2011-10-12 01:42 AM
Our team build yesterday our first Microsoft HyperV solution
Base on FAS2040 Ontap version 8.0.2 and 4 HP servers in one cluster
We follow TR july 2011 (HyperV and NetApp)
So all the LUN made are Window2008 (we don't use snapdrive), igroup type Windows
After start a VM store in CSV, a few minutes later, this VM Hang
Sector 0 Partition crash
With Microsoft support help ,we were able to rewrite it, but now Microsoft support says
May be it's our DSM MPIO or wrong LUN type that is the culprit
They saw some problem with scsi reservation lock ?????
Anyone here has already heard about this?
If LUN type is wrong, the main issue possible will be misalignment and performance issue
But can this break Windows partition table ?
I know that Microsoft CVS is a bit tricky
Your help will be appreciate
2011-10-12 02:41 AM
Are you using NetApp MPIO or Windows native MPIO?
Have you enabled ALUA on the igroups?
You have a GPT partition style and only one big primary partition on it?
You have any other troubles with that cluster, can you browse the C:\ClusterStorage folders fine from all nodes?
2011-10-12 02:54 AM
We installed NetApp MPIO, but remove it for using only MS Native MPIO
As it's iSCSI LUN we don't enable ALUA
Yes it's on big 800 GB GPT partition
And we were able to browse the clusterstorage from all nodes
the big problem appeared a few minutes after starting VMs stored in c:\clusterstorage
according Microsoft support, they suspected scsi reservation or lock issues
My question ,what is the impact of a bad LUN type choice?
As we create WIndow2008 type
We use the last OnCommand System Manager (version 2.0R1), and funny thing, when creating LUN only Windows2008 is available
No HyperV options (like with System Manager 1.1 ou CLI)
But as the TR-3702 says Windows2008 is the right choice (we do not use SnapDrive)
2011-10-12 03:09 AM
as far as i know windows_gpt, windows_2008 and hyper_v are the exact same LUN type, its just for the purpose of having proper reporting how many windows 2008 or how many hyper-v LUNs the customers have and so on.
i prefer microsoft native over dsm, so you fixed that already, besides alua no on iscsi is correct as well. As for reservation conflicts and so on, have you guys properly set up the windows cluster? you are using a quorum i assume and you have NOT disabled Client for Microsoft Netorks on the cluster network interfaces?
Besides when using iscsi i think its recommended to go with target portal groups. Shouldnt case stale scsi reservations tho i guess.
2011-10-12 03:19 AM
Thomas , again thanks for your very fast reply
We're going to follow your advices, for Native MPIO
I work for SAN & virtualisation team (VMware)
It's the firt time that Windows team implements HyperV ,and they try to say "SAN problem" :-)
We need to double check Cluster config
As one guy disable for exemple IPV6 on Windows network inferface, a second said no, you must keep IPV6
I'm not a Windows girl :-)
I'd like to be sure it's not NetApp SAN misconfiguration
2011-10-12 03:51 AM
Just fyi, unlike in previos 2000 and 2003 clusters you MUST keep IPv6 enabled and you MUST keep Client for Microsoft Network enabled as CIFS/SMB plays an important role in a Hyper-V cluster :-)