Today I am going to describe the procedure I am using to upgrade the
VMware environment at work from ESX 2.5 to ESX 3.0 (or
VMware Infrastructure 3). This upgrade has ended up being much more complicated than previous version upgrades for several reasons. The two main issues are that vmfs volumes in ESX 2.5 are only readable (not writable) by ESX 3.0, and the virtual hardware on each virtual machine needs to be upgraded as part of the update.
Before I jump into this I should describe my VMware environment. The servers we use for our host systems recently changed so we are in the middle of a hardware migration as well as the ESX upgrade. Our older ESX hosts are HP BL20p blade servers using 8GB RAM and booting from our HP EVA 8000 SAN. Our new ESX hosts are HP BL25p blades with dual core AMD CPUs and 16GB RAM, booting from the same HP EVA8000 SAN. The new servers run approximately twice as many virtual machines as the older systems do. Our ESX 2.5 systems share a set of 1TB volumes on the SAN for virtual machine storage.
Our data center has two network vlans where half of the ESX systems provide connectivity one vlan and half to the other. We also run Virtual Center and have VMotion configured. There was also an upgrade for Virtual Center to 2.0 which was done earlier which turned out to be helpful because it would have been necessary for the migration.
Since all the ESX hosts share volumes on the SAN and the ESX 3.0 upgrade requires the vmfs volumes to be upgraded I would have to schedule an outage for 2/3 of our systems simultaneously and upgrade all of the ESX hosts at once. This is not a feasible option so I have had to plan a migration path rather than a direct upgrade.
The first step in the process was to install ESX 3.0 on our development systems to test for stability and find any potential issues with migrating to the new version. The systems were very stable and there were no problems found that would prevent us from moving to ESX 3.0 in our production environment.
After testing was done I had to move VMs around so I could clear off one host system. Then I rebuilt the host with ESX 3.0 and created two new vmfs volumes on the SAN. I scheduled individual outages for all of the virtual machines on a second ESX 2.5 host and migrated them to the newly built ESX 3.0 system on the new vmfs volumes. This proceeded until the ESX 2.5 host was free of any virtual machines, at which point the procedure started over with the next system.
I am still in the process of migrating to ESX 3.0 and the only issue I see is with the migration to the new server hardware platform. There are not enough BL25p hosts to hold all of the virtual machines we currently have in production and there is no money left in the budget to purchase additional hardware for this migration. I may end up having to keep a few BL20p hosts in production.
Overall the migration process has been smooth. There haven’t been any problems converting the virtual machines to the ESX 3.0 format and performance seems comparable if not better than ESX 2.5. The only thing that would make the whole procedure easier is additional server hardware which would allow more flexibility in migrating virtual machines.