[one-users] "onevm delete" deletes extra VM

[Shi Jin]

Hi there,

I am running one-1.4beta with KVM hypervisor.
Recently I found a very bizarre behavior of "onevm delete".
For example, initially there are 5 VMs running:72,73,76,77,78.
Then I did "onevm delete 76" and found only 3 VMs are left: the VM 77
is also deleted.
"virsh list" on the host node shows VM 77 is gone and "onevm list"
shows its status to be unknown.
I think this is a serious bug and requires a quick fix.
I tried to dig into the OpenNebula log files but found nothing other
than " Command execution fail: virsh dominfo one-77" after the "onevm
delete 76" command.

On the host node, the syslog shows something interesting:
Sep 30 08:58:59 node1 kernel: [238101.304619] br1: port 5(vnet4)
entering disabled state
Sep 30 08:58:59 node1 kernel: [238101.342851] device vnet4 left promiscuous mode
Sep 30 08:58:59 node1 kernel: [238101.342854] br1: port 5(vnet4)
entering disabled state
Sep 30 08:58:59 node1 kernel: [238101.424851] br1: port 4(vnet3)
entering disabled state
Sep 30 08:59:00 node1 kernel: [238101.463031] device vnet3 left promiscuous mode
Sep 30 08:59:00 node1 kernel: [238101.463035] br1: port 4(vnet3)
entering disabled state
Sep 30 08:59:53 node1 libvirtd: 08:59:53.480: error : Domain not found

Please note that the VM 76,77,78 all use the same vnet (bridged on
br1) and their corresponding ports are vnet3,vnet4 and vnet5.  I am
not exactly sure what  the above message implies but I have a feeling
that something went wrong about the network. In this sense, this
problem may not be a OpenNebula bug, possibly a KVM bug too.

Actually this kind of problem has happened many times for us. This is
my first time to carefully record what has happened.
I am wondering if anyone has encountered this kind of problem before
and is there a fix?

Thanks a lot.

Shi
-- 
Shi Jin, Ph.D.