If you’ve watched an ESXi 5 host boot, you’ve probably seen the bottom right corner that shows <SHIFT+R: Recovery Mode>.
I did a bit of looking in the vSphere 5 documentation, the VMware KB, and Google, and still haven’t found a clear answer as to its purpose. It appears that it offers the ability to rollback to a previous hypervisor. As you can see in the image below, the current hypervisor I was running was one that had been patched with ESXi500-201109001, making it Build 474610. If I’m reading the KB article correctly, that build consists of both patches (more on that below).
If you opt to replace the current default hypervisor, the system will boot on the previous build. If you have no alternate hypervisor (you just installed ESXi 5 GA and haven’t patched yet, for example), then you’ll see a warning in the Recovery Mode screen stating that there isn’t an alternate hypervisor to rollback to. The system will then boot the only hypervisor available.
I discovered an odd thing about using the Recovery Mode feature: when I rolled back to the original hypervisor state, the listed build was 469512, but VMware Update Manager reports that patch ESXi500-20119402-BG was still installed. I don’t know why the system wasn’t fully rolled back to the original build, requiring that both patches would need to be installed. Not sure if this is a bug or a “feature”. 🙂
I also noticed that my SSH settings were reverted back to the defaults after rolling back the hypervisor. Previously, it was set to start with host, and after the rollback it was set to manual.
Due to service changes and what appears to be unreliable patch rollbacks, it might be best to use recovery mode in limited situations, and perhaps only when directed to do so by VMware support. Until some documentation, a kb article, or some other explanation of how this feature works and its potential use cases is released, I’d refrain from using it in a production environment without testing first.