WTF? Why would you stick the service console or Console Operating System (COS) on the VMFS volume?
Previous versions of ESX held the /Boot within the Red Hat formatted disk partition’s. Vsphere differs, it now has the Service Console stored within a VMDK which can be either held on a local VMFS partition or from a Shared Storage VMFS LUN. I guess some of the advantages to this might be:
- Maybe to remove the way the COS in previous releases is tied to CPU0, with it being a VM it can be scheduled.
- Being able to perform a snapshot the service console and have the opportunity to rollback if ever needed after upgrading/updating due to issues on the changes
- Probably means you can now provide dedicated resource throttling for the VM running the service console so it doesn’t have a chance to run away with the whole CPU
The Service console VM is stored on the VMFS volume as host-XX-XXXXXXXXXX and contains a vmdk labeled cos-vmdk. This vmdk includes the /boot, /,/var and swap partitions for the service console.
Service Console (COS) on local VMFS volume
But the downside is the VMFS volume can never be deleted and recreated therefore they cannot be upgraded. After upgrading to vSphere 4 I now have VMFS volumes at 3.21, 3.31 and 3.33 and I’m stuck with them unless I reinstall.
So why do I care about upgrading my VMFS partitions? For locally attached storage, I guess it isn’t a big deal, everything still works, but for shared SAN vmfs volumes or NFS volumes the newer 3.33/3.31 vmfs version introduces distributed locking optimization and can significantly reduce metadata IOs on the shared storage. Read here for a post that touches upon the advantages - VMFS 3.31 distributed locking optimization
This is also a problem if you wanted to ensure that the service console is on a separate VMFS partition away from other VMs. So what if we do a manual install instead of an upgrade? When performing a manual install of Vsphere from DVD/ISO it formats under an extended partition all of the local spare disk space with VMFS. If you want to change this behavior you need to use the scripted install.
So for me I don’t really care if it is isolated but I would like to try to keep all of my VMFS volumes at the most current level to take advantage of performance improvements and bug fixes. To do this I have to reinstall.
