Vcenter server: fault tolerance on service console network

[casperello]

Hi everybody!

I've question about service console network on vCenter server. In my server, I've 4 free NIC and I want to make redundant the NIC witch manage service console network. Is it a good choice to teaming this NIC with other one free with the Intel Advanced Networking Services ( http://www.intel.com/support/network/sb/cs-009747.htm )?

My 2 vShpere host is configured like the Scenario 4 illustrated at this page:

http://kensvirtualreality.wordpress.com/2009/03/09/when-is-it-ok-to-default-on-your-vi/

My idea is to connect the primary service console NIC on Vcenter server to the one switch, and the other one to the otner switch then teaming the 2 NIC.

Is the good choice?

Thanks

[athlon_crazy]

Service console doesn't use much of your network bandwith unless you use SC base backup(VCB nbd mode). So redundant config for SC should be enough for you without nead to team both vmnics.

vcbMC-1.0.6 Beta

vcbMC-1.0.7 Lite

[casperello]

ok...but how I can redundant the config for SC on vCenter without teaming the nic with another nic?

Now, I've 1 nic with the ip 10.0.0.1 used in the host configuration that respond on vcenter.vmware.local...this nic is connected on a gigabit switch...if the switch fails? or the nic fail?

[athlon_crazy]

Like scenario 4, you'll need minimum 2x vmnic (active/passive or active/active) for SC management & vCenter communication. Let say :

• vmnic0 --> pSwitch1

• vmnic1 --> pSwitch2

As long both vmnics physically connected to the same subnet/network with vCenter and in the event when one vmnic or pSwitch failed, your ESX should be able to communicate with vCenter too. Do this at SC port group level but no need to do it on pSwitch level.

vcbMC-1.0.6 Beta

vcbMC-1.0.7 Lite

[casperello]

On the esx hosts, I've already done this configuration...

but, on the vcenter server I've only 1 nic configured in the SC network management...

this is my config...

if the switch 1 fails, the vcenter server is unreachable by esx01 and esx02...

[athlon_crazy]

It's make sense for this setup when vCenter running as VM otherwise, when pSwitch1 downed, your vCenter will don't have any connection to both ESX hosts due to it's ONLY physically connected to one pSwitch. However, since you already have redundancy on ESX & pSwitch, I don't think you need to worry much about your vCenter. You should worry about your VMs though. Yes, when it happened, you unable to do live migration & VM network configuration(vDS) but you should have enough time to get pSwitch replacement without single downtime on your VMs or ESX hosts.

vcbMC-1.0.6 Beta

vcbMC-1.0.7 Lite

[Walfordr · Accepted solution]

You should be able to accomplish what you want using the NIC teaming software from your NIC manufacturer.

Plug the second NIC into pSwitch 2, using the NIC software create a team. Assign both NICs (same make/model I hope) to the team. Give the SC network IP to the teamed NIC - To be clear move the existing IP from the single NIC.

If one of the VC server's NIC port fails and/or one pSwitch fails your VC server should still be up.

Everything else looks good.

----

---

Robert

BSIT, MCP, A+, VCP (VI3)

[casperello]

ok Robert, you have confirmed my firts idea!

thank you so much everyone!

[casperello]

Hi Robert...

I've configured the team with intel software, but I've this problem:

http://communities.intel.com/message/90423#90423

Where did I go wrong?

[Walfordr]

There is nothing wrong with your configuration. Its just how the intel teaming software works and the team type that you choose to use.

Adaptive Load Balancing (ALB)

Offers increased network bandwidth by allowing transmission over 2-8 ports to multiple destination addresses, and also incorporates Adapter Fault Tolerance.+ Only the primary receives incoming traffic. Only the primary transmits broadcasts/multicasts and non routed protocols. +

http://www.intel.com/support/network/sb/cs-009747.htm

You can also look into Switch Fault Tolerance (SFT) .

If your vCenter server is not yet in production give the ALB a test. Start a ping session from another box. Unplug the NIC see if you loose any ping replies while it failsover.

Almost all of my boxes uses Broadcom and their teaming software handles things differently. The team type I use in the Broadcom world is Smart Load Balancing and Failover. It does not show any NIC disabled.

Update: On second look at the "TEAM: sc-team Properties" window I think I am wrong. Because it list the definition for "Disabled: The adapter is in the team, but does not have link, is disabled in Device Manager or the Network Control Panel, or is experiencing driver issues."

-

---

Robert

BSIT, MCP, A+, VCP (VI3)

Please consider awarding points for "helpful" and/or "correct" answers.

Message was edited by: Walfordr. corrected firefox bad formating.

[casperello]

Update: originally the two switch are not connected among them...

but...if I link the two switch,the other NIC switch to ACTIVE state...

it's correct this configuration? (the 2 switch linked among them?)

[Walfordr]

Update: originally the two switch are not connected among them...

but...if I link the two switch,the other NIC switch to ACTIVE state...

it's correct this configuration? (the 2 switch linked among them?)

Yes, the physical switches should have been uplinked to each other with 1 network cable. I did not catch that they weren't because your original question was focused on the vcenter server. Once you have both adapters active you should test, to make sure it works.

----

---

Robert

BSIT, MCP, A+, VCP (VI3)

Please consider awarding points for "helpful" and/or "correct" answers.

[casperello]

Thank you so much Robert!

fault tolerance (and load balancing) with this configuration works fine!

I've tested the environment by unplug the power to both switch (one at a time) and the connectivity from hosts to vcenter and vice versa, work without loss.