RHEV-M offers spice session on wrong (none display) ip address:port

Latest response

 Hi guys,

For this problem I already created a support case, because I think it is a bug in RHEV 3.1 (00818325). I just wanted to share this with you.

I have the following setup running (ip addresses are fictitious):

RHEVM-M server:
-- rhevm network: 10.4.10.75/24
-- rhevm default gw: 10.4.10.1

RHEVH 01 hypervisor:
-- rhevm network: 10.4.10.72/24
-- rhevm default gw: 10.4.10.1
-- display network: 172.16.2.15/27

RHEVH 02 hypervisor:
-- rhevm network: 10.4.10.73/24
-- rhevm default gw: 10.4.10.1
-- display network: 172.16.2.16/27

Client (Fedora):
-- rhevm network: 10.4.10.199
-- rhevm default gw: 10.4.10.1

In the cluster config the logical network "display network" has been assigned the role "Display Network".

When ever I try to connect to the Spice console (or VNC) RHEV-M offers me a session on a port on the wrong IP address, ~/.spicec/spicec.log:
- - - - - - - - - - - - - - - - - - - - - - - -
1365576329 INFO [24060:24060] Application::main: starting 0.8.1
1365576330 INFO [24060:24060] init_key_map: using evdev mapping
1365576330 INFO [24060:24060] MultyMonScreen::MultyMonScreen: platform_win: 54525953
1365576330 INFO [24060:24060] GUI::GUI:
1365576330 INFO [24060:24060] ForeignMenu::ForeignMenu: Creating a foreign menu connection /tmp/SpiceForeignMenu-24060.uds
1365576330 INFO [24060:24060] Controller::Controller: Creating a controller connection /tmp/spicec-NIPMXW/spice-xpi
1365576333 INFO [24060:24061] RedPeer::connect_unsecure: Trying 10.4.10.72 5901
1365576333 INFO [24060:24061] RedPeer::connect_unsecure: Connect failed: Connection refused (111)
1365576333 WARN [24060:24061] RedChannel::run: failed to connect: Connection refused (111)
1365576333 INFO [24060:24060] main: Spice client terminated (exitcode = 3)
- - - - - - - - - - - - - - - - - - - - - - - -

Why is a spice session offered on 10.4.10.72:5901? No spice sessions run on that IP, RHEVH 01:
- - - - - - - - - - - - - - - - - - - - - - - -
[root@hit-demo-rhevh01 admin]# netstat -lnp|egrep "^tcp.*qemu-kvm"                                    
tcp        0      0 172.16.2.15:5900         0.0.0.0:*                   LISTEN      13809/qemu-kvm
tcp        0      0 172.16.2.15:5901         0.0.0.0:*                   LISTEN      13809/qemu-kvm
tcp        0      0 172.16.2.15:5902         0.0.0.0:*                   LISTEN      8718/qemu-kvm
- - - - - - - - - - - - - - - - - - - - - - - -

The temporary workaround is to set the displaynetwork on the rhevm network. We use tagged VLAN's to seperate rhevm and displaynetwork.

Is this a bug in RHEV 3.1?

 

-- Vincent

Responses

This is a known bug if display network is configured as non-vm network. This will be fixed in a later release.

You can work around it for now by re-configuring display network as a vm network.

Thnx for you fast response!