Percona Search deployment

Percona Search for MongoDB runs as a separate mongot process alongside Percona Server for MongoDB. The deployment topology determines how many mongot instances are required and how search requests are routed.

Applications continue to connect to mongod in a replica set or to mongos in a sharded cluster. They never connect directly to mongot.

Replica set deployment

One mongot deployment serves the complete data-bearing replica set. All mongod members use the same mongot endpoint for search operations.

When the application submits a search query:

1. The application sends the query to a mongod member.
2. mongod forwards the search stage to mongot over gRPC.
3. mongot searches its indexes and returns matching document identifiers and scores.
4. mongod retrieves the matching documents and returns them to the application. A three-member replica set therefore has three mongod processes and one mongot deployment.

Sharded cluster deployment

In a sharded cluster, deploy one mongot for each data-bearing shard. The config server replica set does not require mongot.

The application sends search queries to mongos, which coordinates the request:

1. mongos forwards the query to each relevant shard.
2. A mongod member on each shard forwards the search stage to that shard’s mongot.
3. Each mongot searches its local indexes and returns document identifiers and scores.
4. The shards retrieve the matching documents and return their partial results to mongos.
5. mongos merges and ranks the results before returning them to the application.

For example, a cluster with two three-member shards has six data-bearing mongod processes and two mongot deployments.