I recently added a number of vdi images from virtualbox to kvm in unraid.

These are two windows 2016 machines.

I did NOT convert them, just added them as their .vdi images.

I start them up, log on and all is fine until I start using them under any sort of load.

One of them (VASENT3-32bit) - the main database server is dropped from VNC Viewer, 

They have two CPU's each that are not used by anything else.

I can no longer ping the machine so it is gone, but on the unraid dashboard, they are both still showing as runnng.

The VMS tab in unraid however is unresponsive and shows nothing but the jumping lines forever.

The ONLY way I can recover this is to go into settings, turn off VM's, delete the libvirt.img file, then restart and re-add the vm's.

A turn off/on on its own of the VMs does not help - has to be a delete and re-create - which is a pain as I have about 10 VM's configured there.

Diagnostics is attached.

Anyone give me an idea of what is going on ?

 

This is an extract from the libvirt.log :

 

2020-04-09 10:15:34.467+0000: 14305: error : qemuDomainObjBeginJobInternal:7209 : Timed out during operation: cannot acquire state change lock (held by monitor=remoteDispatchDomainGetBlockInfo)
2020-04-09 10:15:46.176+0000: 14304: warning : qemuDomainObjBeginJobInternal:7187 : Cannot start job (modify, none, none) for domain VASENT3-32bit; current job is (query, none, none) owned by (14306 remoteDispatchDomainGetBlockInfo, 0 <null>, 0 <null> (flags=0x0)) for (1287s, 0s, 0s)
2020-04-09 10:15:46.176+0000: 14304: error : qemuDomainObjBeginJobInternal:7209 : Timed out during operation: cannot acquire state change lock (held by monitor=remoteDispatchDomainGetBlockInfo)
2020-04-09 10:16:04.468+0000: 14308: warning : qemuDomainObjBeginJobInternal:7187 : Cannot start job (query, none, none) for domain VASENT3-32bit; current job is (query, none, none) owned by (14306 remoteDispatchDomainGetBlockInfo, 0 <null>, 0 <null> (flags=0x0)) for (1306s, 0s, 0s)
2020-04-09 10:16:04.468+0000: 14308: error : qemuDomainObjBeginJobInternal:7209 : Timed out during operation: cannot acquire state change lock (held by monitor=remoteDispatchDomainGetBlockInfo)
2020-04-09 10:16:56.961+0000: 14303: error : qemuMonitorIORead:611 : Unable to read from monitor: Connection reset by peer
2020-04-09 10:19:00.047+0000: 17921: info : libvirt version: 5.1.0
2020-04-09 10:19:00.047+0000: 17921: info : hostname: Tower
2020-04-09 10:19:00.047+0000: 17921: warning : qemuDomainObjTaint:7986 : Domain id=1 name='VASENT3-32bit' uuid=c752cd72-ff12-d630-06be-91c1200f1974 is tainted: high-privileges
2020-04-09 10:19:00.047+0000: 17921: warning : qemuDomainObjTaint:7986 : Domain id=1 name='VASENT3-32bit' uuid=c752cd72-ff12-d630-06be-91c1200f1974 is tainted: host-cpu
2020-04-09 10:20:14.826+0000: 17922: warning : qemuDomainObjTaint:7986 : Domain id=2 name='VASDEP3-32bit' uuid=65d3d573-4a1f-fb65-aa6b-a721ad2ca435 is tainted: high-privileges
2020-04-09 10:20:14.826+0000: 17922: warning : qemuDomainObjTaint:7986 : Domain id=2 name='VASDEP3-32bit' uuid=65d3d573-4a1f-fb65-aa6b-a721ad2ca435 is tainted: host-cputower-diagnostics-20200410-0830.zip

I seem to have fixed this issue.

I converted these two vdi images to qcow2 with :

qemu-img convert -f vdi -O qcow2 VASDEP3.vdi VASDEP3.qcow2

qemu-img convert -f vdi -O qcow2 VASENT3.vdi VASENT3.qcow2

As qcow2, the servers are stable and dont crash and then bring down the VM system in unraid.

Maybe there is some sort of compatibility issue with vurtualbox VDI files ?

I have now converted all VM's to qcow2.  I have a shedload, so I am never going to do the vdi thing again.

You should probably post this under bug reports or a mod should move this to bug reports.