I have a two host cluster of 2008 r2 DC, a Cisco 6513 switch and a PS4000 that is the CSV
My questions have to do with three things; jumbo frames, STP and flow control
Do I need to configure Jumbo on all the host NICs as well as the switch and SAN?
I disabled STP on the SAN switch ports. Does that pertain to STP bpdufilter/bpduguard as well? I should disable these on the port too?
With flow control, should I enable for both send and receive? Not clear on how to confiogure flow control for ISCSI.
Thanks for any help.
Re: Jumbo Frames. With JF, everything in the path must be enabled for jumbo frames in order to have them negotiated. On the 6513, it depends on the SUP card and Line cards. Not all of them support both Jumbo Frames and flowcontrol. A really good combo is SUP720's and X6748 series line cards. It also depends on what slots the cards are in. The last few have limited connectivity to the backplane.
Re: STP. On the array ports and server NIC ports, setting spanning tree pvst (or portfast) is sufficient
On the 65xx series set flowcontrol receive desired and flowcontrol send desired.
Social Media and Community Professional
Get Support on Twitter - @dellcarespro
Jumbo setting for EQL is default to 9k. If you wish to take advantage of that frame size then your initiator MTU must also be set (1519-9000 range - in your advanced tab under device manager for COMMs devices <jumbo>). You should know that this feature deployment NEEDS to propogate through the entire connection path. Meaning all connection points must be set to that frame size or higher; that would be for Inband Switchports, VLANs, and even virtual interfaces (SVI's/Portchannels). Do a "show vlan mtu" to see the stats for the switch MTU sizes.
STP Portfast prefered, or even STP disable is suggested for iSCSI Hosts "directly" attached to switches. This is done to exclude the host from any long standing STP convergence that may occur. Depending on the STP mode this can take anywhere between 15 seconds for RSTP to 60 for STP. Normally you wouldn't disable STP from switch to switch (to avoid accidental config leading to storm - up to your net admin setting that up).
As for flow control this is typically default enabled these days (if your on the latest IOS then it is). This feature is a Layer 2 function and where iSCSI resides ontop of transport just under the session. ISCSI is not L2 aware. You enable this feature only to mitigate potential congestion (bottle-neck) to a point (i.e. several initiators to 1 EQL array for example).
I hope this helps explain.