We did recently purchase a new HP BladeSystem c7000 Enclosure that was ordered with HP OneView. We started the deployment when the version 1.05 was out, to configure it directly in Active-Active mode. After a few problems with the iLO firmware we started a lot of tests to ensure high availability of the solution.
A major problem was found during the tests, when removing a network interface that was directly connected to the HP c7000 the network connection was lost and all the infrastructure was down. The HP c7000 is connected to two Cisco switches that does not support stacking. One Virtual Connect FlexFabric is connected to the switch one and the second Virtual Connect FlexFabric is connected to the switch two. The two switches are linked by a 10GbE link too.
The PDF “HP OneView 1.05 User Guide” explains how to configure it correctly for Active-Active mode, that was what we followed and we noticed that the result was not really good. Perhaps this documentation is only working for stacked switches ?
Finally we found a solution for this problem in an old documentation for Virtual Connect. When configuring a new Interconnects Group in OneView you’ve some settings by default. OneView 1.05 is set up to runs in Active/Active mode, but in the settings we noticed that the checkbox “Fast MAC cache failover” was checked. When referring to the Virtual Connect documentation’s it’s written that this checkbox need to be used in Active/Standby configuration. Something is wrong here…
We have made a new test without this option and the high availability was working fine… So if you’re working with HP OneView it’s important to test how this checkbox is reacting in your environment. Because you can be sure that the day you go to meet a problem, it’s may be too late.