Issues arise when clustered applications are started manually outside of an active cluster.
- The service may recognise that it is clustered and fail to start with a error stating it is a clustered application and must be started under cluster control.
- The application may depend on a virtual context and if the virtual server name via the Lanman resource is not available the service may not start, or may start in the context of the local server.
- If needing to test remote connection then the virtual IP addresses may not be available for client connection
Errors vary, and many application and related services messages can be viewed in the application and system event viewer logs respectively.
Starting applications while the cluster service (High Availability Daemon - HAD)is running.
As a general rule, only online/offline clustered apps via the Veritas Cluster console.
Should there be issues with a specific clustered resource, process or application, then follow the procedure below:
1. Stop the cluster service using the force command e.g. hastop -local -force, on the specific node with the issue. This ensure that other active service groups remain online and do not fault due to absence of the cluster service. It may be necessary to online the VMDG, IP or Lanman resources (or any other supporting resources) before issuing this command depending on the test being carried out.
2. Start or troubleshoot the process or application directly e.g. Oracle Listener service via Windows Services
3. Once the process or application is working or troubleshooting has been completed, start the Veritas Cluster service again e.g hastart
It is not best practice to troubleshoot or start/stop applications which are part of an active cluster; as this will hamper the correct functionality of Veritas Cluster Server and may limit correct failover functions.