Problem with live migration

Latest response

Respected Sir,

 

I have RHEVM manager version 3.1.4

and 2 RHEV-H Host 6.4

1) first rhev-h node name "vm1"

2) second rhev-h node name "vm2"

both the rhev-h host are in the single cluster

Now i have create 1 vm on "vm1 rhev-h node"

and try to manuall  migratie to the second node "vm2" it working fine

now for testing i have disconnected the network from my one of the rhev-h host "vm2" on which the vm is running

so on the manager it shows that  rhev-h host vm2 is down

but the problem is the vm which is running on vm2 node is not migrating automatically

do i required any special configuration for that

i have also keep that vm on high availabilty also

still it is not migrating automatically

Please give the solution for that

Regards

 

 

 

Responses

Hi Pratik,

Sorry the community hasn't been able to help you out with these issues. They are probably beyond the scope of what can be easily handled in a discussion forum so I would suggest that you open a support case for this, and for the issue you outlined in https://access.redhat.com/discussion/problem-while-extending-fiberchannel-storage-domain

I think by default the virtual machine will not be migrated unless RHEV-M can verify that the non-responsive host is not running the virtual machine.  Do you have power management enabled for both hosts?  For HA virtual machine power management is required.

Take a look at this document: https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Virtualization/3.1/html/Administration_Guide/High_availability_considerations.html

Accidentally double posted.

I don't think you need power management for HA to work but now I'm not so sure.  I've had HA virtual machines come up automatically and I definitely do not have power management turned on here.   But my scenario was different and at least for me, I found HA to be a little over-eager.  When I shut down an HA VM, it starts right back up again automatically and I have to uncheck the HA box to make it stay shut down.  I haven't tried killing one of my own hosts yet.  I am running RHEV-M 3.2 but my hosts, cluster, and datacenter are still at 3.1 compatibility.  

If I'm reading how you're testing:

Two hosts named vm1 and vm2.  

Virtual machines running on each host.  Manual virtual machine live migrations work fine.  Isolate host vm2 from the network and its virtual machine does not automatically migrate.  

I think that's the sequence of events.  

What are you doing for storage?  

Even though hosts vm1 and vm2 are isolated from each other on, say, the rhevm network, can they both still see your shared storage?  Maybe they can still kind-of see each other over the storage network.  

I wonder what would happen if you just manually killed the power to, say, host vm2, leaving host vm1 as the surviving host?  Thinking this through, I'm not sure you want the virtual machines that were running on host vm2 to automatically boot on the surviving host, because without power management, I don't know how RHEV-M and the surviving host can determine the state of the other host.

- Greg

Power Management is required for HA to work.

If you define  a vm is highly available and the hypervisor on which it runs goes down, then the question of migration does not come into picture. We cannot migrate a vm from a host which is down. We can only restart the vm on another hypervisor in the cluster.

To make sure that the vm does not end up running on multiple hypervisors, ie the hypervisor and went down and the new hypervisor that RHEV-M decided to restart it, we need to fence the down hypervisor using power management. So a successful power management is a pre-requisite for successful HA.

> When I shut down an HA VM, it starts right back up again automatically and I have to uncheck the HA box to make it stay shut down.

If this reliably reproducible, please open a case with details. An HA vm should not be automatically started after a regular shutdown.

> If this reliably reproducible, please open a case with details. An HA vm should not be automatically started after a regular shutdown.

OK.  Just opened case number 00905091.

- Greg

Close

Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.