Hi all
I were testing the new RPC calls and it seems that some of them have changed so we have found some backwards incompatibilities between the 4.x and 5.x versions. In our case the most important change is related with the host management.
it seems that one.host.allocate has changed since the last version:
http://docs.opennebula.org/5.0/integration/system_interfaces/api.html
http://docs.opennebula.org/4.12/integration/system_interfaces/api.html
Using the 4.x rpc call we get this error:
[ERROR] _rpc failed to make request faultCode -501 faultString Parameter that is supposed to be integer is not method one.host.allocate args [string, hyp101.altaria.os], [string, kvm], [string, kvm], [string, ovswitch], [int, -1]
The reason is that vn_mad is not used by the new onehost command, and also the VNET template should be changed. This is done on purpose? It’s not commented in the compatibility guide and it breaks the compatibility with some RPC clients.
This change also raises another question, how is possible now to setup an heteregenous hyp cluster with OVS and linux bridges hyps? In OpenNebula 4.x we were able to setup the vn_mad for each host, so we were able to include linux and OVS bridges in different hypervisors, thanks to OVS_BRIDGE variable, as example we had for an specific
VNET:
NAME = “vscnetwork”
BRIDGE=“br101”
OVS_BRIDGE=“ovsbr”
VLAN=“YES”
VLAN_ID=“295”
and we had hosts with dummy and also hyps with ovsswitch network drivers, so OVS_BRIDGE was used for OVS hyps, how is possible to keep this heterogeneous configuration now? It’s still supported?
Thanks in advance!
Alvaro