MTU!!! That was the problem. As we've always previously done layer 2 (flat) network and our iSCSI switches are configured for jumbo frames, we always set MTU to 9000.
So when we flipped over to layer 3, I didn't even think about MTU. I just flipped over the IP address, set netmask, added the gw per iSCSI NIC. What confused me was how I could ping, traceroute, discover down the iSCSI NICs. But not login. Since these are TCP connections, I'm surprised path MTU discovery didn't save me. I'm sure it's that one hop that's the culprit -- when I was layer 2 that same iSCSI box was working fine w/ jumbo frames. Thanks for responding. Wireshark was an excellent idea BTW -- I would have been the problem immediately w/ wireshark. I'm guessing my discover packets were <1500 bytes, but my login packets must be >1500 bytes. Spike On Fri, Jun 22, 2012 at 1:29 AM, Ulrich Windl < ulrich.wi...@rz.uni-regensburg.de> wrote: > Hi! > > Did you check for MTU problems? Like: The router cannot handle MTUs that > are sent, and does not fragment, but discard them? Anyway tcpdump or > wireshark might help. > > Regards, > Ulrich > > > >>> Spike White <spikewhit...@gmail.com> schrieb am 21.06.2012 um 04:41 in > Nachricht > <CAO2Co24d=lc29vxtxmoa0n2wdk3lhoq+cp2q7avefzekv3t...@mail.gmail.com>: > > Hi. > > > > I have dual redundant dedicated iSCSI NICs with RHEL5.7. So I'm running > > the iscsi-initiator-utils--6.2.0.872-10.0.1.el5 that comes w/ RHEL 5.7. > > It's based directly on open-iscsi. > > > > Using CHAP authentication. No iSCSI off-loading. > > > > When I do layer 2 (flat routing) where the iSCSI target portal and iSCSI > > NICs are on the same subnet, all works great. > > > > When I move the NICs one subnet over (one hop to the iSCSI target > portal), > > then ping to portal works. Discovery of iSCSI targets works. But logins > > to all targets hang and ultimately fail. Why? > > > > I've defined the gw for each iSCSI NIC in the > > /etc/sysconfig/network/ifcfg-eth* files. (That's why ping and traceroute > > works). > > > > Why did I move one subnet over? Because my network guy wants me to > measure > > the increased latency introduced by going from zero hops to one hop. > > (They're not a fan of flat networks, too inflexible in large env's). > > > > Spike > > > > > > -- > You received this message because you are subscribed to the Google Groups > "open-iscsi" group. > To post to this group, send email to open-iscsi@googlegroups.com. > To unsubscribe from this group, send email to > open-iscsi+unsubscr...@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/open-iscsi?hl=en. > > -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To post to this group, send email to open-iscsi@googlegroups.com. To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
