Hi!
I have noticed a very odd behaviour of the Hyper-V agent while monitoring neutron agent-list output.
I have two Hyper-V hosts, they're both available in Nova and Neutron, however, about every 10 seconds the status of the HyperV agent changes from alive to dead and vice-versa.
I am using VLAN segmentation and here's my Get-NetAdapter output on one of the Hyper-V hosts:
Name InterfaceDescription ifIndex Statu s ---- -------------------- ------- ----- vEthernet (NovaNetwork) Hyper-V Virtual Ethernet Adapter #2 56 Up Port1 - VLAN 252 Microsoft Network Adapter Multiple...#3 57 Up Port1 - VLAN 251 Microsoft Network Adapter Multiple...#2 52 Up Port1 Microsoft Network Adapter Multiplexo... 29 Up Ethernet 3 HP FlexFabric 10Gb 2-port 534FLB ...#81 14 Up Ethernet 2 HP FlexFabric 10Gb 2-port 534FLB ...#80 13 Up Ethernet 4 HP FlexFabric 10Gb 2-port 534FLB ...#84 17 Di... Ethernet 5 HP FlexFabric 10Gb 2-port 534FLB ...#85 18 Di... Ethernet 7 HP FlexFabric 10Gb 2-port 534FLB ...#87 20 Di... Ethernet 6 HP FlexFabric 10Gb 2-port 534FLB ...#86 19 Di...
On the controller side I have Mirantis OpenStack 6.1 and the network setup is identical to this one: http://www.cloudbase.it/wp-content/uploads/2015/08/fuelnetlist.png with the exception of VLAN IDs.
Hyper-V server and OpenStack controller can "see" each other via all interfaces. VMs, depending on the state of the agent, either get network connectivity on boot (if agent is alive), or not (if agent is dead).
Could you please suggest where the problem might be?
Thanks!
Justinas
