Connection Pools
On this page
mongocxx::client vs mongocxx::pool
A standalone mongocxx::client uses a single-threaded algorithm to
monitor the state of the cluster it's connected to. When connected to a
replica set, the thread "stops the world" every 60 seconds to check the
status of the cluster. A mongocxx::pool, on the other hand, uses a
separate background thread for each server in the cluster, each of which
checks the status of the server it monitors every 10 seconds. Because of
the performance advantages of monitoring the cluster in the background
rather than "stopping the world", it's highly recommended to use a
mongocxx::pool rather than a set of standalone clients if your
application has access to multiple threads, even if your application only
uses one thread.
Using Threads with Connection Pools
A mongocxx::pool can be shared across multiple threads and used to create
clients. However, each mongocxx::client can only be used in a single
thread. See the thread safety documentation for
details on how to use a mongocxx::client in a thread-safe manner.
Using Forks with Connection Pools
A mongocxx::pool cannot be shared between a parent and a fork.
You must create your connection pool after forking. See the fork safety
documentation documentation.
Configuring a Connection Pool
The number of clients in a connection pool is limited by the URI parameter
maxPoolSize. After the number of clients created by a mongocxx::pool
(both in the pool and checked out) reaches the value of maxPoolSize,
mongocxx::pool::acquire blocks until another thread returns a client to the
pool. The default value is 100.
Using a Connection Pool
To use a connection pool, first create a mongocxx::pool, passing the URI
as an argument. Then, call mongocxx::pool::acquire to receive a client
from the pool. The client will automatically be returned to the pool when
it goes out of scope.
See the connection pool example for more details.