How to Setup and Test Failover of Parallel concurrent Processing PCP

Visit the Below Website to access unlimited exam questions for all IT vendors and Get Oracle Certifications for FREE
http://www.free-online-exams.com

Problem:     How to Setup and Test Failover of Parallel concurrent Processing PCP

Solution:

·         Setup the primary node with 11.5.10.2 ATG RUP6 installed and having CP,Admin,Forms,Web and DB set on it.

·         Add and register a new secondary node which has just CP and Admin on it. (You can use shared APPL_TOP  for this setup)

·         If using Virtual Hostnames, please ensure that the Physical Hostname of the CP node with virtual host is registered from OAM.

·         Once both nodes are ready, set the APPLDCP variable to ON in the context file of both nodes.

·         Shutdown the application tier services of both the nodes and run AutoConfig on each node. After AutoConfig completes successfully verify that the tnsnames.ora (on both CP nodes) has the FNDFS entries of both the nodes.

·         Ensure that the Internal Monitors on both nodes defined properly and have workshifts assigned to them. Also make sure the Internal Monitor manager is activated by going into Concurrent -> Manager -> Adminitrator and activate the manager as they need to be active on the nodes where the ICM can start in case of a failure..

·         Define the Primary and Secondary nodes for your Internal Concurrent Manager and Standard Managers (Concurrent Manager Define form)    

To ensure that failover of the concurrent managers when one concurrent node fails happens in time,ensure following steps are followed :

·         The sqlnet.expire_time in sqlnet.ora is set to 1. This is to be done for the database listeners (not apps listener).

·         Ensure that One-off Patch 6495206 and Patch 9074947 have been applied.

Manager Setup :

Assuming the 2 CP nodes defined on the environment are Primary Node1(CP,Web,DB,forms,Admin) and Secondary Node2(CP,Admin) the below table gives a snapshot of few of the Manager definitions used for the failover testing.

Manager Name	Primary Node	Secondary Node
Internal Manager (ICM)	Node2	Node1
Standard Manager	Node2	Node1
Conflict Resolution Manager(CRM)	Node2	Node1
Scheduler/Prereleaser Manager	Node1	Node2

Failover Tests :

Failover is the process in which the concurrent managers are migrating from primary node to secondary node because of a node or listener failure. Failback is when the primary node becomes available again and the concurrent managers need to migrate back to original primary node.

Use the below scenarios to ensure that PCP setup has been done correctly and the failover functionality is working as expected.

Please note that since PCP here is configured in a non RAC environment you cannot have an instance failover , that is if your DB Node is down the concurrent managers will not failover.

Primary Testcase: Node does respond to OS ping anymore.
Please note for the current setup we can only test if Node 2 does not respond to the OS ping anymore .This can be done by disconnecting the network cable for the Node2 and subsequently issuing an operating system ping on the machine name to ensure that it is not available/reachable anymore.
Expected Result :
As a result of this the managers running on Node2 (including ICM,CRM, etc, as defined in previous table) are all migrated to Node1. Once the Node2 is available again, i.e responds to the OS ping, the managers failback to it.

The Node pingable test as mentioned above is the ideal/recommended testcase to be used for checking PCP failover in non-RAC instances. The alternative testcase explained below should be used just to check if the PCP setup is done correctly for having a failover/failback of concurrent managers.

Alternative Testcase: Listener Failure test:

a. When both Nodes have Apps Listeners running and respond to OS ping, and we try to kill the ICM process by issuing a kill -9 command for the FNDLIBR process id , the ICM will start on its primary node again(Node2) and not failover to secondary node(Node1) because Node2 is still available and pingable.

b. When Apps Listener on Node2 is brought down and we try to kill the ICM process by issuing a kill -9 command for the FNDLIBR process id , the ICM will still start on Node2 however the Service Manager on Node2 will go down . As a result ICM will not be able to spawn the concurrent managers running on Node2 anymore, so the managers on Node2 will failover to Node1. Once the Apps Listener is brought up (using adalnctl.sh) on Node2 these managers having primary node defined as Node2 will failback to Node2.

References:

 Note 743716.1 How to Setup and Test Failover of PCP on Non-RAC Environments.

Get Oracle Certifications for all Exams
Free Online Exams.com