push.scheduler.killed-percentage-for-reindex

Background

The percentage of documents which must already be killed in a set of merging generations for the new generation to re-index the merged WARC file (rather than merging the indexes).

A lower value will result in more frequent re-indexing, producing improved ranking at the cost of more computation time per merge. A higher value reduces merge times at the cost of lower search quality, increased disk space usage and slower query processing times.

Setting the key

Set this configuration key in the search package or data source configuration.

Use the configuration key editor to add or edit the push.scheduler.killed-percentage-for-reindex key, and set the value. This can be set to any valid Double value.

Default value

By default, a push data source will re-index when 50% of the documents have been killed.

push.scheduler.killed-percentage-for-reindex=50

Examples

To set a push data source to re-index on every merge:

push.scheduler.killed-percentage-for-reindex=0