Applies to Pega Platform versions 7.2.2 through 8.3.1
When one or more nodes in a clustered environment is restarted, Searches fail across all nodes and the Search landing page reports that there is no indexing host node.
Search fails across all nodes after one or more nodes is restarted.
Steps to Reproduce
- Using the appropriate JVM argument for specifying index directory, set up a cluster with at least three (3) nodes, where at least one node is a search index host node and the other nodes are not.
- Leave the search index host node (or nodes) up and simultaneously restart two or more other nodes.
The problem is caused by inconsistent results from search initialization cluster logic.
Search initialization has logic to remove offline nodes during startup, which the Hazelcast API was being used for. In this case, the Hazelcast API is not reliable because both search initialization and Hazelcast cluster membership occur concurrently as startup tasks.
Therefore, search initialization cluster logic needs to rely on the Elasticsearch API (not Hazelcast cluster membership) to determine offline nodes.
Update to the latest Pega Platform release. See Keeping current with Pega.
The issue was resolved, and the correction was delivered in Pega Platform versions 8.1.6 and 8.2.2.
On the Resolved Issues page, select Pega Platform version 8.2.2 and search for Issue 440079, Search initialization logic updated to ensure consistent node member information.
If you cannot update your Pega deployment and need to continue using Pega Platform version 7.2.2 or a later Pega Platform 7 release, apply the appropriate hotfix.
- Obtain and install the appropriate hotfix for the version of the Pega Platform that you are using. Pega 7.2.2 HFix-53861 Pega 7.3 HFix-53862 Pega 7.3.1 HFix-53540 Pega 7.4 HFix-53422
- After installing the hotfix, restart the server.