Oh, learning on the job with production hardware...
I have a fileserver VM that boots from one datastore and keeps its giant pile of data on another. I created a snapshot of the computer to perform a hot backup. Everything was going well until I got the terrible "There is no more space for the redo log" error. The differential data was being placed on the first (very small) datastore and had filled it to the brim.
"Oh, I'll just delete the snapshot and try again later." Ha.
I'll skip the details that everyone else has already covered and get to the part that counts. When I was browsing around the datastore configuration, I saw a little box that said datastores could span multiple extents. (ie: hardware RAID sets) I found some fugly old SCSI drives (one was 11 years old!) and scraped together a 9GB RAID-1 on the second channel of my SCSI card. I added this extent to the existing datastore, providing a litlte breathing room for the merge process. I'm running ESXi 3.5, so I don't have any of the fancy backup tools and partition managers. I created a new snapshot of the machine, and then pressed Delete All Snapshots. Some nail biting and lots of disk grinding later, all of my disks were merged and I didn't have to commit seppuku in the server closet.
Note: after you add an extent to a datastore, you can't remove it. I've got an external power supply and two drives precariously placed inside a milk crate near my server. I'm migrating to new hardware in a week or two so it's no big deal for me. (The old physical file server will be phased out soon - when it goes away, its hardware will become the final home of the virtual machines.)
