H165895:  CLIENTS CAN NOT CONNECT TO AN ORACLE PARALLEL SERVER NODE


 SYMPTOM:

 Client sessions can no longer connect to one of the nodes in
 the database, and/or some rows or tables cannot be accessed
 from the other nodes.

 The node to which the client sessions can no longer connect
 cannot see the shared storage when viewing the drives with the
 NT Disk Administrator. This node had been up and running OK.

 This scenario has occurred without rebooting this node.

 PROBLEM ISOLATION AIDS:

 - The system is a Netfinity 7000 M-10 server, Type 8680,
  any supported Model configured with Oracle Parallel Server.
 
 Note: Supported configurations are listed at the following URL:
       http://www.pc.ibm.com/us/compat/clustering/matrix.shtml
       look for "Oracle Parallel Server"

  - The system is configured with the following option:

       IBM Netfinity Cluster Enabler Software

  - NOS affected:

       Windows NT 4.0 Enterprise Edition with
       Service Pack 4 applied.

 FIX:

 The action to take depends on the state of the
 OracleServiceOPSn service on the node that clients can no
 longer connect to. "n" will be the number of the database
 instance so the actual service name will be something like
 OracleServiceOPS1.

 - If the database instance has stopped and the node has been
 evicted from the cluster, then no action is required.

 - If the database instance has not stopped completely (even
 after 15 minutes), then the OracleServiceOPSn service should be
 stopped with the "net stop" command or from the Windows NT
 Services window.

 You can determine if the OracleServiceOPSn service has been
 stopped by using the Windows NT Services window. The "Status"
 field of the OracleServiceOPSn service will be blank when the
 service is successfully stopped. If the OracleServiceOPSn
 service cannot be stopped with one of these commands, then the
 ORACLE80.exe process should be terminated from the Windows NT
 Task Manager.

 DETAILS:

 A hardware failure may have occurred that prevents access to the
 shared storage on this node. If this has occurred, the database
 instance on this node may behave in one of two ways:

      1) The database instance terminates. This node is
 evicted from the cluster. The Oracle reconfiguration process
 resulting from a node being evicted from the cluster may take
 as long as 15 minutes to complete.

      2) The database instance terminates, but not completely.

 In both cases, provided the client application is written
 properly, the clients will failover from the node that can no
 longer communicate with the shared storage. In the first case
 above, no problems will occur. In the second case above, the
 ORACLE80.exe process may continue to use up CPU and memory on
 this node, but the clients should not see a problem. The fix
 addresses the problem by terminating an Oracle service that may
 not have shut down correctly and is hung holding a database
 lock.

 TRADEMARKS:

 Microsoft, Windows, Windows NT, and the Windows logo are
 trademarks of Microsoft Corporation in the United States and/or
 other countries.

 Other company, product and service names may be the trademarks
 or service marks of others.




DATE: February 17, 1999