If your Search index is corrupt and cannot be used, you
can recreate it by first deleting any existing indexes, and then either
waiting for the next scheduled indexing task to run or running a one-off
indexing task.
Before you begin
Important: When you follow the steps described
in the following procedure, Search functionality is not available
to your users. If you want to recreate the index without the need
for Search downtime, follow the steps described in the Creating
a background index topic instead.
About this task
During
the indexing update process, documents are first written to a cache
table in the HOMEPAGE database and then written to each index across
the nodes. When a new index needs to be built, the database cache
is skipped, and the crawling and indexing process writes directly
to the index directory on the node that is performing the indexing
task.
Procedure
- Stop all the nodes that are running the Search application. If
there are existing search indexes on these nodes, delete them by performing
the steps described in Deleting the index.
- Start all the Search nodes in the cluster.
- Recreate the index by completing one of the following steps:
- Create a one-off task that indexes all the installed IBM® Connections applications in
your deployment. For more information, see Running one-off tasks.
- Wait for the next scheduled indexing task to run.
You can tell that 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 IBM WebSphere® Application Server 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 WebSphere Application Server variable SEARCH_INDEX_SHARED_COPY_LOCATION.
The files in the Search staging folder are then copied to each index
folder on the remaining nodes.
Important: Do not stop
your 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, you need to manually copy
the index from the staging location to the other nodes.