on the disk drive IS NOT CRASH-CONSISTENT. Indeed, this can make the data look 'corrupted'.