We are a small-time customer with a MSA 1500 fc (2 shelfs) and a VMware vSphere 4.0 upd. 1 env. installed on 3x HP DL380 servers. Guest are with various OS all supported by VMware. We have a very annoying problem which we really need some help with from you experts out there!
The problem in a nutshell:
We suddenly experienced a very noticeable decrease in performance when we backed up our .vmdk files (guests) - approx 5% of normal speed. We have tried both vRanger, Veeam and a third program (can't remember the name)
1. When we cold migrate a guest between MSA LUN's the performance is extremely slow - also if LUN's is on diff. MSA disk-shelves.
2. If we migrate it to a QNAP NAS with a 1Gb NIC from MSA LUN's using NFS the performance would be considered to be at a normal rate/speed. If we reverse and migrate it back from QNAP to MSA it is again extremely slow.
3. If we try to copy a .iso from laptop to ESX folder through WinSCP the tranfer rate is 8MB p/s.
4. If we copy an .iso from a presented disk (located on MSA LUN) on a VM to another ditto the performance >30MB p/s - which would be considered acceptable in this scenario...
5. Users don't have any performance issues on a daily basis
A. Firstly we assumed the problem to be zoning in FC switch but it seems ok. We have experimented with not allowing arbitraded loops on the switch which didn't return any increase or decrease in performance.
B. Secondly we suspected the MSA but when we saw the result from above test 4 we found it more likely to be somewhere in the VMware env.?
Can anyone here please help us out? Are we correct when we point our fingers at the VMware env. instead of the MSA?
Firstly the MSA1500 Array is not in the supported list for ESX 4.0.
MSA 1500 with Active/Active firmware supports upto ESX 3.5.
What is the verison,measns active/active or active/passive for this MSA.
If it is active/passive, try to upgrade to active/active and create more datastores in which u may get some improvement inseted of active/passive arravs in which all IO is taken care by single controller.
