Rotune Live Insertion/Removal of Disks on DS21XX Enclosures - HP-UX Environment

[Alzhy]

Amigos y Amigas! Prospero Año!

I am re-crafting my client's HP-UX Disaster Recovery methods. Our DR is a cold one - meaning everything needs to be restored from scratch. For quick re-establishment of the OS - I am planning to just simply do periodic off-siting of clones of disks (cloned via dd or some other means) - say every month or whenever patches are applied. All our servers use dual DS21XX enclosures as boot subsystems used exclusively for the OS mirrors and swap - leaving free slots.

My questions then are:

1. Are these enclosures built to handle a dozen or more inserts and pulls a year?

2. Is it truly safe - from the running OS perspective, to do live pulls and inserts? I think it is but what safeguards should I observe?

3. Is this a crazy idea? (we're already doing it successfully on Solaris environments). I figure using this approach, I can restore the OS, acclimate it to the target server's I/O config and have the OS up and running in under 30 minutes - ready for non-OS recovery stuff.

Thanks.

[Luk Vandenbussche]

Hi Nelson,

1) It is possible to remove and insert disks in DS21XX, but I have never done it on regular base

2) it is perfectly save from OS perspective, if you export the vg before removing the disk. (vgexport). You will see a SCSI reset in the syslog each time you remove and insert a disk

3) I don't find it a crazy idea

Good Luck

[baiju_3]

Hi ,

My personal suggtion is , HDDs need to be handled carefully .Frequent add/remove may result in damage of disk , since disks are delicate and small jurk can destroy the media . Better to depend on ignite , create a Ignite server and take golden image after you do any major change .Rebuild the new server using this image . Configure a backup ignite server at DR site and push this golden image to DR site .

thx,
bl.

[Pete Randall]

Nelson,

I have been warned by HP over the years that you are better off to keep the disks spinning and avoid shutting them down at all costs. Typically, if there are going to be problems, they arise when the disk spin back up to speed. Since you would be doing the same thing on an individual disk basis, I would advise against it.


Pete

[Alzhy]

Thanks so far for your thoughts/replies.

As I've said we've been doing these approach on our Solaris environments for years. At one time we had a disk at the DR site that missed its rotation and was "dormant but stored at normal data center temperature and humidity" for about 2 years. It booted and spinned just fine.

Pete - I was also "told" of the same by another HP person surprisingly but have not heard this from another vendor that disks need to be kept powered on and spinning. These days.. there are now emerging new array technologies called "MAID" -- called "Massive Array Idle Disks" that are gaining prominence as VTL (virtual tape library) or near-line storage. What it does is it only "power on" disks that are actually needed to - as the technologist behind this new architecture claims - lengthen the service life of the disks.

BTW, since we're still using LVM on our OS disks - a colleague of mine warns that I may have issues should I go the route of "dd" to clone my OS disks on the same server. That will essentially mean my clone disk will have the same VGID as my live OS disks and might corrupt the running OS. The solution I am thinking then will be to just do dd accross the network on to our backup server.