Bug 1094882

Description of problem:

After a redundant storage path outage, customer has 2 hypervisors with VMs reporting as paused, but are in fact still running.


Version-Release number of selected component (if applicable):

* vdsm-4.13.2-0.13.el6ev.x86_64
* Red Hat Enterprise Virtualization Hypervisor release 6.5 (20140407.0.el6ev)

How reproducible:

Not reproducible at this point.  We have been having to rely on customer environment.

Steps to Reproduce:

* Disabled one of the switch network redundant paths

* RHEV detected a fault and started automatically migrating everything at once, overloading the network.

* now many virtual machines are set in a paused state, but they are not paused, and we can't unpause them, can't migrate them, can only shut them down

Actual results:

VMs in paused state according to RHEVM, vdsClient, but in gfact running, and reported as running via virsh directly on host.

Expected results:

vdsm reads state from qemu/libvirt properly

Additional info:

Logs will be attached soon.

Here are the steps we used to attempt to get the status updated:

* Unpause from RHEVM - no change

* Set state to up in DB directly - changed back by vdsm

* Shutdown supervdsm, vdsm, and libvirtd on host, restarted all. - VMs still reporting paused

* Via virsh directly, all VMs report "running", and are reachable on the network.