i’m upgrading 2 PowerEdge R620 servers at work. both had originally 32GB of RAM, 10x 500GB SATA disks running in RAID10 and run vmware esxi. in the first step i’ve upgraded esxi from 5.1 to 5.5 [ bye bye 32GB memory limit ], in the 2nd – i’ve added 8x DELL SNP20D6FC/16G memory sticks to each of the servers [ according to dell’s manuals larger – 16GB units should go to slots with lower numbers – A1, B1, A2, B2, A3, B3, A4, B4; and smaller – old 8GB sticks – to A5, B5, A6, B6 ]. that worked flawlessly.
in the next step i want to migrate away from RAID10 on 10 disks to RAID10 on 8 drives + RAID1 on 2x 600-800GB SSDs [ most probably intel’s s3500 ]. i’ve taken a backup of all vms, rebooted the machine, entered bios, removed the old array and recreated new one. system booted finely [ esxi with all its settings sits on dual sd cards so there was no need to reinstall and reconfigure it ], but i could not create a new datastore on the now smaller RAID10. apparently some leftovers from old GPT prevented vmware from correctly recognizing the available virtual disk. i was getting:
Call "HostDatastoreSystem.QueryVmfsDatastoreCreateOptions" for object "ha-datastoresystem" on ESXi "serverName" failed.
solution? boot to some live cd [ i’ve used gparted live ], and in console use parted and create gpt label on the virtual disk provided by raid:
sudo bash # make sure i'll work on the right device dmesg|grep sda parted /dev/sda mklabel gpt quit reboot
after reboot vmware let me to create new data store without any errors.