Setting a redundant Management Network

[RobLI]

Currently, I have one physical nic on vSwitch0 for the management network which is set with "Management Traffic" enabled. After reading some of the best practices docs, I see it would be nice to have a second, redundant network configured for Management Traffic for VMHA. Unfortunately, I don't have another physical nic to add to the system. Is it wise to configure the VMotion Network to act as the backup for this?

Any ideas would be appreciated.

Thanks

Rob

[kegwell]

Is this production or test/dev?

[RobLI]

This is a brand new installation and it's being set up for production. At the moment, I have a few tertiary production VM's running on it. But nothing business critcal yet.

I have 8 nics per server, I've got one for Management, one for VMotion, one for Fault Tolerance, two for iSCSI and three for the VM Network.

[wpatton]

RobLI,

I would definitely recommend setting a second pNIC to the Standby Adapters for vSwitch0.

If the pNIC for the vMotion network is cabled and available for use as a Standby Adapter, in case of failover, it would be a good idea.

This is the configuration we use in our ESX blades, it separates the traffic of Service Console "SC" and vMotion traffic during normal operation, but in the event of an adapter/link/port failure we can maintain functionality of both vMotion (with SC pNIC as Standby Adapter) and SC (with vMotion pNIC as Standby Adapter) to remove the Host from the environment until repaired.

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".

[RobLI]

So then are you advocating that I add more pNICS to this box, or do you think that my current configuration should be changed to remove a NIC from say the three I have on the VM Network? Using the three on the VM Network was a recommendation by a VMWare support tech. It seems to make sense since there will be many VMs running, and the more pNICS on that network, the better the performance.

Thanks

[wpatton]

RobLI,

I am not advocating more pNICs. Rather, some redundancy changes.

First question, how are you handling the 3 VM pNICs for Load Balancing? If they are simply "Route Based on the originating virtual port ID" Load Balancing with "Link Status Only" failover you may not be getting the performance boost you are expecting.

Second, let me clarify my above post. I am simply saying that you could benefit from additional redundancy, depending on your vSwitch configuration.

1. vSwitch0 - Portgroups "SC" and "vMotion" - pNICs "vmnic0 and vmnic1"

(Sidenote, depending on your configuration of pNICs, singleport or dualport or quadport, this may need to be "vmnic0 and vmnic2" (Dual Port Adapter) to ensure protection against adapter failure.

Now, with the above vSwitch, Portgroups, and pNICs added to the vSwitch. You can configure the two Portgroups to route over separate vmnics.

Go to the Portgroup for "SC" and Edit. Under the NIC Teaming Tab, you can then select "vmnic0" as Active Adapter, and "vmnic1" as Standby Adapter.

Go to the Portgroup for "vMotion" and Edit. Under the NIC Teaming Tab, you can select "vmnic1" as Active Adapater, and "vmnic0" as the Standby Adapter.

This will then compensate for failure of a port, and allow for either Portgroup to continue routing traffic. With your VM vSwitch, or what I would call now "vSwitch1" I would leave all 3 vmnic adapters (vmnic2, vmnic3, vmnic4) Active Adapters. I would ensure you have the vmnics on physically distinct adapters to ensure against an adapter failure.

I would do the same as "vSwitch1" for what I will call your iSCSI "vSwitch2" (vmnic5, vmnic6, vmnic7)

FT is the tricky one with only one remaining pNIC. "vSwitch3" could be consolidated to a Portgroup of "vSwitch0" or kept separately. To help keep FT in a FT state, and highly available, I would encourage you to look at making it a Portgroup of "vSwitch0" and giving it "vmnic8" as its Active Adapter, using "vmnic0" and "vmnic1" from "SC" and "vMotion" as your Standby Adapters. However, this will mix what some may call "VM" traffic into your "Management" network. I will leave this final architecture choice up to you and given your exact environment.

Little bit older doc, but still a good overview from VMware of what I propose above: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=100272...

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".

[RobLI]

wpatton

Thank you for the clarification on this. I like these ideas, the only thing that doesn't compute with me is how you can use these adapters as backup for each other like this. I neglected to tell you that I am using Physical VLANs set up on our Extreme Networks switching fabric. I have 3 VLANs.

1. Management/VM Network,

2. VMOTION,

3. iSCSI.

So is it still really possible to use something like your vSwitch0 idea?

Thanks

[wpatton]

RobLI,

Probably not from what you are describing in your environment.

For example, the two vmnics in vSwitch0 above:

If vmnic0 which is currently dedicated to SC does not also have the VLAN for vMotion trunked to it, then no, my configuration above will not work if a Link Status failover occurred, the adapter could not route the VLAN tagged traffic for SC over vmnic1 if it only has the vMotion VLAN trunked.

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".