10-19-2012 08:06 AM
I got a performance related problem on an aplication: Domino Lotus. The aplication is on IBM servers running Windows 2008 Servers.
They are connected trough a Fabric Module Switch on the Blade Chassis H. These switches are 8GB Speed, and connected trough an HP Director and then go to the Disk on a Eva HP Storage.
The mismatch that I see is regarding the fillword parameter, on the Brocade Switch in the IBM Chassis are configured with 4GB speed and fillword 3, the ISL is connected to the port on the HP Director with the configuration 4GB speed and fillword 0.
I don't understand if the fillword parameters say the following:
0 | -idle-idle
1 | -arbff-arbff
2 | -idlef-arbff
3 | -aa-then-ia
Attempts hardware arbff-arbff (mode 1) first. If the attempt fails to go into active state, this command executes software idle-arb (mode 2). Mode 3 is the preferable to modes 1 and 2 as it captures more cases.
So, both ports are on E_port right now.
The Director is on -idle-idle while the Switch on IBM Chassis are on -aa-then-ia.
Is this configuration ok? This 0 and 3 configuration wouldn't cause a mismatch? If we change the parameter on the HP end for 1 or 2 what would be spect to produce? Should that change the port status from E-port to G-port? If we change it on the Switch IBM for a 0, what could happend?
Are the fillword responsible in any ways on the performance and errors count for the ISL interface?
Thanks for your comments.
10-22-2012 02:46 AM
AFAIK, fillword is only required when a port is working at 8G, so with 4G links, it is not used, and the fillword is IDLE (no matter what it is put in the fillword parameter). So I would not worry about that.
To see if everything's ok on this port, check the output of portstastsshow <port> on the Director, you'll see that er_bad_os counter does not increase.
I don't think that your perf issue could have relation with fillword.