ALB bond only using one nic

[danletkeman]

Hello,

I have setup an ALB bond on our P4500 two node san.  I have the switch configuration all done, with a trunk between the two switches, flow control all the way from the esx hosts to the san.

Nodes are connected as follows:

node1 nic1 - switch1

node1 nic2 - switch2

node2 nic1 - switch1

node2 nic2 - switch2

There is zero traffic on nic1 on both nodes.  I have three volumes setup on this san and there are two iscsi connecections from each host to each volume.  The traffic never goes in or out of the first nic on either of the nodes.

When I look at the ALB bond the mac address of the bond is the mac address of the second nic.  I knew this before.  I also understood that the san was supposed to take care of the load balancing and at this point I see no load balancing whatsoever.

esxi 5 w/dual vmk ports, p4500 dual node 15k 600g sas, 3 volumes, version 9.5

Any idea what I am doing wrong?

Thanks,

Dan.

[Bryan McMullan]

Do you have MPIO setup correctly on ESX?  Did you change it to Round Robin?  You can check in your paths to see if all of them are active. 

I know that sometimes it looks like ALB isn't working because there isn't enough traffic for it to really register.  You can try to slam it to see if you notice more traffic on the second NIC.

And just fyi...unless it has changed recently, ALB only uses one interface to receive, but uses the additional interfaces to send.  So depending on the traffic you are looking at, things will not line up perfectly.

[danletkeman]

Yes MPIO is setup correctly.  I can verify that.

I have slammed it, and there is zero traffic on the other nic.  Read or write.  Even the HP tech couldn't explain it.

Everything apears correct, but the san's are not load balancing the traffic.

[Bryan McMullan]

What is your switch setup?  I know I'm grabbing at straws, but maybe there's something there. 

Another funny thing is the ALB Mac is the primary NIC for the selection of nodes of mine that I checked (9 nodes).  I typically assign the IP of the node to the Primary NIC before I create the NLB.   You could always try deleting the ALB setup and recreate it. 

[danletkeman]

The switch setup is dual 3560G's, with a 4gig etherchannel between them.  Flow control all the way through from the hosts to the nodes.  Not much too it.

I my case it has chosen the second nic on each node.  I have not tried rebooting it as we have production VM's on it already, but I will adding a couple of nodes soon and I will be doing some updates so maybe it will fix itself after a reboot.

The one thing that I notice on the switches is that the two nic's that are not being used at all actually don't even exist in the mac address table on the switch.  So that tells you that they are definetly not being used at all.

If I wanted to change it to a LACP instead of ALB what would be the best process to do that?

[danletkeman · Accepted solution]

Rebooting the san solved the problem.