L3_L4 LAG setup with Intel 82575EB 1G NIC.

  • 0
  • 1
  • Question
  • Updated 3 years ago
  • Answered
I trying to create team using the L3_L4 algorithm on a Summit x440, but I had issues with NIC side.  So I was wondering which one method do I choose on the NIC side of the team on the server. I original choose ALB on the NIC side but then when I was looking at the NIC teaming setup I noticed one nice was set to primary and the other was disabled, so while trying to figure it out I ran the test option on the NIC and it suggested using the SLA so I made some changes and it broke.  Now I am at square one.  But based on what I know of L3_L4 I should be using ALB. But every time I set it up one NIC is disabled.  Was hoping someone here had some similar experience.
Photo of Joe

Joe

  • 430 Points 250 badge 2x thumb

Posted 3 years ago

  • 0
  • 1
Photo of Bill Stritzinger

Bill Stritzinger, Alum

  • 6,016 Points 5k badge 2x thumb
Joe, 

Sorry for the delay, just found this... what is the OS you are running?  Based on some of your comments assume it is Windows Server?

Bill
Photo of Joe

Joe

  • 430 Points 250 badge 2x thumb
Sorry for the delayed response.  Been busy.. We are running Windows Server 2008r2. All my other servers are using the Broadcom NICs and the same version of OS.
Photo of Joe

Joe

  • 430 Points 250 badge 2x thumb

Our 440's are currently using EXOS 15.3.1.4. We have the edge license.  We will be adding in a 460Gen2 as the stack master in an upcoming project, with the Edge license as well.  

After reading this:  https://community.extremenetworks.com/extreme/topics/port-load-sharing
I noted they are using L3_L4 algorithm with LACP between switches.  Is that possible to do between a switch and a server?  I would have sworn I tried this.

My original post above is about what mode to choose on the Server NIC when trying to use L3_L4, but now I am wondering now if tried to do "L3_L4 LACP."  I will have to try again.

Our goal here is to load balance our data flow to the servers primarily for optimizing our backups but we are also looking at implementing DFS with replication.

I have my current test LAGS working with LACP which defaults to the L2.  I am not currently set up on all my primary servers been testing with lesser used servers to avoid disruption to our users.  When I look at my LACP Counters things look evenly spread out between my ports.

What exactly is this " LACP PDUs dropped on Non-LACP ports:" telling me?

Photo of Joe

Joe

  • 430 Points 250 badge 2x thumb
So I was able to switch my lacp to L3_L4 LACP.. everything went fine but I noticed this output when I run "sh lacp lag 3:26 detail "

LAG Id flag      : Yes
  S.pri:0   , S.id:XXXXXX
  T.pri:65536, T.id:XXXXXX

I assume that the T.pri:65536 is some sort of priority indicator.  And when I run "sh shar" the output for the associated port and MAC address corresponding to the T.pri has less link up transitions, then the other NIC port.   This is my INTEL NIC Team.  The port coorsponding to the T.id MAC is my TEAM IP and associated MAC and when I doubled checked things this is also the first member of my TEAM.

I reverted back to the LACP with the default L2 as the algorithm and noticed the same thing in the details the Tpri 65536.  Do I need to play around with any NIC settings on this TEAM? 

My Broadcom is great. The S.pri and T.pri are 0 and 1 respectively on that team and the link up transitions are about equal., the counters are almost balanced. for btoh setups.

The only issues I see is with the first TEAM using the Intel NICs. If not for the high T.pri number and the low value for the transition everything looks like it is working.