ESX 4.0 iSCSI initiator jumbo frame connection reset by peer - PS6000 SAN

[joemazur]

ESX 4 hosts, vswitch with 4 vmknics configured correctly for jumbo frames (MTU=9000 at vswitch / vmk, etc), teaming set correctly - everything per vsp_40_iscsi_san_cfg.pdf . Dell PowerConnect 6248 switches, VLANs for private iSCSI network, flow control enabled, STP disabled, etc, ports set to MTU 9000 / auto neg / max. Attempting to connect to LUNS on PS6000XV / E, initially connects, correctly showing jumbo frames, then disconnects - reset by peer. After much work, set PowerConnect switch ports to 1518 (standard) - connects with no problem, using standard frame length. Having connected to LUN, I can then set ports back to 9000, rescan iSCSI and everything is still there but cannot connect initially at 9000. Anyone have any ideas?

[beyondvm]

What do the switch logs say. I have not use the PowerConnect switches myself but it sounds like there is an issue with it negotiating the jumbo frames. How is the teaming set up on the vSwitch?

---

If you found any of my comments helpful please consider awarding points for "Correct" or "Helpful". Thanks!!!

www.beyondvm.com

[joemazur]

Thanks for your input but I just solved the problem after a bunch of research and experimentation. I had read several posts regarding various jumbo frames MTU settings ranging from 8000 to 9216. I then did a vmkping -s 9000 from one ESX host's vmkernel to another (to verify MTU) and got back 9008 as a response. So then I created a new LUN on the PS6000, initiated a rescan from the ESX host and then logged into the switch and kept changing the port speed until it connected without dropping. The actual value ended up at 8000 MTU. So for anyone else out there with ESX 4, PS6000 iSCSI SANS and 6248 switches, this should be helpful info.

[joemazur]

Please see my reply for details.