Setup and Configuration of the AGLT2 MD3600i - July 25, 2011
This details our installation and configuration of our new MD3600i (
UMVMSTOR02) plus MD1200 shelf. We received both units on July 22, 2011 and both were installed in Rack4-2 at 31-34 U. A diagram on the network and power connections is shown in this figure:
- UMVMSTOR02 power and network connectivity diagram:
In addition to the physical setup there were a number of additional configuration details to attend to:
- Setup of storage and associated LUNS
- Creation of Host Groups to manage shared access via iSCSI
- Integration with AGLT2's VMware system
There are some useful documents that describe most of the details:
A number of items were configured to optimize our setup for use with VMware. Summarizing them:
- Made sure we had the most recent firmware installed (it arrived with the most recent)
- All ESX hosts, switches and the MD3600i were configured to use jumbo frames of 9000 bytes
- The ESX hosts had a VMkernel device "bound" to each of the available 10GE nics (vmnic12 and vmnic13)
- Made sure the MD3600i is recognized by VMware. On each ESX host do:
- The MD3600i had the following disk groups created:
- Three AFS disk groups, each RAID-10 and enclosure redundant (4 disks each): AFS_vicep[e|f|g]
- Two VM images disk groups, each RAID-5 (5 disks each): VMImages and VMImages2
- The "High-performance Tier" and "Snapshot and Disk Copy" licenses were loaded on the MD3600i
- We installed Dell's MD VMware plugin in our vCenter which allows easier management of the iSCSI system
- Note this is available from the regular support/downloads area for the MD3600i
- Enabled the ESX hosts access to the relevant disk-groups and LUNs on MD3600i via the Dell MD plugin
- Setup host-group VMware_AGLT2 with UMVM01/02/03 as members
- Created VMFS 3.46 datastores for AFS_vicep[e|f|g] and VMImages/VMImages2 from UMVM01 are rescanning
- Rescanned iSCSI on all ESX hosts
- All ESX hosts had their storage adapters configured to use "Round-Robin" path selection on all iSCSI targets
- Forced all disk-groups back to their preferred controllers
- The VMImages and AFS_vicepf areas are served via controller #1 while the rest are service via #0
--
ShawnMcKee - 25 Jul 2011