I'm working on a large-scale netbooted cluster configuration, which has it's persistent data stored on a clustering filesystem for redundancy. There is only one physical network link, where the untagged vlan is used for management/filesystem replication, and tagged vlans are used for other things.
Is there a good way of getting this interface set up in a basic fashion using OVS before the OVS database server is running? It looks like I can't simply use the underlying physical device with an IP address attached as normal, as apparently OVS takes complete control of the device (or so I hear). For example I'm hoping maybe I could do something like this: 1. Netboot happens 2. Interface eth0 (the only physical link) is setup as basic as possible to allow a sub-interface eth0.mgmt (which only needs to allow untagged packets) to be created. 3. Run "dhclient eth0.mgmt" to get an IP for the node (the mac address on this interface will need to match the real MAC of the NIC). 4. Start clustering filesystem, which uses the eth0.mgmt interface and associated IP address. 5. Start ovsdb-server on one node (on a floating IP which would be an alias on eth0.mgmt), with a persistent storage data directory. If this server dies it will be automatically master elected (using our internal automation framework) on another node and the floating IP moved to the new node. 6. Start ovs-vswitchd on all nodes, pointing them to the single ovsdb-server IP. 7. Do any other things needed to tell the running OVS daemon what I did before it started, assuming it can't examine the state of the system to figure out the sub interface I created. Assuming this is possible, how would I go about doing #2 and #7? If the above isn't possible, any other ideas how to get OVS working in this environment?
_______________________________________________ discuss mailing list [email protected] http://openvswitch.org/mailman/listinfo/discuss
