[Linux-cluster] gfs performance

[Jure Pečar]

On Tue, 31 Jan 2006 09:02:08 -0500
Bowie Bailey <Bowie_Bailey at BUC.com> wrote:

> Just for informational purposes, what was the problem and how did you
> fix it?

I did mention it there on the page. The fact is that DS4100 does not like having two nodes connected to it each via single fibre to separate controller ... it does weird things(tm) when both nodes want to do i/o to a volume that has one controller assigned as primary, but is receiving simultanoeus requests also via secondary controller ... 

One has to either connect both nodes to the storage via the same controller or use dual channel HBAs to connect each node to both controllers and do some multipath magic. I was somewhere inbetween :)

-- 

Jure Pečar
http://jure.pecar.org