UCS FI's Active-Active

[K P SIM]

Hi, Cisco Gurus:

We all know that the UCS FI is active-active for Ethernet & FC data. It is only Active-Passive for Management Plane.

1.

Is there a command we can run from our FI to show customer that UCS FI is Active-Active in terms of FC and Ethernet traffic? Customer wants to see such command outputs to compare with their SAN Switches. Assuming that we have a job which will generate some traffic likes copy data from 1 VM to the other, ftp from 1 VM to the other, etc.

2.

Is there a white paper which shows UCS FI is Active-Active for Ethernet and FC Data Traffic?

Thanks.

SiM

[Jeremy Waldrop]

I don't know of a single command but there are a few ways you could demonstrate this.

1. Build a Service Profile that has 2 vNICs, 1 on Fabric A and 1 on Fabric B
2. Installed ESXi on the server
3. In VMware add both vNICs to vSwitch0
4. Build 2 VMs and place them on this vSwitch
5. On the uplink network switch look at the mac address tables for the interface connecting to UCS. You should see the mac of 1 VM on each fabric/switch
6. In VMware take 1 of the vmnic uplinks away and both VMs will start using the other link and then do the same with the other vmnic
7. Also in VMware if an uplink is not active VMware will place a red X on the vmnic showing that it is down.

1. On the SAN side build a Service Profile that has a vHBA in each SAN fabric
2. In VMware on the Storage Adapters section you can show that both HBAs see the SAN paths (assuming zoning and SAN LUN masking are properly configured)
3. Also on the SAN/SAN FC switch you can show them that both vHBAs are logged into the fabric and SAN. If they have MDS for FC switching the commands "show flogi database" and "show zoneset active" will show the logged in initiators.
4. If UCS were an active/standby system then in VMware you would only see SAN paths/LUNs on 1 HBA and you wouldn't see both logged into the the FC fabric or SAN.

[K P SIM]

Hi, Jeremy:

Thank. I am still thinking of running some commands to check likes show interface fc2/1 and others to prove to customer there is some FC frames being sent northwards. Brocade engineer claims that they only saw 1 side of the FC traffic, but again I need to verify likes:

1.

how long they run the command as when they only ran once, the FC traffic may have been coincidentally sent through the same FC uplink.

2.

what is the server's IO profile during the time they ran the command from Brocade switch?

3.

what is the Brocade command they use?

Please help as you are from Cisco and we need to prove FC data is really being sent to both FC uplinks though it does not have to be concurrently. I think we can achieve that by having more blades issuing or simulating FC IO, etc.

Thanks.

SiM

[cparik]

Hi SiM,

For question number 1, you can run the command "show service-profile circuit name " to see the status of the links on the two fabrics. You can also run the commands "show npv flogi" on the nxos of each FI to see that the vhba on the server has logged in. Similarly "show mac address-table" on the nxos of each FI to check if the mac address of the vnic of a server are learned.

You can use the following example to display to that the FIs are in active-active configuration for ethernet. For this example you need a blade with ESX installed on it.

1. Configure a service profile with two vNICs connected to two different fabrics. i.e. vNIC 1 connected to fabric A and vNIC 2 connected to fabric B.

2. Create two vSwitches on the ESX host, vSwitch 0 and vSwitch 1.

3. Create VM network port groups on each vSwitch.

4. Add one vmnic as uplink for vSwitch 0 and the other vmnic as uplink for vSwitch 1.

5. Connect a virtual machine to each port group.

6. Start a continuous ping from each virtual machine to any routable IP address. If the configuration is proper then you should get replies for the ping requests, thus proving that the FIs work in an active active configuration.

For FC, you can try configuring one vhba each on two different blades. Connect the vhba on one blade to fabric A and the vhba on another blade to fabric B.

Present two separate luns to these two vhba. If your zoning and FC is configured properly, you should be able to access the respective LUNs on both server, thus proving that the FIs are in an active-active configuration for FC also.

For question 2, I will have to check if there is any whitepaper. I will update you if I find anything helpful.

[K P SIM]

Hi, Chetan:

Thank a lot.

I am still looking for commands just to prove to Brocade engineer. See my reply to Jeremy to see whether you can help me to help Cisco or not.

SiM

[Robert Burns]

KP,

Understand that though UCS provides two active paths for both Ethernet & FC, it's up to the Host OS to handle multipathing and load balancing - not UCS.

We can "prove" both paths are active by looking at the interface counters, but you first need to ensure both paths are being used by the OS.

The examples Jeremy and Chetan have suggested will show the failover capability.  Using the "show service-profile circuit" you can determine the interface #s.  From there you can look at the interface counters and see them incrementing.

Regards,

Robert

[K P SIM]

Rob:

Thank a lot as ususal. I will probe further and will come back if I have been overshelmed by customer questions.

KP

[Jeremy Waldrop]

It is also important that UCS and Service Profiles are properly configured with northbound LAN/SAN from each Fabric Interconnect. On the Service Profiles make sure there is a vNIC/vHBA on each Fabric (A/B).

Data planes are active/active but configured incorrectly you could end up with a non-redundant active/passive or worse active/no-paths system.

[K P SIM]

Jeremy:

Thank to you also. Will get back if still being able to articulate and convince customer successfully.

SiM