Hi
I have 2 Dell R610's and a MD3200i which I have up and running using the Dell guides for running ESXi and multi pathing. I have cabled the R610's direct into the MD3200i and left the default addresses on the storage array. I've a diagram of my setup at http://www.rckeith.co.uk/images/stories/vmware/vm.jpg
Hosts are configured as screen shot shows http://www.rckeith.co.uk/images/stories/vmware/esxi_host_fw190d.jpg and http://www.rckeith.co.uk/images/stories/vmware/esxi_host_spitfire.jpg
When I try to migrate a VM I get the "ESX hosst not able to connect over vmotion network". Which looking at how I've cabled up the iSCSI makes sense Do I need to use a single subnet
You need to enable vmotion on ONE vmkernel interface on each host, they need to be in the same subnet and should have a short path,
You surely tried to enable it on more than one uplink then it vanish from the previous one, this makes it easy to mess it up.
I suggest you add a dedicated vmkernel for vmotion (use a dedicated subnet for it) to vswitch0 and override the Teaming settings on the managment network to but vmnic3 to standby group and put vmnic0 to standby group on the new created vmotion network - this isolate the vmotion traffic a bit from disturbing managment network traffic.
Thanks for your advice, but I do have a dedicated vmkernel for vmotion. I followed Dell esxi configuration white paper http://www.dell.com/downloads/global/products/pvaul/en/powervault-md32xxi-deployment-guide-for-vmware-esx4.1.pdf
Spent most of today trying to get it to work, but I'm coming to the conclusion that I can't get it to work with direct cabling and failover. The best document I seen so far is this
http://www.delltechcenter.com/page/VMware+ESX+4.0+and+PowerVault+MD3000i which I think still applies to the MD3200i. I don't think I can achieve that with direct cabling.
"Thanks for your advice, but I do have a dedicated vmkernel for vmotion. I followed Dell esxi configuration white paper http://www.dell.com/downloads/global/products/pvaul/en/powervault-md32xxi-deployment-guide-for-vmware-esx4.1.pdf
Spent most of today trying to get it to work, but I'm coming to the conclusion that I can't get it to work with direct cabling and failover. The best document I seen so far is this
http://www.delltechcenter.com/page/VMware+ESX+4.0+and+PowerVault+MD3000i which I think still applies to the MD3200i. I don't think I can achieve that with direct cabling.
Keith"
On the MID3000i? If each individual host involved is attached to each controller in a dual controller setup, I can't see why this would not work. The error your getting indicates the hosts are not communicating on your dedicated vmotion network. Are the hosts direct connected to each other on that dedicated vmotion net or do they all go to a gig switch?
We have a small data center with two hosts direct attached to the MD3000i that services it and vomotion works fine. I share my dedicated vmotion network across all my data centers so there are lots of hosts attached to the vmotion network switches on my side.
Hi
The host are direct connected into the MD3200i on nic 1 and 4.
Today I connected a Gig switch between the hosts on nic1 and 4 to the MD3200i port and reconfigured as
http://www.delltechcenter.com/page/VMware+ESX+4.0+and+PowerVault+MD3000i and everything works fine.
I'm a bit puzzled as why I can't get it to work by direct connection. Would you be able to share your config so I can compare and see if I've got something wrong. I would like to go back to direct cabling if possible
Host 1 (iSCSI Connections)
Vswitch1 - 192.168.130.4 hooked to MD3000i Controller 0 iSCSI IN Port 0 with default 192.168.130.101 on the port
Vswitch3 - 192.168.131.4 hooked to MD3000i Controller1 iSCSI IN Port 1 with default 192.168.131.102 on the port
Host 2 (iSCSI connections)
Vswitch1 - 192.168.131.8 hooked to MD3000i Controller 0 iSCSI IN Port 1 with default 192.168.131.101 on the port
Vswitch3 - 192.168.130.8 hooked to MD3000i Controller1 iSCSI IN Port 0 with default 192.168.130.102 on the port
My Vmotion network is 192.1.1.0/24
Host1
Vswitch2 - 192.1.1.6 (Vmotion dedicated)
Host2
Vswitch2 - 192.1.1.8 (Vmotion dedicated)
I checked this post many times, I think I have everything right. Pretty easy to mix it up and post it wrong. I did give it a good hearted effort to doccument it correctly though.
Hi John
Thanks for your support. I've realised where I was going wrong. I was trying to use the iSCSI direct attached network for vmotion. yes the host were direct connected to the vmotion network. Using the default setup of MD3200i the hosts couldn't connect as they were on different subnets. I've now moved the vmotion to another nic pair and direct attached the MD3200i to the hosts again and everything works properly. Thanks once again, your post made me realise were I was going wrong.
Hello Keith, I think there is a slight design flaw in your iSCSI IP assignment. Normally you would have one of each subnet on each of the ESX hosts. To me it looks like you have the same subnets on the one host, eg: FW190D has 192.168.131.x and Spitfire has 192.168.130.x.
What I think you should have is 192.168.130.1 and 192.168.131.1 on FW190D and 192.168.130.2 and 192.168.131.2 on Spitfire (or visa versa).
From my understanding this will give better iSCSI performance, correctly load balance the traffic between controllers and if you place a switch in there in the future it will isolate these redundant paths between the two switches.
It also looks like you have a spare NIC on each server, which you could use for the vMotion subnet. That would isolate vMotion traffic from your production traffic.
In my setup I find that vMotion is very infrequent and not time critical, so I place vMotion onto the same vSwitch as the Vitrual Machine data. In your setup you have six NICs, so you could extend the iSCSI performance by adding one or two NICs (192.168.132.x & 192.168.133.x). If it was me I would configure the three NICs for iSCSI and the other three NICs for vMotion and Virtual Machins Data. This would give a good balance between iSCSI performance and Virtual Machine performance, plus fully utilise the NICs. Unless you are going to be vMotion heavily it will also provide redundancy for your vMotion traffic. The key is performance and redundancy. If you identify any paths that can fail then you can improve on that design, eg: SW7224-2 fails you just lost management connection between that host and the MD3200, plus I suspect you just lost vMotion, but by putting another connection to sw7224-6 you just made improvement on redundancy without any extra hardware.
Thanks for your reply. Yes I think you have a point. I did try that setup but I had problems seeing all the disk. I think maybe I had a configuration error. I will try again. I won't be doing very much vmotion so I have moved that off to another nic pair and that works OK.
MarioGrunert
2 Posts
642
0
Posted January 24th, 2011 01:00
You surely tried to enable it on more than one uplink then it vanish from the previous one, this makes it easy to mess it up.
I suggest you add a dedicated vmkernel for vmotion (use a dedicated subnet for it) to vswitch0 and override the Teaming settings on the managment network to but vmnic3 to standby group and put vmnic0 to standby group on the new created vmotion network - this isolate the vmotion traffic a bit from disturbing managment network traffic.