topic difference between block size and extent size in Operating System - HP-UX

difference between block size and extent size

khilari — Tue, 21 Jul 2009 13:14:11 GMT

Hi Guys, just wondering what is the difference between block size and extent size. And how much of an impact it has in a DR situation that they should be the same as in the prod boxes, if not then there can be a problem at the DR site.
Thanks

Re: difference between block size and extent size

Ganesan R — Tue, 21 Jul 2009 13:27:49 GMT

Hi,

block size is at filesystem level and extend size at lvm level. Note that there is a block size in lvm level which is not used widely because it is not supported with mirroring.

Re: difference between block size and extent size

Mel Burslan — Tue, 21 Jul 2009 13:29:21 GMT

Block size is the minimum amount of space tha you can read from or write on to a *filesystem* whereas extent size is pretty much the same thing to a volume before the filesystem has created atop.

The sizing of these are somewhat important. For instance wildly different extent sizes on DR site may cause not being able to create large enough volume groups. Block size on the other hand may effect how the data laid out on the filesystem and size difference may cause wasted space (although not by much) and performance differences (positively or negatively depending on the application behavior).

Re: difference between block size and extent size

Steven E. Protter — Tue, 21 Jul 2009 13:30:34 GMT

Shalom,

http://docs.hp.com/en/6054/LVM_Limits_White_Paper_V4.pdf

http://docs.hp.com/en/5992-3818/5992-3818.pdf

Reading on the topic.

Impact on DR depends how you do DR.

If you have systems hot and built, and keep the data updated live, no impact at all. It really depends on how you replicate your data.

If you have different hardware at the DR center than production and use Ignite backups after the disaster to build the DR systems, you may run into issues building machines and need to run through a few test scenarios.

Then create notes on the assumption that you are dead and someone else will need to build the machines.

If feasible, I always like my DR boxes to be the same hardware as production and every detail storage, block size, extent size to be the same.

Of course that would cost a lot of money.

SEP