Shared Uplink for VC in Bay2 is the Linked/Active, not VC1

[Clark Hamilton]

I am running two Virtual Connects into two Different Cisco 6509's for redundancy.

The shared uplink set in VC2 Bay 2 is defaulting to Linked/Active instead of VC1 Bay 1.

VC1's connections are two 1GB fiber cables which is a 2GB port-channel from the switch, into a shared-uplink set.

VC2's connections are two 1GB copper cables which is also a 2GB port-channel from the switch, into a shared-uplink set.

All four of the cables are part of the same shared uplink set (all vlans available on switch are trunked and tagged from the switch)for redundancy if the active VC fails.

Why is it defaulting to VC2 as the linked/active pair and VC1 as the linked/standby?

[WFHC-WI]

Does resetting module 2 resolve this or does the traffic move back onto the module following the reboot?

[HEM_2]

I think when the LACP channels are equal bandwidth as in your case, the decision criterion falls back to something arbitrary like the lowest MAC address of the connected switch (or if they are all connected to the same switch it uses the lowest port ID).

At one point I think it used to calculate the number of downlink ports on the modules themselves and chose the active uplinks based on the module with the most number of downlinks in those networks but that was probably many fw revs ago.

Look at the Hardware Overview and what VC reports as the MAC address of the connected switch.

[The Brit]

Clark,
I believe that this is the correct behaviour when uplinks on different vc modules are contained in the same uplink set.
(I may be wrong, and if I am, I'm sure that someone with better information will respond).

I assume that you are doing NIC Teaming on the host, so that you can have network redundancy? If this is so, then this is the configuration (we) use.

The problem can be resolved by creating an Uplink Set on each VC module (with different names, e.g. UPLNK_A, UPLNK_B).

For each vLAN on the current Uplink set, create an ethernet network, say _A, _B.

Assign the _A to UPLNK_A (on VC Module 1 say), and _B to UPLNK_B (on VC Module 2).

Now, in your server profile, assign (Teamed) NIC 1 to _A, and (Teamed) NIC 2 to _B.

You should find that both Uplink Sets show up as Linked/Active.

Also, if you think it through you will find that this configuration provides the network redundancy that you are after.

Note: 1. VC Modules failover horizontally (not vertically, so your VC modules should be installed in horizontal pairs.

2. Failover between VC modules takes ~25 seconds, so unless your systems/clusters can tolerate this, NIC Teaming on the blade is a must.

3. Considering Note 1. Your teamed NIC's must be assigned to VC modules which are a horizontal pair, in which case (with the above configuration) failover is immediate.

HTH

Dave.

[Clark Hamilton]

Thank you, all, for your responses thus far.

WFH-WI: Powering down the active VC2 fails it over to VC1, but then fails back to VC2 once it is powered back on.

HEM: This is interesting, I may have to call HP Engineering support on this one. Thanks for the tip.

The Brit: All four of my connections are in the same uplink set so I can have VC redundancy and not NIC redundancy as my server NICs are dual-homed with different vlans/IPs. I have port-channeled each pair of 1GB cables from the 6509 to channel 2GB to each VC. Should I lose once of my 1GB port-channels I still have 1GB of bandwidth and I won't have to failover the VC. All vlans available on my 6509s are trunked to each VC, and each vlan has a tagged ethernet entry created in the uplink set. My VCs are in BAY1 and BAY2 for the redundancy.

-------------------------------------------

Long story short, there is really no risk to VC2/BAY2 being the linked/active connections, but more of a "clean-living" scenario. However, I do want to get to the bottom of this. I will start with HEM's suggestion first at this point.