NIC config of VMware installation of Call Manager

[David Hunt]

System version: 8.6.2.22900-9

VMware Installation: 1 vCPU Intel(R) Xeon(R) CPU E5-2609 0 @ 2.40GHz, disk 1: 80Gbytes, 4096Mbytes RAM

A few month ago we had a vendor come in and upgrade our old VOIP system with a VMware solution of Call Manager. Versions are shown above.

I use Veeam to backup these VM's and it was taking hours to finsih so I starting digging into VCenter to look at the netwroking setup.

Even though the server has gigabit ethernet cards they "hard coded" the speed down to 100 Meg and changed the duplex to half.

Does anyone know of any reason why they might have done that? Will it fail if I change the speed to GB and utilze full duplex?

thanks...

[joe.lazzaro]

Absolutely no reason to have the NICs hardcoded to 100/half. I would personally just leave them at Auto.

If you decide to hardcode 1000/full on the Hypervisor, make sure you hardcode the switch-side to the same settings otherwise you will still get half-duplex operation.

[David Hunt]

Thanks Joe,

through the command line interface, each of the 4 VM's

Call Manager 1

Call Manager 2

Cisco Emergengy Responder

Cisco Unified Messaging

are setup as below...

So, from Cisco Side, things look okay correct?

[joe.lazzaro]

Yes that looks fine.

Just a tid-bit of information, if you were to change any eth0 settings within CUCM/UCxn/etc, your product licenses would become invalid. The system MAC address (which is used to license the product) is generated by hashing all the NIC specific information. Virtual server = no physical MAC of course.

[joe.lazzaro · Accepted solution]

Actually, according to CUCM 9.1 (and 8.6) installation guide @

http://www.cisco.com/en/US/docs/voice_ip_comm/cucm/virtual/CUCM_BK_CA526319_00_cucm-on-virtualized-servers.pdf

"You should configure the same NIC settings on the server and on the switch port. For GigE (1000/FULL), you should set NIC and switch port settings to Auto/Auto; do not set hard values."

I have done 100+ virtual CUCM deployments and always use Auto across the board (Application, Hypervisor, Switchport). At this point, I would probably try the Hypervisor and Switchport hardcoded at 1000/Full to match CUCM. Check interface counters and watch for errors and runts.

I can't recommend mismatching CUCM at 1000/Full and Hypervisor/Switchport at Auto. It's not something I have ever done and YMMV.

Changing CUCM back to Auto would require a license change...if it comes to that have you integrator come back correct their installation.

[David Hunt]

Thank you very much Joe.

[joe.lazzaro]

Glad to help, David.

Something else to note, your 2609 CPU @ 2.4Ghz is considered a "Restricted UC Performance" processor. Are you running this on C220 M3 servers? See virtualization CPU table and the 8.6 OVA template table with the "Restricted UC Perf. CPU" heading.

http://docwiki.cisco.com/wiki/UC_Virtualization_Supported_Hardware#Processors_.2F_CPUs

http://docwiki.cisco.com/wiki/Virtualization_for_Cisco_Unified_Communications_Manager_(CUCM)#Version_8.6.28x.29

You are essentially stuck at BE6k templates (1000 user for CUCM). There are no valid configurations with this processor that should run CUCM with only 1vCPU. If you encounter performance issues with CUCM, you will likely NOT be supported by Cisco TAC.

[David Hunt]

Joe...

Questions and a comments

Questions in the post above you mention...

"At this point, I would probably try the Hypervisor and Switchport hardcoded at 1000/Full to match CUCM. Check interface counters and watch for errors and runts."

When you refer to "Switchport" I am assuming you are you talking about the physical port on our physical switch?

which is interesting becasue I just looked only to discover that on our physical cisco switch (Catalyst 4506) that our phone system is physcally plugged into our 10/100 module and not into our gig module. So even if I made all of those changes I still would have been at a 100 MPs speed becasue of the physical connection to our switch. So obviously... I would need to change ports on the physical switch to the gig module, correct?

question 2...

monitoring the ports for errors and runts, again I am assuming you are talking about the physical Cat 4506 switch?

Assuming then I move the connections into gig ports 5/26 & 5/28, etc...

From a telnet session into physical switch... after enable mode type...

switch#show int gigeth5/26

and I will see stats from that interface correct? or do you look for these in a different mannor?

comments... first of all, thank you....

second....

I am not sure I quite understand anything about you very last post.... about the restricted UC and Cisco support..

yes, I am running USC C220 M3 servers... are you saying that if I have over 1000 users that TAC will not support performance issue? I ask becasue we only have about 125 users....

[Jaime Valencia]

On a 6K deployment you're limited to a maximum of 1000 users configured on CUCM, and 1250 devices.

If you go over any of those figures, then we won't consider that a 6K CUCM anymore, but a regular CUCM cluster and you might need to switch to a different HW if the one you're using is 6K specific.

HTH

java

if this helps, please rate

www.cisco.com/go/pdihelpdesk