Elasticsearch and Indexing#
There is a multi-step process for copying data from an upstream source and loading it into the Openverse API using ingestion server:
The data is copied from the upstream catalog database and into the downstream API database into new tables.
Data from the downstream API database gets indexed in Elasticsearch in new indexes.
Aliases in the API database and Elasticsearch are updated to point to the new tables and indexes. We refer to this as “promotion”.
The previous API tables and Elasticsearch indexes are deleted.
Performance is dependent on the size of the target Elasticsearch cluster, database throughput, and bandwidth available to the ingestion server. The primary bottleneck is indexing to Elasticsearch.
How indexing works#