This topic applies only to the IBM Business Process Manager Standard configuration.This topic applies only to the IBM Business Process Manager Advanced configuration.

Configuring the Elasticsearch service

You can configure some aspects of the Elasticsearch service behavior.

About this task

The Elasticsearch service on Process Federation Server provides a distributed index that contains the data from the IBM® BPM systems in the federated environment. The distributed index ensures that users have fast access to federated data, and relieves IBM BPM systems from expensive queries. There is an indexing service for each IBM BPM server in the federated environment, and a common REST-based query service that provides data to users. The Elasticsearch service can automatically rebalance the distributed index. For example, as you add more process federation servers to handle load, partitions of the federated index (shards) are automatically relocated to the new servers.
The server.xml configuration file on Process Federation Server contains configuration properties with default values for the Elasticsearch service. By default, the Elasticsearch service is configured for a single-server, quick-start environment. To configure the Elasticsearch service for a production environment with multiple servers, review the following considerations:
  • The cluster.name property must be the same on all servers. The Elasticsearch service uses this property to identify a group of servers as being in the same cluster.
  • The node.name property must be unique on all servers in the cluster.
  • The Elasticsearch service discovers other clusters members based on the configured TCP ports. By default, multicast discovery is disabled (discovery.zen.ping.multicast.enabled=”false”). Use unicast discovery instead so that you can control the exact composition of your cluster membership, by configuring a list of all cluster members in the discovery.zen.ping.unicast.hosts property. The list is a comma-separated list of other Elasticsearch services in the cluster that is based on the host TCP port.
  • The Elasticsearch service requires a quorum of cluster members to be available before it can service requests and perform indexing operations. To provide failover support, a production topology requires three or more process federation servers with Elasticsearch services. With three process federation servers, a server can go offline without impacting the quorum (two or more). Configure the discovery.zen.minimum_master_nodes property to indicate the number of servers that is required to form a quorum. For example, set the value to “2” if you have a total of three process federation servers in your cluster, or “3” if you have a total of five process federation servers in your cluster.

Procedure

  1. Open the server.xml configuration file for editing. By default, the configuration file is in the pfs_install_root/usr/servers/server_name directory on Process Federation Server.
  2. Update the ibmPfs_elasticSearch element with configuration properties for your environment. For information about the properties that you can set, see Elasticsearch service configuration properties.
  3. Save the changes to the server.xml configuration file and restart Process Federation Server.

What to do next

If you experience problems with the index, you can delete the index, then rebuild it. For more information, see Maintaining the Process Federation Server index.
Parent topic: Configuring and maintaining federated environments