SC and VMotion, same vSwitch, failover ?

[HendersonD]

Below is my networking configuration. I have trunking (802.1Q) setup on the switches that feed the two pNics on my VM Network. I also have etherchannel setup on both of these nics for bonding. I have load balancing set to IP Hash, failover to link status only, notify and failback both to yes. This setup I feel comfortable with and from what I gather is best practice for the VM Network. If I am off on this, let me know.

The switches that feed the pNics for vSwitch0 are not setup for trunking, they are just access ports. I do have etherchannel setup for bonding. The Service Console on this vSwitch is the only thing that needs to be routed so I did not think that I needed trunking. vMotion traffic does not get routed so I do not even have a gateway setup for this in its properties. Right now I do not have any load balancing set but I do have failover set for Link Status only, notify to yes and failback to no. Is this setup correct? My understanding with having my pNics setup with etherchannel that failover might just occur anyways so I am questioning my configuration.

[thickclouds]

You are in good shape. Just watch out fo too much vmotion traffic flooding your SC network.

http://communities.vmware.com/blogs/chuckgman

[HendersonD]

Do I even need to setup failover within the portgroups in vSwitch0 since the physical switches that support it are bonded with etherchannel and if one fails the whole thing should just keep going anyways?

Even though vMotion traffic is on the same vSwitch with my Service Console, since they are on separate vLans (vMotion traffic is also not even routed) can vMotion traffic even affect my Service Console? If you notice I do have my vMotion traffic on the same vlan as my VMs but of course they are on different vSwitches

[thickclouds]

you do need them both in vswitch0. also, if you are vlan tagging, both pnics for vswitch0 need to be able to send over both vlans for redundancy. SC/Vmotion redundancy is extremely important.

http://communities.vmware.com/blogs/chuckgman

[lldmka]

I have a very similar vSwitch0 configuration to this (with no VLAN tagging) and have proven redundancy - the VMotion pNIC has currently lost connectivity but VMotion and Service Console are both fine. Failover is set within the portgroup.

[jbruelasdgo]

a good configuration for avoiding traffic contention would be to put 1 NIC active por service console and the other stand by in the service console portgroup

and viceversa in the vmotion portgroup

and you still would have redundancy

regards

Jose

[HendersonD]

I am not vlan tagging. Even though my Service Console is in vlan 125 and vMotion is in vlan 127, my phyiscal switch ports are setup as access ports for vlan 125 only. For the vMotion portgroup on all three of my ESX servers, I do not have a gateway entered since this traffic is never routed. My understanding is that vMotion traffic never actually traverses the physical switch as a VM is moved from one ESX server to another. If that is the case then my physical switch never sees traffic from vlan 127 so I do not need vlan tagging. In other words, the only traffic that my physical switch will see is from the Service Console, vlan 125. Does this sound correct?

[tjuksa]

I second this jbruelasdgo. It's the way we've configured our system as well. And it works perfectly.

Use one nic as active for SC and the other as stand-by and the opposite order for your vmotion.

We've tested with pulling out the nics and it switches over seamlessly, even during a vmotion job

Regards

Hyper-T

[jguzmanr]

How could vmotion flood the network?.......never mind

[thickclouds]

check out page 68 in the config guide.

http://communities.vmware.com/blogs/chuckgman

[HendersonD]

What config guide?