Jump to content


Photo

VM loses network after migration

Started by Niklas Åhden , 21 April 2017 - 12:27 PM
4 replies to this topic

Niklas Åhden Members

Niklas Åhden
  • 258 posts

Posted 21 April 2017 - 12:27 PM

Hi,

 

We have 2 pools with 4 active physical hosts in each. The physical hosts are all identical with the following specs:

 

Dell Poweredge R630

2st Intel Xeon E5-2630 v3 @ 2.40GHz

256Gb RAM

Intel E350 Quadport -Port 1 & 2 are in a LACP where mgmt is used on the native VLAN. Then the VM networks are added as VLAN's ontop of that LACP. The LACP configuration used is LACP Active/Active where the mode is "LACP with load balancing based on IP and port of source and destionation".

Port 3 & 4 are used for storage - iSCSI with multipathing enabled against 2st Nexentastor machines.

 

The latest firmware pack has been installed on all hosts as well.

All servers are running XenServer 6.5 SP1 with all hotfixes applied.

 

The switches are 4st Juniper EX3300 in a stack of 2 switches each.

 

Now to the problem, which is really weird:

 

When doing a live migration of a VM running Debian Linux or Windows Server 2008 R2 the VM is fine afterwards. Even migrating Windows Server 2012 R2 works fine in almost all cases.

 

But when migrating a VM running Windows Server 2012 R2 which is acting as a remote desktop gateway (RDS Gateway) with the Windows Network Load balancer the VM looses network connectivity after being migrated within the pool.

To resolve it I simply have to go to the console, login and ping the default gateway and it is all good again.

 

My guess is that this is some what related to the Windows Network Load balancer but I am not really sure, I am out of luck and this is my last resort even tho I dont think this is XenServers fault at all :-)

 

So any suggestions are appreciated!

//Niklas


Edited by Niklas Åhden, 21 April 2017 - 12:56 PM.


Alan Lantz Members

Alan Lantz
  • 6,989 posts

Posted 21 April 2017 - 01:19 PM

Why do you need NLB ? Not sure how thats fitting into the picture.

 

--Alan--



Marc O'Neill Citrix Employees

Marc O'Neill
  • 198 posts

Posted 21 April 2017 - 01:47 PM

As with my other reply multicast/unicast isnt supported on XS. I seem to recall this caused specific problems with windows NLB as it uses unicast.



Niklas Åhden Members

Niklas Åhden
  • 258 posts

Posted 21 April 2017 - 01:53 PM

Why do you need NLB ? Not sure how thats fitting into the picture.

 

--Alan--

 

I have 2st RDGateway's which uses a NLB to share an IP-adress - This was according to some best practice when we did this setup a couple of years ago with Windows Server 2008 R2.

Maybe this is the wrong way to achieve a HA RDP-Gateway setup?
//Niklas



Niklas Åhden Members

Niklas Åhden
  • 258 posts

Posted 21 April 2017 - 01:56 PM

As with my other reply multicast/unicast isnt supported on XS. I seem to recall this caused specific problems with windows NLB as it uses unicast.

 

Oh really?
This is more than I know!

 

If I choose Cluster Properties and click on the Cluster Parameters-tab I can choose the following under Cluster operation mode:

 

Unicast (Selected)

Multicast

IGMP multicast

 

Do you think choosing another option (IGMP multicast) would resolve this issue or how do you suggest to achieve the same thing without using Windows Network Load balancer?
//Niklas