Video Screencast Help
Symantec to Separate Into Two Focused, Industry-Leading Technology Companies. Learn more.

Oracle Shutdown During Failover with Link Based IPMP

Created: 28 Feb 2012 | 3 comments

I am running an active/active VCS cluster (6.0) running Oracler(HA) in one service group and WebLogic Server (WLS) in a second. Oracle service group is mastered on node 1 and WLS on node 2. We are using link based IPMP with MultiNICB. Failover of the service groups appear to function well with one exception. If ... the Oracle and WLS service groups occupy the same node and the Oracle VIP is on :2 with WLS VIP on :1 and the WLS service group is moved to the other cluster node, Oracle shutsdown. No errors, the cluster frame work just reports that Oracle shut itself down. If the Oracle VIP is on :1 and WLS VIP is on :2, when WLS is moved to the other cluster node, all is well. This problem only occured when we moved from MultiNICA to MultiNICB with IPMP. So, we are looking at this being an issue with MultiNICB with link based IPMP. The Oracle Listener appears to be OK and does not shutdown. Oracle has a dependency on the Listener being active.

Below are the MultiNIC statements from the main.cf:

 IPMultiNICB ip_app_somenet (
  BaseResName = mnic_app_somenet
  Address = "xx.yyy.47.184"
  NetMask = "255.255.255.0"
  )

  MultiNICB mnic_app_somenet (
  UseMpathd = 1
  MpathdCommand = "/usr/lib/inet/in.mpathd -a"
  ConfigCheck = 0
  Device @someapp01-c01 = { nxge0 = "", nxge8 = "" }
  Device @somedb01-c01 = { nxge0 = "", nxge8 = "" }
  IgnoreLinkStatus = 0
  )

  IPMultiNICB ip_ora_somenet (
  BaseResName = mnic_ora_somenet
  Address = "xx.yyy.47.183"
  NetMask = "255.255.255.0"
  )

  MultiNICB mnic_ora_somenet (
  UseMpathd = 1
  MpathdCommand = "/usr/lib/inet/in.mpathd -a"
  ConfigCheck = 0
  Device @somedb01-c01 = { nxge0 = "", nxge8 = "" }
  Device @someapp01-c01 = { nxge0 = "", nxge8 = "" }
  IgnoreLinkStatus = 0
  )

Comments 3 CommentsJump to latest comment

Ron Cline's picture

Additional information on IPMP Issue:  This is using Veritas 5.0 MP3 and Oracle 11.2.0.2.0.  The WLS (Application) service group includes Weblogic 10.3.2, custom applications, and application filesystems.  The Oracle Resource includes Oracle engine, listener, and Oracle filesystems.

jstucki's picture

I think your problem is having multiple (two) MultiNICB resources for the same IPMP NIC devices.  You should only have 1 resource for that pair of IPMP NICs.

I typically put my MultiNICB resource in the ClusterService service group.  Then I put Proxy resources in the Oracle and WLS service groups, to represent the MultiNICB resource.  Some people like to create a separate parallel service group for the MultiNICB resource.  I don't see a reason for it.  I haven't had any issues in putting it in the ClusterService service group.  Creating the separate parallel service group just adds more complexity, in my opinion, and then you have to add service group dependencies.

Give this a try, and see if it solves your problem.

-John