0 Replies Latest reply: Nov 21, 2012 11:49 AM by 974667 RSS

    Network configuration under 3.2.1 OVS and template problem

    974667
      Hi all,
      Sorry I changed my E-Mail address and actually they are not able to move my history from User "SPA2" to this one.

      I do a setup with the OVS early access version 3.2.1. After this setup I use a RAC template and after this I would do a deploycluster(.py) Setup by script.
      My Server has two network cards one for the maintenance, Cluster heartbeat and also for virtual machines. The second network card is for the RAC interconnect eth1.
      -----------
      So I setup eth0 which is setup by the server as bond0.
      ------------
      By config the OVM manager I do myEth0 as 172.16.0.x and myEth1 as 10.1.1.x network. So far so good.
      ------------
      After loading the RAC templates and change of the network parameters before doing the clone template job I configure reh racnode.0 and racnode.1 by setting up networks as
      myEth0 and myEth1. Also this works fine.
      -----------
      Then I start the deploycluster.py with the parameters -u admin -p xxxxx -M racnode.0,racnode.1 -N netconfig.ini and here with -D (for dryrun).
      -----------
      Now I saw on the OVS Server that a interface was created on the 172.16.0.20 (which is the server itself) and 172.16.0.151 for the first public IP of RAC node.
      But this was created as bond0:0 and is created after first dryrun of deploycluster.py script.
      -----------
      First of all I do the deploycluster.py ... -N netconfig.ini -B yes (which mean buildcluster = yes) but if I start this there will be no cluster build. Nothing happend!
      -----------
      While both VM's racnode.0 and racnode.1 are started I take a look to the first console.
      ------------
      And here I saw a conflict during the boot of racnode.0 cames a message the IP address 172.16.0.151 is in use!
      I don't understand why but I saw on the server that the bond0 interface starts bond0:0 with this IP address.
      Also it is not possible to do a ifconfig bond0:0 down while it is directly restarted by the server.
      ------------
      I try to start the /u01/racovm/buildcluster.sh but it also stops with network conflict and unknown hostname.
      ------------
      It seems that it is not possible to use the first network interface also for "virtual machines".
      ------------
      And here the questions:

      - How can I do a network setup for a RAC Cluster without using the first network card in the server eth0. It seems to be that it does not work correctly.
      Is it a 3.2.1 problem?
      - For me it looks like to setup a VLAN with for example 2 virtuell networks vlan1 for 172.16.0.x (RAC public and vip and scan) and vlan2 for 10.1.1.x (RAC-priv interconnect)
      - How can I configure this and who had a best practice for this?
      - Why does the deploycluster.py script not build the cluster?
      I started the deploycluster.py script on the Oracle VM Manager machine which is a separate machine.
      -------

      many thanks
      SPA2