VNXe 3300 stops communicating after binding NIC's

Thank you for the response.  I am looking at two scenarios....

Option 1 - 2GB path from each SP

VNXE SPA – Two ports bound on EMC and connected to iSCSI Switch 1 Ports 1-2

VNXE SPA – Two ports bound on EMC and connected to iSCSI Switch 2 Ports 1-2

VNXE SPB – Two ports bound on EMC and connected to iSCSI Switch 1 Ports 3-4

VNXE SPB – Two ports bound on EMC and connected to iSCSI Switch 2 Ports 3-4

Option 2:  Since the SP's are active/active, can I just do this for a full 4GB path?

VNXE SPA – All four ports bound on EMC and connected to iSCSI Switch 1 Ports 1-4

VNXE SPB – All four ports bound on EMC and connected to iSCSI Switch 2 Ports 1-4

I spoke to someone on chat tech support and they said that I needed to bind the ports on
the EMC side AND at the switch side.  It sounds like that may not be correct?  Looking at the document you highlighted, it does not appear as though there is any aggregation/binding happening on the switch at all.

Ultimately, I want as big a data path coming out of my SAN as possible while maintaining redundancy.

For Option 1:

• It is not possible to link aggregate eth4 and eth5 only. Any aggregation must be with port eth2. i.e. eth2+eth3, eth2+eth4, eth2+eth5, eth2+eth3+eth4, eth2+eth3+eth5, eth2+eth3+eth4+eth5 etc.
• When your link aggregation on the switch side spans across two switches, then switch must support that. Usually low end unmanaged switches don't support it. I think it is called switch stacking in which both switches work together as one.

For Option 2:

This provides fault tolerance against port and switch failure as you needed. However please note binding (LACP) 4 1GB ports may not necessarily give 4GB bandwidth to your servers. The LACP algorithm in VNXe is based on MAC address which means, if there are two ESX hosts connected to VNXe they will use one 1GB link each to connect. This is maintained until that link is faulted or disconnected for some reason. For this reason LACP is best bet for CIFS/NFS kind of clients where there may be hundreds of clients connecting to storage which will be load balanced across links.

About what tech support said - yes, you need to link aggregate the ports on VNXe side as well as switch side. On your switch it is called - LAG. For example, you connected eth2 and eth3 of SPA to Switch1 on Ports 1 & 2 respectively. You need to aggregate eth2 & eth3 on Unisphere and then on your switch side add Ports 1 and 2 to same LAG group. Please refer to your switch manual for more accurate details.

Ok, so Option 2 is what I currently have with the 4 ports bound on the emc SPA and connected to one switch, and the same with SPB on a different switch.  My switch does support LAG however when I put the four ports into a lag group, I can no longer communicate with the VNXe...  I think I am going to try a different pair of switches.  I have never really been happy with LAG on the mediocre model Dell switches...  But it seems to communicate fine when the ports are bound at the emc but not at the switch. If I reverse that and turn on LAG at the switch and turn it off on the EMC it still does not work.  So I will see if replacing the switches with a pair of procurves makes a difference.

So the problem turned out to be the switches.  I swapped out my Dell switches and used a couple update procurve switches and my link aggregation now works without an issue.  Thank you for the help!