Recreate the Search index after migrating from a Connections 4.0 deployment

Overview

When we migrate Connections from a previous release, we may need to rebuild the search index. This task is mandatory when migrating from IBM Connections 4.0 or earlier to Connections 5.0, since the index structure was changed between the Connections 4.0 and 4.5 releases.

If we are upgrading from Connections 4.5, do not use this method of recreating the Search Index. Instead, refer to the method described in Migrate the search index from the Connections 4.5 deployment to Connections 5.0.

While performing the steps described in this procedure, search functionality is not available to the users.

Where there is no suitable index available for use in the migrated system (such as when migrating from the Connections 4.0 release, which uses an older index structure), the search index needs to be rebuilt in full, using a process known as initial indexing. A full new search index is build on a single node to cover all content on the system. When this operation is complete, the new index is rolled out to the other nodes in the Search cluster (each Search node maintains its own local index). This happens automatically if all the nodes are online at the time of indexing completion, but if some of those nodes were not online at that time, the index can be copied manually.

Recreate the Search index

1. Stop all the nodes running the Search application.

2. If there are existing search indexes on these nodes, delete them

3. Start all the Search nodes in the cluster.

4. Support for indexing of new content types has been added between the Connections 4.0 and 5.0 releases. To ensure that all content is included in the indexing process, restore the full set of default indexing tasks using...
   SearchService.resetAllTasks()

   If we have configured custom scheduled tasks and do not want to restore the default scheduled tasks, and if the indexing task names specific application names (instead of 'all_configured'), then add the people_finder application name (added in Connections 5.0) to this list. We can update the task to include 'people_finder', by deleting it and adding it back.

   See:

   • List existing indexing tasks using...
     SearchService.listIndexingTasks()

   • Remove indexing tasks using...
     SearchService.deleteTask(String taskName)

   • Add an indexing task using...
     SearchService.addIndexingTask(String taskName, String schedule, String startBy, String applicationNames, Boolean optimizeFlag)

5. Recreate the index by completing one of the following steps:

   • Create a one-off task that indexes all the installed Connections applications

   • Wait for the next scheduled indexing task to run.

   We can tell the index is built on the indexing node when the INDEX.READY and CRAWLING_VERSION files are present in the index directory. The Search index directory is defined by the WAS variable SEARCH_INDEX_DIR.

   After the index is built, the next phase is index roll-out. During this phase, the files in the index directory are automatically copied to the Search staging folder, which is defined by the WAS variable...

   SEARCH_INDEX_SHARED_COPY_LOCATION

   The files in the Search staging folder are then copied to each index folder on the remaining nodes.

   Do not stop the deployment until the index has been copied to all nodes. If the server is stopped during this process, the index will not be successfully rolled out to all nodes. In this event, manually copy the index from the staging location to the other nodes.

Related:
Run one-off tasks
Restore the default scheduled tasks for Search
Delete the index