Understanding AutoRAID 12H module...

[Jonathan Caplette]

Hi all,

I have a K370 with an AutoRAID 12H module, in that module, I have 8 x 9GB and 4 x 18GB for a total of 12 disks. When I do a "ioscan -C disk" I only see 6 disks. When I do a "vgdisplay -v" to see where are my volume group, I see that my volume groups are using 2 disks each, 2 disks that I can see in /dev/dsk.

What I want to know, is how this AutoRAID manage the disks... Does it have docs on that...

I've attached the result of the two command "ioscan" and "vgdisplay".

Why I want to know that, is because I got I/O problems when I do a copy... I found that I/O prob with Glance...

Thanks
Jonathan

[A. Clay Stephenson]

Hi Jonathan:

Your UNIX box knows only about configured LUNS; on the AutoRAID you can't point to disks and say that LUN3 is on this disk. All of the LUN's are distributed across all the disks. This distribution can change 'on the fly' as well as blocks shift from RAID 1/0 to RAID 5 automatically. When you say than Glance shows you problems, do you mean possible disk bottlenecks? Glance can do this with any array - all it knows it that an awful lot of I/O is going to /dev/dsk/c5t6d0 which it thinks is a single disk. In dealing with arrays, things can APPEAR to be bad when they are fine. If you want to know how your array is actually configured, use the arraydsp command. Man arraydsp for details - it will tell you anything you want to know about your 12H.

For an understanding of AutoRAID's go the docs.hp.com and search on 'AutoRAID' -several papers are available.

Regards, Clay

[harry d brown jr]

Here's a quick link:

http://www.systrade.de/mainger/info/daten/handbuch/storage/A3700A_Admin.pdf

[linuxfan]

Hi Jonathan,

Here is a manual for Managing Autoraid 12H
http://www.hp.com/cposupport/manual_set/lpg28365.pdf

Someother useful links
http://www.hp.com/cposupport/manindex/hpsurestor19162_eng_man.html

http://www.hp.com/products1/storage/disk_arrays/midrange/12h/index.html

Like Clay mentioned, you see the disks for the number of LUNs that were created. Mind you, there is a maximum limit of 8 LUNS that can be created, so you have to plan your layout before you start creating your LUNS on the autoraid.

Some of the commonly used commands
arraydsp -i
(gives the serial number of the array)
arraydsp -a
(gives the whole configuration of the array)
arraycfg
(for creating new LUNS)

Most of the array commands are in /opt/hparray/bin, so look at their man pages. Also check out the manual.


-HTH
Ramesh

[Jonathan Caplette]

Hi all...

Thanks for the answers... But here's a little specification on what I've in mind...

The bottleneck on the disks, appears when I'm doing copy of my databases, to take backups... I've already submited that problem, but I was asking for monitoring utilities, now I know exactly where are my bottlenecked disks. But it seems that it's "OK" that Glance show me bottlenecked disks, cause of my AutoRAID. Now I understand a little bit more the concept of the AutoRAID.

But what my problem is, is that 2 weeks ago I install ArcServe 2000 on a NT Server to centralized my NT backups, and I've a project of centralizing my UNIX backup with OmniBack, but before I can implement and get OmniBack, I have to take my backups of UNIX server, and on one of them, I got a Oracle DB of 13GB. Cause with Arcserve I can't take online Oracle backup, that's what Computer Associates told me, I do a "cp" of my database to eliminate the down time of the Oracle DB. When I started to take backup with ArcServe, my DB got 11GB and the copy was taking about 1 hour, but now the DBA's have augmented the size of the DB of 2GB so the DB take now 13GB and my copy takes about 2h45min to complete. It's about 1h45mins more for only 2GB. And I got my bottleneck disks with Glance at the time of the copy. My backup is starting after the "cp" is over.

So I don't see why it could take 1h45mins more time to copy 2Gb more... Do you see what I mean???

Thanks..
Jonathan

[A. Clay Stephenson]

Hi Jonathan,

I've got a better idea for you. Use St. Bernard's 'Open File Manager'. http://www.stbernard.com/products/ofm/products_ofm.asp

The idea is that you down your database, snapshot it, and bring your database back up.
The entire process takes less than a minute. You backup the snapshots and you have a nice safe COLD backup that's very easy to restore.
You have almost all the benefits of Online Backups with the safety of a cold backup. The good news is that OFM works with both ArcServe and OmniBack. You can then forget about how fast your copies or your backups are. The database is back up in a minute so the backup (or copy) speed doesn't matter.

Food for though, Clay

[Jonathan Caplette]

Hi Clay,

is that OFM from St-Bernard, using the ArcServe Client Agent??? They don't tell on the web Site...

[A. Clay Stephenson]

Hi again Jonathan,

On their webpage, look for Agents Update and that will show you the versions of ArcServer supported. However, after reading your posting again, I realize that you don't need OFM since the Oracle data resides on your UNIX box. If you have OnlineJFS, you can do a snapshot mount of each filesystem. The idea is exactly the same as the OFM method: Shutdown the database, snapshot mount the Oracle filesystems, startup the database - should be less than 1 minute. You then backup the snapshots. If you don't have OnlineJFS get it; it is useful for much more than snapshots.
man mount_vxfs for details on the snapshot process. Typically, the snapshot buffer needs to be no more than about 15% (and that is generous) the size of the main filesystem.

The answer why only 2 GB could make your copies so much longer is that you almost certainly crossed the threshhold of RAID 0/1 (where the AutoRAID is quite fast) to RAID 5 (where it can be a real dog).

Regards, Clay