VMware Cloud Community
jgarvey2
Contributor
Contributor

Teaming NIC problems

Having problems with my nic teaming. Found out that we were dropping packets using freeping (free utility to constantly ping addresses). The reason we started to investigate was that a developer called and said they were losing connections to one of there VM servers. We have had this setup for over 2 years with out realizing there were any problems. Would not have known until we started pinging all of the servers on the dual nic setups. We realized many of them were dropping ping requests. The problem is it is very intermittent. If I break the pair and remove one nic from the switch everything runs clean. If I add a nic back the problem eventually comes back. It has not ever lasted for longer than a day as far as I can tell. I was in with support while this was going on and was told they do not have a clue. The problem being it would straighten itself out while we were trouble shooting. I have HP 380 servers, with dual 4 port Gig ethernet cards. I have created virtual switches teaming 2 nics per switch. I called support and verified all of my settings, using Cisco equipment those ports are configured correctly. Has anyone else experienced these issues or are not aware of them and may not know it. We are not showing any dropped packets on the Cisco side it seems to be in the virtual switch,

Reply
0 Kudos
6 Replies
Cruicer
Enthusiast
Enthusiast

You said if you break the team, it runs fine...is that regardless of which NIC is plugged in? You could have a bad NIC / bad Cable.

On your vSwitch Properties you have Load Balancing check with Route based of the originating virtual port ID set. How many port group are assigned to the vSwitch...anything trunked? How many VMs attached to that vSwitch?

Reply
0 Kudos
jgarvey2
Contributor
Contributor

It does not matter which nic you pull, we have changed cables,changed the ports they are using on the Cisco switch even tried different switches. There is no trunking. Two nics per vswitch and 3 vswitches per server. We had one switch with just 2 servers. This is happening within 2 different clusters. 5 different servers. Only happening on servers with dual nics setup for the virtual switches. Support was pretty thorough about checking all of our settings as I was on the phone with them for a total of 10 hours and have still not resolved. At this point I have just broken the teaming on all of my teamed nics. I have also now tried this in my testbed and am getting the same results. Would not really notice any problems except for the timeouts during the ping requests.

Reply
0 Kudos
paul_xtravirt
Expert
Expert

On the Cisco switches do you have port security enabled? If so, if you show the config for the port what config is set?

I think the command is something like this (it's been a while) sh int G/1/0/5 : for example for port 5 on switch 1

If you found this helpful, please consider awarding points

If you found this helpful, please consider awarding some points
Reply
0 Kudos
efarzad
Contributor
Contributor

How have you configured your teaming on ESX host and physical switch ports?

1. What model of Cisco switch do you have?

2. Have you configured EtherChannel on the Cisco side? If yes, what mode is your virtual switch load balancing?

Reply
0 Kudos
estebs1978
Contributor
Contributor

Maybe this is too basic but; have you checked your firewall and also the host to make sure that the cable/patch panel/router port is not creating the issue? Do you have a VMotion port? Sometimes those ports can create all kinds of hell when you are having network issues (speaking from experience).

Hope this helps...

Esteban Blanco, MCP, MCSE, MCTS, MCITP

Esteban Blanco, MCP, MCSE, MCTS, MCITP
Reply
0 Kudos
dlhartley
Contributor
Contributor

Hi jgarvey,

We had a similar problem a little while ago, and

we're using Dell servers with quad-port gigabit cards and Cisco

switches (3750s). It's a little hard to try and troubleshoot without

posting any config, so here's what you'll need to check to make sure it

works..

It really depends on your choice of load-balancing and that kind of thing, but this is how we set it up and it works well.

Cisco side:

Configure the type of load-balancing to IP by running this command: port-channel load-balance src-dst-ip

On each of your port-channels make sure you include the command switchport nonegotiate - we found that this was a big help to stop packets dropping!

Make sure that you also include the last command (switchport nonegotiate) on the ethernet ports that you've assigned to their respective port-channels.

On the ESX host:

On the "NIC Teaming" tab of your virtual switch, make sure the load balancing has route based on IP hash.

(When you're changing these settings make sure you do it on the vSwitch object not the service console)

Ultimately you can use whatever load balancing you like here as long as you match it up with the load-balancing on the switch. Seeing as we're using src-dst-ip on the switch we've chosen "IP hash" here.

That should be all you have to change - the only difference between that config and ours is that we're using trunking for the switchports so we can assign any VLAN to any VM's network card.. but it also works when you use "switchport mode access" on the switch, which i'm assuming you are because you stated that there's no trunking.

See if that makes a difference - it worked for us.

David

Reply
0 Kudos