Operating System - HP-UX
1753792 Members
7075 Online
108799 Solutions
New Discussion юеВ

Re: Huge EMC disks creation

 
STC ITRC user_1
Occasional Advisor

Huge EMC disks creation

I have 4 EMC disks each with size 133GB and I want to distribute that into three logical volumes:
/u01 100GB
/u02 200GB
/u03 200GB
But to do that I have to change many parameters such as max_ph_extents to 65535 ...etc, so I need the recommended parameters that can do that with a good performance.

~Abdullah
Champion
7 REPLIES 7
Rita C Workman
Honored Contributor

Re: Huge EMC disks creation

Well first when you had these EMC disks added to your disk array....how did you have them 'sliced' up? or presented? Generally, the CE will ask how do you want these disks done - and will create the binfile accordingly. Unless you the customer were expected to create the binfile with the new disks. Which can be done if your experienced in creating binfiles....

If your not certain, maybe you should contact your Tech Support person from EMC, and possibly they can tell you.

Making the logical volumes the size you mention is easy enough....once your disks are set up as you want in the binfile & luns created.

Regards,
Rita
Geoff Wild
Honored Contributor

Re: Huge EMC disks creation

We recently upgraded to DMX 1000's and went with LUN's of 33GB - striped across 112 x 146GB mirrored physical disks.

So, as long as EMC has striped them, you will have good performance - unless you were mislead and they used "parity RAID". Performance with parity raid is awful.

For 132 GB LUNS, you will need to set your PE Size to 32

vgcreate -s 32 /dev/dsk/cXtXdX

With LUNS your size, you probably can keep the default of MAX PV's at 16....


Rgds...Geoff
Proverbs 3:5,6 Trust in the Lord with all your heart and lean not on your own understanding; in all your ways acknowledge him, and he will make all your paths straight.
Todd McDaniel_1
Honored Contributor

Re: Huge EMC disks creation

Abdullah,

I am of the school of thought to break those large disks into Virtual LUNS of a manageable size, anywhere from 9GB to 30GB in size.

If this is possible, then you merely have to assign disks based on ctd numbers till you get your desired FS size. also, you dont have to jump thru hoops to change standard configurations for your diks. In addition, the EMC tech can Raid-s for striping on the H/w side, if you desire to stripe. OR they can setup a H/w mirror to aid with bad disk failures. Not sure if you have the disks for mirroring though.

While you certainly can do the whole disk configuration, it seems rather self-defeating to me to do so, when that is a great functionality of a Frame.


However, if you stick with your current configuration wishes. I would suggest increasing the extent size to 64 maybe 128, depending on your data. also, I would recommend setting max PV to 64 at least. You never know when you might need to add disks.

Also, I didnt see where you were planning on mirroring or striping these disks. What is your plan there?
Unix, the other white meat.
George Petrides_1
Honored Contributor

Re: Huge EMC disks creation

I agree with Daniel. You'll also have problems backing up these filesystems with many backup software unless you specify special configuration and it's maintainance will make your life miserable. I wouldn't go over 50-60 GB per filesystem.
George
STC ITRC user_1
Occasional Advisor

Re: Huge EMC disks creation

Thanks all,

Backup is not a problem because our ATL will take care of that.

~Abdulla
Champion
DIPAK KUMAR ROY
Frequent Advisor

Re: Huge EMC disks creation

I don't think 200GB filesystem will create any performance issue. You just need to make sure that taking backup is not a problem. Are you using these filesystems for database dumps ?

Thanks
Todd McDaniel_1
Honored Contributor

Re: Huge EMC disks creation

Dipak, I beg to differ on the performance to a small degree...

A single 200GB FS on one disk is not going to respond as well as 20~ 9GB virtual LUNS would... Seek time would be lower and reads faster.

IF you raided the LUNS, then your writes might be a bit slower, but reads would be blazing fast!


Sure you can make a 200GB single disk filesystem, and it would be okay. But I just can't buy into that line of thinking. In my mind, an architecture based on Virtual LUNS makes good practice. You get much more benefit and you also get a standard LUN size and you also gain benefits of down the line when you add disks to your environment.

I dont know about you guys but my EMC frames have 3 different sized disks 32 81 171GB drives... you cant have that range and be able to manage whole disks on a VG or LVOL. UNLESS you have a standard LUN size.

With a standard LUN size, in my case of 9GB, It is amazingly easy to manage my environment. As my 11 hosts/5 Frame SAN Grows, 11TB at present, i find it very easy to expand using the virtual LUN architecture.

Please at least consider using LUNS, of course that will mean you will need to do additional work and probably dont have time to do it now.

BTW. I am actually inheriting 4 boxes that are setup this way, with whole disks allocated for Filesystems... and I have tried to get them to consider rebuilding them. I just see the whole disk setup as a hindrance to good System Administration.

Sorry if I am so passionate on this, I just dont want you to do this without considering the alternatives.
Unix, the other white meat.