Just a quick post to remind myself of the gotchas when you do some routine maintenance on VMs in vSphere.
Disks
- Make sure the underlying datastore has capacity! An obvious one but probably the most important.
- Before ESXi 6.5 you couldn’t extend a VM online past 2TB, it has to be shut down and then extended. Link
- Don’t forget about the partition types, Windows MBR only supports up to 2TB if you want to go past that you need to format (and lose all data) then convert to GPT.
- If you are running VMware SRM (Site Recovery Manager) then you can’t edit a protected VM, you may need to stop the replication (unprotect the VM) make your adjustments then enable replication again which if it’s a large VM or slow network can take a significant amount of time and may require out of hours scheduling.
CPU
- Hot Add CPU is a great feature however NUMA isn’t supported with hot add, it’s a convenience vs performance argument. Link
- More isn’t always better, generally the less CPUs the better. Link
- Be careful with OS and/or application licensing, some apps are licensed per CPU so make sure to check beforehand.
RAM
- If you increase the amount of virtual RAM don’t forget that the swap file on the datastore will also grow equal to the size of the vRAM, which if you’re increasing a VM or multiple VMs at once can fill a datastore.
Final thoughts, hopefully these tasks are automated but good to remember it’s not just the VMware environment that you are working on.
- Update any monitoring tools if required to ensure the counters are accurate
- Update your CMDB to reflect the changes
- Update any design/build documentation if required