Connection Pooling

A connection pool is a standard technique used to maintain long running connections in memory for efficient re-use, as well as to provide management for the total number of connections an application might use simultaneously.

Particularly for server-side web applications, a connection pool is the standard way to maintain a “pool” of active database connections in memory which are reused across requests.

SQLAlchemy includes several connection pool implementations which integrate with the Engine. They can also be used directly for applications that want to add pooling to an otherwise plain DBAPI approach.

Connection Pool Configuration

The Engine returned by the create_engine() function in most cases has a QueuePool integrated, pre-configured with reasonable pooling defaults. If you’re reading this section only to learn how to enable pooling - congratulations! You’re already done.

The most common QueuePool tuning parameters can be passed directly to create_engine() as keyword arguments: pool_size, max_overflow, pool_recycle and pool_timeout. For example:

engine = create_engine(
    "postgresql+psycopg2://me@localhost/mydb", pool_size=20, max_overflow=0
)

All SQLAlchemy pool implementations have in common that none of them “pre create” connections - all implementations wait until first use before creating a connection. At that point, if no additional concurrent checkout requests for more connections are made, no additional connections are created. This is why it’s perfectly fine for create_engine() to default to using a QueuePool of size five without regard to whether or not the application really needs five connections queued up - the pool would only grow to that size if the application actually used five connections concurrently, in which case the usage of a small pool is an entirely appropriate default behavior.

Note

The QueuePool class is not compatible with asyncio. When using create_async_engine to create an instance of AsyncEngine, the AsyncAdaptedQueuePool class, which makes use of an asyncio-compatible queue implementation, is used instead.

Switching Pool Implementations

The usual way to use a different kind of pool with create_engine() is to use the poolclass argument. This argument accepts a class imported from the sqlalchemy.pool module, and handles the details of building the pool for you. A common use case here is when connection pooling is to be disabled, which can be achieved by using the NullPool implementation:

from sqlalchemy.pool import NullPool

engine = create_engine(
    "postgresql+psycopg2://scott:tiger@localhost/test", poolclass=NullPool
)

Using a Custom Connection Function

See the section Custom DBAPI connect() arguments / on-connect routines for a rundown of the various connection customization routines.

Constructing a Pool

To use a Pool by itself, the creator function is the only argument that’s required and is passed first, followed by any additional options:

import sqlalchemy.pool as pool
import psycopg2


def getconn():
    c = psycopg2.connect(user="ed", host="127.0.0.1", dbname="test")
    return c


mypool = pool.QueuePool(getconn, max_overflow=10, pool_size=5)

DBAPI connections can then be procured from the pool using the Pool.connect() function. The return value of this method is a DBAPI connection that’s contained within a transparent proxy:

# get a connection
conn = mypool.connect()

# use it
cursor_obj = conn.cursor()
cursor_obj.execute("select foo")

The purpose of the transparent proxy is to intercept the close() call, such that instead of the DBAPI connection being closed, it is returned to the pool:

# "close" the connection.  Returns
# it to the pool.
conn.close()

The proxy also returns its contained DBAPI connection to the pool when it is garbage collected, though it’s not deterministic in Python that this occurs immediately (though it is typical with cPython). This usage is not recommended however and in particular is not supported with asyncio DBAPI drivers.

Reset On Return

The pool includes “reset on return” behavior which will call the rollback() method of the DBAPI connection when the connection is returned to the pool. This is so that any existing transactional state is removed from the connection, which includes not just uncommitted data but table and row locks as well. For most DBAPIs, the call to rollback() is inexpensive, and if the DBAPI has already completed a transaction, the method should be a no-op.

Disabling Reset on Return for non-transactional connections

For very specific cases where this rollback() is not useful, such as when using a connection that is configured for autocommit or when using a database that has no ACID capabilities such as the MyISAM engine of MySQL, the reset-on-return behavior can be disabled, which is typically done for performance reasons. This can be affected by using the Pool.reset_on_return parameter of Pool, which is also available from create_engine() as create_engine.pool_reset_on_return, passing a value of None. This is illustrated in the example below, in conjunction with the create_engine.isolation_level parameter setting of AUTOCOMMIT:

non_acid_engine = create_engine(
    "mysql://scott:tiger@host/db",
    pool_reset_on_return=None,
    isolation_level="AUTOCOMMIT",
)

The above engine won’t actually perform ROLLBACK when connections are returned to the pool; since AUTOCOMMIT is enabled, the driver will also not perform any BEGIN operation.

Custom Reset-on-Return Schemes

“reset on return” consisting of a single rollback() may not be sufficient for some use cases; in particular, applications which make use of temporary tables may wish for these tables to be automatically removed on connection checkin. Some (but notably not all) backends include features that can “reset” such tables within the scope of a database connection, which may be a desirable behavior for connection pool reset. Other server resources such as prepared statement handles and server-side statement caches may persist beyond the checkin process, which may or may not be desirable, depending on specifics. Again, some (but again not all) backends may provide for a means of resetting this state. The two SQLAlchemy included dialects which are known to have such reset schemes include Microsoft SQL Server, where an undocumented but widely known stored procedure called sp_reset_connection is often used, and PostgreSQL, which has a well-documented series of commands including DISCARD RESET, DEALLOCATE, and UNLISTEN.

The following example illustrates how to replace reset on return with the Microsoft SQL Server sp_reset_connection stored procedure, using the PoolEvents.reset() event hook. The create_engine.pool_reset_on_return parameter is set to None so that the custom scheme can replace the default behavior completely. The custom hook implementation calls .rollback() in any case, as it’s usually important that the DBAPI’s own tracking of commit/rollback will remain consistent with the state of the transaction:

from sqlalchemy import create_engine
from sqlalchemy import event

mssql_engine = create_engine(
    "mssql+pyodbc://scott:tiger^5HHH@mssql2017:1433/test?driver=ODBC+Driver+17+for+SQL+Server",
    # disable default reset-on-return scheme
    pool_reset_on_return=None,
)


@event.listens_for(mssql_engine, "reset")
def _reset_mssql(dbapi_connection, connection_record, reset_state):
    if not reset_state.terminate_only:
        dbapi_connection.execute("{call sys.sp_reset_connection}")

    # so that the DBAPI itself knows that the connection has been
    # reset
    dbapi_connection.rollback()

Changed in version 2.0.0b3: Added additional state arguments to the PoolEvents.reset() event and additionally ensured the event is invoked for all “reset” occurrences, so that it’s appropriate as a place for custom “reset” handlers. Previous schemes which use the PoolEvents.checkin() handler remain usable as well.

Logging reset-on-return events

Logging for pool events including reset on return can be set logging.DEBUG log level along with the sqlalchemy.pool logger, or by setting create_engine.echo_pool to "debug" when using create_engine():

>>> from sqlalchemy import create_engine
>>> engine = create_engine("postgresql://scott:tiger@localhost/test", echo_pool="debug")

The above pool will show verbose logging including reset on return:

>>> c1 = engine.connect()
DEBUG sqlalchemy.pool.impl.QueuePool Created new connection <connection object ...>
DEBUG sqlalchemy.pool.impl.QueuePool Connection <connection object ...> checked out from pool
>>> c1.close()
DEBUG sqlalchemy.pool.impl.QueuePool Connection <connection object ...> being returned to pool
DEBUG sqlalchemy.pool.impl.QueuePool Connection <connection object ...> rollback-on-return

Pool Events

Connection pools support an event interface that allows hooks to execute upon first connect, upon each new connection, and upon checkout and checkin of connections. See PoolEvents for details.

Dealing with Disconnects

The connection pool has the ability to refresh individual connections as well as its entire set of connections, setting the previously pooled connections as “invalid”. A common use case is allow the connection pool to gracefully recover when the database server has been restarted, and all previously established connections are no longer functional. There are two approaches to this.

Disconnect Handling - Pessimistic

The pessimistic approach refers to emitting a test statement on the SQL connection at the start of each connection pool checkout, to test that the database connection is still viable. The implementation is dialect-specific, and makes use of either a DBAPI-specific ping method, or by using a simple SQL statement like “SELECT 1”, in order to test the connection for liveness.

The approach adds a small bit of overhead to the connection checkout process, however is otherwise the most simple and reliable approach to completely eliminating database errors due to stale pooled connections. The calling application does not need to be concerned about organizing operations to be able to recover from stale connections checked out from the pool.

Pessimistic testing of connections upon checkout is achievable by using the Pool.pre_ping argument, available from create_engine() via the create_engine.pool_pre_ping argument:

engine = create_engine("mysql+pymysql://user:pw@host/db", pool_pre_ping=True)

The “pre ping” feature operates on a per-dialect basis either by invoking a DBAPI-specific “ping” method, or if not available will emit SQL equivalent to “SELECT 1”, catching any errors and detecting the error as a “disconnect” situation. If the ping / error check determines that the connection is not usable, the connection will be immediately recycled, and all other pooled connections older than the current time are invalidated, so that the next time they are checked out, they will also be recycled before use.

If the database is still not available when “pre ping” runs, then the initial connect will fail and the error for failure to connect will be propagated normally. In the uncommon situation that the database is available for connections, but is not able to respond to a “ping”, the “pre_ping” will try up to three times before giving up, propagating the database error last received.

It is critical to note that the pre-ping approach does not accommodate for connections dropped in the middle of transactions or other SQL operations. If the database becomes unavailable while a transaction is in progress, the transaction will be lost and the database error will be raised. While the Connection object will detect a “disconnect” situation and recycle the connection as well as invalidate the rest of the connection pool when this condition occurs, the individual operation where the exception was raised will be lost, and it’s up to the application to either abandon the operation, or retry the whole transaction again. If the engine is configured using DBAPI-level autocommit connections, as described at Setting Transaction Isolation Levels including DBAPI Autocommit, a connection may be reconnected transparently mid-operation using events. See the section How Do I “Retry” a Statement Execution Automatically? for an example.

For dialects that make use of “SELECT 1” and catch errors in order to detect disconnects, the disconnection test may be augmented for new backend-specific error messages using the DialectEvents.handle_error() hook.

Custom / Legacy Pessimistic Ping

Before create_engine.pool_pre_ping was added, the “pre-ping” approach historically has been performed manually using the ConnectionEvents.engine_connect() engine event. The most common recipe for this is below, for reference purposes in case an application is already using such a recipe, or special behaviors are needed:

from sqlalchemy import exc
from sqlalchemy import event
from sqlalchemy import select

some_engine = create_engine(...)


@event.listens_for(some_engine, "engine_connect")
def ping_connection(connection, branch):
    if branch:
        # this parameter is always False as of SQLAlchemy 2.0,
        # but is still accepted by the event hook.  In 1.x versions
        # of SQLAlchemy, "branched" connections should be skipped.
        return

    try:
        # run a SELECT 1.   use a core select() so that
        # the SELECT of a scalar value without a table is
        # appropriately formatted for the backend
        connection.scalar(select(1))
    except exc.DBAPIError as err:
        # catch SQLAlchemy's DBAPIError, which is a wrapper
        # for the DBAPI's exception.  It includes a .connection_invalidated
        # attribute which specifies if this connection is a "disconnect"
        # condition, which is based on inspection of the original exception
        # by the dialect in use.
        if err.connection_invalidated:
            # run the same SELECT again - the connection will re-validate
            # itself and establish a new connection.  The disconnect detection
            # here also causes the whole connection pool to be invalidated
            # so that all stale connections are discarded.
            connection.scalar(select(1))
        else:
            raise

The above recipe has the advantage that we are making use of SQLAlchemy’s facilities for detecting those DBAPI exceptions that are known to indicate a “disconnect” situation, as well as the Engine object’s ability to correctly invalidate the current connection pool when this condition occurs and allowing the current Connection to re-validate onto a new DBAPI connection.

Disconnect Handling - Optimistic

When pessimistic handling is not employed, as well as when the database is shutdown and/or restarted in the middle of a connection’s period of use within a transaction, the other approach to dealing with stale / closed connections is to let SQLAlchemy handle disconnects as they occur, at which point all connections in the pool are invalidated, meaning they are assumed to be stale and will be refreshed upon next checkout. This behavior assumes the Pool is used in conjunction with a Engine. The Engine has logic which can detect disconnection events and refresh the pool automatically.

When the Connection attempts to use a DBAPI connection, and an exception is raised that corresponds to a “disconnect” event, the connection is invalidated. The Connection then calls the Pool.recreate() method, effectively invalidating all connections not currently checked out so that they are replaced with new ones upon next checkout. This flow is illustrated by the code example below:

from sqlalchemy import create_engine, exc

e = create_engine(...)
c = e.connect()

try:
    # suppose the database has been restarted.
    c.execute(text("SELECT * FROM table"))
    c.close()
except exc.DBAPIError as e:
    # an exception is raised, Connection is invalidated.
    if e.connection_invalidated:
        print("Connection was invalidated!")

# after the invalidate event, a new connection
# starts with a new Pool
c = e.connect()
c.execute(text("SELECT * FROM table"))

The above example illustrates that no special intervention is needed to refresh the pool, which continues normally after a disconnection event is detected. However, one database exception is raised, per each connection that is in use while the database unavailability event occurred. In a typical web application using an ORM Session, the above condition would correspond to a single request failing with a 500 error, then the web application continuing normally beyond that. Hence the approach is “optimistic” in that frequent database restarts are not anticipated.

Setting Pool Recycle

An additional setting that can augment the “optimistic” approach is to set the pool recycle parameter. This parameter prevents the pool from using a particular connection that has passed a certain age, and is appropriate for database backends such as MySQL that automatically close connections that have been stale after a particular period of time:

from sqlalchemy import create_engine

e = create_engine("mysql+mysqldb://scott:tiger@localhost/test", pool_recycle=3600)

Above, any DBAPI connection that has been open for more than one hour will be invalidated and replaced, upon next checkout. Note that the invalidation only occurs during checkout - not on any connections that are held in a checked out state. pool_recycle is a function of the Pool itself, independent of whether or not an Engine is in use.

More on Invalidation

The Pool provides “connection invalidation” services which allow both explicit invalidation of a connection as well as automatic invalidation in response to conditions that are determined to render a connection unusable.

“Invalidation” means that a particular DBAPI connection is removed from the pool and discarded. The .close() method is called on this connection if it is not clear that the connection itself might not be closed, however if this method fails, the exception is logged but the operation still proceeds.

When using a Engine, the Connection.invalidate() method is the usual entrypoint to explicit invalidation. Other conditions by which a DBAPI connection might be invalidated include:

  • a DBAPI exception such as OperationalError, raised when a method like connection.execute() is called, is detected as indicating a so-called “disconnect” condition. As the Python DBAPI provides no standard system for determining the nature of an exception, all SQLAlchemy dialects include a system called is_disconnect() which will examine the contents of an exception object, including the string message and any potential error codes included with it, in order to determine if this exception indicates that the connection is no longer usable. If this is the case, the _ConnectionFairy.invalidate() method is called and the DBAPI connection is then discarded.

  • When the connection is returned to the pool, and calling the connection.rollback() or connection.commit() methods, as dictated by the pool’s “reset on return” behavior, throws an exception. A final attempt at calling .close() on the connection will be made, and it is then discarded.

  • When a listener implementing PoolEvents.checkout() raises the DisconnectionError exception, indicating that the connection won’t be usable and a new connection attempt needs to be made.

All invalidations which occur will invoke the PoolEvents.invalidate() event.

Supporting new database error codes for disconnect scenarios

SQLAlchemy dialects each include a routine called is_disconnect() that is invoked whenever a DBAPI exception is encountered. The DBAPI exception object is passed to this method, where dialect-specific heuristics will then determine if the error code received indicates that the database connection has been “disconnected”, or is in an otherwise unusable state which indicates it should be recycled. The heuristics applied here may be customized using the DialectEvents.handle_error() event hook, which is typically established via the owning Engine object. Using this hook, all errors which occur are delivered passing along a contextual object known as ExceptionContext. Custom event hooks may control whether or not a particular error should be considered a “disconnect” situation or not, as well as if this disconnect should cause the entire connection pool to be invalidated or not.

For example, to add support to consider the Oracle Database driver error codes DPY-1001 and DPY-4011 to be handled as disconnect codes, apply an event handler to the engine after creation:

import re

from sqlalchemy import create_engine

engine = create_engine(
    "oracle+oracledb://scott:tiger@localhost:1521?service_name=freepdb1"
)


@event.listens_for(engine, "handle_error")
def handle_exception(context: ExceptionContext) -> None:
    if not context.is_disconnect and re.match(
        r"^(?:DPY-1001|DPY-4011)", str(context.original_exception)
    ):
        context.is_disconnect = True

    return None

The above error processing function will be invoked for all Oracle Database errors raised, including those caught when using the pool pre ping feature for those backends that rely upon disconnect error handling (new in 2.0).

Using FIFO vs. LIFO

The QueuePool class features a flag called QueuePool.use_lifo, which can also be accessed from create_engine() via the flag create_engine.pool_use_lifo. Setting this flag to True causes the pool’s “queue” behavior to instead be that of a “stack”, e.g. the last connection to be returned to the pool is the first one to be used on the next request. In contrast to the pool’s long- standing behavior of first-in-first-out, which produces a round-robin effect of using each connection in the pool in series, lifo mode allows excess connections to remain idle in the pool, allowing server-side timeout schemes to close these connections out. The difference between FIFO and LIFO is basically whether or not its desirable for the pool to keep a full set of connections ready to go even during idle periods:

engine = create_engine("postgreql://", pool_use_lifo=True, pool_pre_ping=True)

Above, we also make use of the create_engine.pool_pre_ping flag so that connections which are closed from the server side are gracefully handled by the connection pool and replaced with a new connection.

Note that the flag only applies to QueuePool use.

New in version 1.3.

Using Connection Pools with Multiprocessing or os.fork()

It’s critical that when using a connection pool, and by extension when using an Engine created via create_engine(), that the pooled connections are not shared to a forked process. TCP connections are represented as file descriptors, which usually work across process boundaries, meaning this will cause concurrent access to the file descriptor on behalf of two or more entirely independent Python interpreter states.

Depending on specifics of the driver and OS, the issues that arise here range from non-working connections to socket connections that are used by multiple processes concurrently, leading to broken messaging (the latter case is typically the most common).

The SQLAlchemy Engine object refers to a connection pool of existing database connections. So when this object is replicated to a child process, the goal is to ensure that no database connections are carried over. There are four general approaches to this:

  1. Disable pooling using NullPool. This is the most simplistic, one shot system that prevents the Engine from using any connection more than once:

    from sqlalchemy.pool import NullPool
    
    engine = create_engine("mysql+mysqldb://user:pass@host/dbname", poolclass=NullPool)
  2. Call Engine.dispose() on any given Engine, passing the Engine.dispose.close parameter with a value of False, within the initialize phase of the child process. This is so that the new process will not touch any of the parent process’ connections and will instead start with new connections. This is the recommended approach:

    from multiprocessing import Pool
    
    engine = create_engine("mysql+mysqldb://user:pass@host/dbname")
    
    
    def run_in_process(some_data_record):
        with engine.connect() as conn:
            conn.execute(text("..."))
    
    
    def initializer():
        """ensure the parent proc's database connections are not touched
        in the new connection pool"""
        engine.dispose(close=False)
    
    
    with Pool(10, initializer=initializer) as p:
        p.map(run_in_process, data)

    New in version 1.4.33: Added the Engine.dispose.close parameter to allow the replacement of a connection pool in a child process without interfering with the connections used by the parent process.

  3. Call Engine.dispose() directly before the child process is created. This will also cause the child process to start with a new connection pool, while ensuring the parent connections are not transferred to the child process:

    engine = create_engine("mysql://user:pass@host/dbname")
    
    
    def run_in_process():
        with engine.connect() as conn:
            conn.execute(text("..."))
    
    
    # before process starts, ensure engine.dispose() is called
    engine.dispose()
    p = Process(target=run_in_process)
    p.start()
  4. An event handler can be applied to the connection pool that tests for connections being shared across process boundaries, and invalidates them:

    from sqlalchemy import event
    from sqlalchemy import exc
    import os
    
    engine = create_engine("...")
    
    
    @event.listens_for(engine, "connect")
    def connect(dbapi_connection, connection_record):
        connection_record.info["pid"] = os.getpid()
    
    
    @event.listens_for(engine, "checkout")
    def checkout(dbapi_connection, connection_record, connection_proxy):
        pid = os.getpid()
        if connection_record.info["pid"] != pid:
            connection_record.dbapi_connection = connection_proxy.dbapi_connection = None
            raise exc.DisconnectionError(
                "Connection record belongs to pid %s, "
                "attempting to check out in pid %s" % (connection_record.info["pid"], pid)
            )

    Above, we use an approach similar to that described in Disconnect Handling - Pessimistic to treat a DBAPI connection that originated in a different parent process as an “invalid” connection, coercing the pool to recycle the connection record to make a new connection.

The above strategies will accommodate the case of an Engine being shared among processes. The above steps alone are not sufficient for the case of sharing a specific Connection over a process boundary; prefer to keep the scope of a particular Connection local to a single process (and thread). It’s additionally not supported to share any kind of ongoing transactional state directly across a process boundary, such as an ORM Session object that’s begun a transaction and references active Connection instances; again prefer to create new Session objects in new processes.

Using a pool instance directly

A pool implementation can be used directly without an engine. This could be used in applications that just wish to use the pool behavior without all other SQLAlchemy features. In the example below the default pool for the MySQLdb dialect is obtained using create_pool_from_url():

from sqlalchemy import create_pool_from_url

my_pool = create_pool_from_url(
    "mysql+mysqldb://", max_overflow=5, pool_size=5, pre_ping=True
)

con = my_pool.connect()
# use the connection
...
# then close it
con.close()

If the type of pool to create is not specified, the default one for the dialect will be used. To specify it directly the poolclass argument can be used, like in the following example:

from sqlalchemy import create_pool_from_url
from sqlalchemy import NullPool

my_pool = create_pool_from_url("mysql+mysqldb://", poolclass=NullPool)

API Documentation - Available Pool Implementations

Object Name Description

_ConnectionFairy

Proxies a DBAPI connection and provides return-on-dereference support.

_ConnectionRecord

Maintains a position in a connection pool which references a pooled connection.

AssertionPool

A Pool that allows at most one checked out connection at any given time.

AsyncAdaptedQueuePool

An asyncio-compatible version of QueuePool.

ConnectionPoolEntry

Interface for the object that maintains an individual database connection on behalf of a Pool instance.

ManagesConnection

Common base for the two connection-management interfaces PoolProxiedConnection and ConnectionPoolEntry.

NullPool

A Pool which does not pool connections.

Pool

Abstract base class for connection pools.

PoolProxiedConnection

A connection-like adapter for a PEP 249 DBAPI connection, which includes additional methods specific to the Pool implementation.

QueuePool

A Pool that imposes a limit on the number of open connections.

SingletonThreadPool

A Pool that maintains one connection per thread.

StaticPool

A Pool of exactly one connection, used for all requests.

class sqlalchemy.pool.Pool

Abstract base class for connection pools.

Class signature

class sqlalchemy.pool.Pool (sqlalchemy.log.Identified, sqlalchemy.event.registry.EventTarget)

method sqlalchemy.pool.Pool.__init__(creator: _CreatorFnType | _CreatorWRecFnType, recycle: int = -1, echo: log._EchoFlagType = None, logging_name: str | None = None, reset_on_return: _ResetStyleArgType = True, events: List[Tuple[_ListenerFnType, str]] | None = None, dialect: _ConnDialect | Dialect | None = None, pre_ping: bool = False, _dispatch: _DispatchCommon[Pool] | None = None)

Construct a Pool.

Parameters:
  • creator – a callable function that returns a DB-API connection object. The function will be called with parameters.

  • recycle – If set to a value other than -1, number of seconds between connection recycling, which means upon checkout, if this timeout is surpassed the connection will be closed and replaced with a newly opened connection. Defaults to -1.

  • logging_name – String identifier which will be used within the “name” field of logging records generated within the “sqlalchemy.pool” logger. Defaults to a hexstring of the object’s id.

  • echo

    if True, the connection pool will log informational output such as when connections are invalidated as well as when connections are recycled to the default log handler, which defaults to sys.stdout for output.. If set to the string "debug", the logging will include pool checkouts and checkins.

    The Pool.echo parameter can also be set from the create_engine() call by using the create_engine.echo_pool parameter.

    See also

    Configuring Logging - further detail on how to configure logging.

  • reset_on_return

    Determine steps to take on connections as they are returned to the pool, which were not otherwise handled by a Connection. Available from create_engine() via the create_engine.pool_reset_on_return parameter.

    Pool.reset_on_return can have any of these values:

    • "rollback" - call rollback() on the connection, to release locks and transaction resources. This is the default value. The vast majority of use cases should leave this value set.

    • "commit" - call commit() on the connection, to release locks and transaction resources. A commit here may be desirable for databases that cache query plans if a commit is emitted, such as Microsoft SQL Server. However, this value is more dangerous than ‘rollback’ because any data changes present on the transaction are committed unconditionally.

    • None - don’t do anything on the connection. This setting may be appropriate if the database / DBAPI works in pure “autocommit” mode at all times, or if a custom reset handler is established using the PoolEvents.reset() event handler.

    • True - same as ‘rollback’, this is here for backwards compatibility.

    • False - same as None, this is here for backwards compatibility.

    For further customization of reset on return, the PoolEvents.reset() event hook may be used which can perform any connection activity desired on reset.

  • events – a list of 2-tuples, each of the form (callable, target) which will be passed to listen() upon construction. Provided here so that event listeners can be assigned via create_engine() before dialect-level listeners are applied.

  • dialect – a Dialect that will handle the job of calling rollback(), close(), or commit() on DBAPI connections. If omitted, a built-in “stub” dialect is used. Applications that make use of create_engine() should not use this parameter as it is handled by the engine creation strategy.

  • pre_ping

    if True, the pool will emit a “ping” (typically “SELECT 1”, but is dialect-specific) on the connection upon checkout, to test if the connection is alive or not. If not, the connection is transparently re-connected and upon success, all other pooled connections established prior to that timestamp are invalidated. Requires that a dialect is passed as well to interpret the disconnection error.

    New in version 1.2.

method sqlalchemy.pool.Pool.connect() PoolProxiedConnection

Return a DBAPI connection from the pool.

The connection is instrumented such that when its close() method is called, the connection will be returned to the pool.

method sqlalchemy.pool.Pool.dispose() None

Dispose of this pool.

This method leaves the possibility of checked-out connections remaining open, as it only affects connections that are idle in the pool.

See also

Pool.recreate()

method sqlalchemy.pool.Pool.recreate() Pool

Return a new Pool, of the same class as this one and configured with identical creation arguments.

This method is used in conjunction with dispose() to close out an entire Pool and create a new one in its place.

class sqlalchemy.pool.QueuePool

A Pool that imposes a limit on the number of open connections.

QueuePool is the default pooling implementation used for all Engine objects other than SQLite with a :memory: database.

The QueuePool class is not compatible with asyncio and create_async_engine(). The AsyncAdaptedQueuePool class is used automatically when using create_async_engine(), if no other kind of pool is specified.

method sqlalchemy.pool.QueuePool.__init__(creator: _CreatorFnType | _CreatorWRecFnType, pool_size: int = 5, max_overflow: int = 10, timeout: float = 30.0, use_lifo: bool = False, **kw: Any)

Construct a QueuePool.

Parameters:
  • creator – a callable function that returns a DB-API connection object, same as that of Pool.creator.

  • pool_size – The size of the pool to be maintained, defaults to 5. This is the largest number of connections that will be kept persistently in the pool. Note that the pool begins with no connections; once this number of connections is requested, that number of connections will remain. pool_size can be set to 0 to indicate no size limit; to disable pooling, use a NullPool instead.

  • max_overflow – The maximum overflow size of the pool. When the number of checked-out connections reaches the size set in pool_size, additional connections will be returned up to this limit. When those additional connections are returned to the pool, they are disconnected and discarded. It follows then that the total number of simultaneous connections the pool will allow is pool_size + max_overflow, and the total number of “sleeping” connections the pool will allow is pool_size. max_overflow can be set to -1 to indicate no overflow limit; no limit will be placed on the total number of concurrent connections. Defaults to 10.

  • timeout – The number of seconds to wait before giving up on returning a connection. Defaults to 30.0. This can be a float but is subject to the limitations of Python time functions which may not be reliable in the tens of milliseconds.

  • use_lifo

    use LIFO (last-in-first-out) when retrieving connections instead of FIFO (first-in-first-out). Using LIFO, a server-side timeout scheme can reduce the number of connections used during non-peak periods of use. When planning for server-side timeouts, ensure that a recycle or pre-ping strategy is in use to gracefully handle stale connections.

    New in version 1.3.

  • **kw – Other keyword arguments including Pool.recycle, Pool.echo, Pool.reset_on_return and others are passed to the Pool constructor.

method sqlalchemy.pool.QueuePool.dispose() None

Dispose of this pool.

This method leaves the possibility of checked-out connections remaining open, as it only affects connections that are idle in the pool.

See also

Pool.recreate()

method sqlalchemy.pool.QueuePool.recreate() QueuePool

Return a new Pool, of the same class as this one and configured with identical creation arguments.

This method is used in conjunction with dispose() to close out an entire Pool and create a new one in its place.

class sqlalchemy.pool.AsyncAdaptedQueuePool

An asyncio-compatible version of QueuePool.

This pool is used by default when using AsyncEngine engines that were generated from create_async_engine(). It uses an asyncio-compatible queue implementation that does not use threading.Lock.

The arguments and operation of AsyncAdaptedQueuePool are otherwise identical to that of QueuePool.

class sqlalchemy.pool.SingletonThreadPool

A Pool that maintains one connection per thread.

Maintains one connection per each thread, never moving a connection to a thread other than the one which it was created in.

Warning

the SingletonThreadPool will call .close() on arbitrary connections that exist beyond the size setting of pool_size, e.g. if more unique thread identities than what pool_size states are used. This cleanup is non-deterministic and not sensitive to whether or not the connections linked to those thread identities are currently in use.

SingletonThreadPool may be improved in a future release, however in its current status it is generally used only for test scenarios using a SQLite :memory: database and is not recommended for production use.

The SingletonThreadPool class is not compatible with asyncio and create_async_engine().

Options are the same as those of Pool, as well as:

Parameters:

pool_size – The number of threads in which to maintain connections at once. Defaults to five.

SingletonThreadPool is used by the SQLite dialect automatically when a memory-based database is used. See SQLite.

method sqlalchemy.pool.SingletonThreadPool.connect() PoolProxiedConnection

Return a DBAPI connection from the pool.

The connection is instrumented such that when its close() method is called, the connection will be returned to the pool.

method sqlalchemy.pool.SingletonThreadPool.dispose() None

Dispose of this pool.

method sqlalchemy.pool.SingletonThreadPool.recreate() SingletonThreadPool

Return a new Pool, of the same class as this one and configured with identical creation arguments.

This method is used in conjunction with dispose() to close out an entire Pool and create a new one in its place.

class sqlalchemy.pool.AssertionPool

A Pool that allows at most one checked out connection at any given time.

This will raise an exception if more than one connection is checked out at a time. Useful for debugging code that is using more connections than desired.

The AssertionPool class is compatible with asyncio and create_async_engine().

method sqlalchemy.pool.AssertionPool.dispose() None

Dispose of this pool.

This method leaves the possibility of checked-out connections remaining open, as it only affects connections that are idle in the pool.

See also

Pool.recreate()

method sqlalchemy.pool.AssertionPool.recreate() AssertionPool

Return a new Pool, of the same class as this one and configured with identical creation arguments.

This method is used in conjunction with dispose() to close out an entire Pool and create a new one in its place.

class sqlalchemy.pool.NullPool

A Pool which does not pool connections.

Instead it literally opens and closes the underlying DB-API connection per each connection open/close.

Reconnect-related functions such as recycle and connection invalidation are not supported by this Pool implementation, since no connections are held persistently.

The NullPool class is compatible with asyncio and create_async_engine().

method sqlalchemy.pool.NullPool.dispose() None

Dispose of this pool.

This method leaves the possibility of checked-out connections remaining open, as it only affects connections that are idle in the pool.

See also

Pool.recreate()

method sqlalchemy.pool.NullPool.recreate() NullPool

Return a new Pool, of the same class as this one and configured with identical creation arguments.

This method is used in conjunction with dispose() to close out an entire Pool and create a new one in its place.

class sqlalchemy.pool.StaticPool

A Pool of exactly one connection, used for all requests.

Reconnect-related functions such as recycle and connection invalidation (which is also used to support auto-reconnect) are only partially supported right now and may not yield good results.

The StaticPool class is compatible with asyncio and create_async_engine().

method sqlalchemy.pool.StaticPool.dispose() None

Dispose of this pool.

This method leaves the possibility of checked-out connections remaining open, as it only affects connections that are idle in the pool.

See also

Pool.recreate()

method sqlalchemy.pool.StaticPool.recreate() StaticPool

Return a new Pool, of the same class as this one and configured with identical creation arguments.

This method is used in conjunction with dispose() to close out an entire Pool and create a new one in its place.

class sqlalchemy.pool.ManagesConnection

Common base for the two connection-management interfaces PoolProxiedConnection and ConnectionPoolEntry.

These two objects are typically exposed in the public facing API via the connection pool event hooks, documented at PoolEvents.

New in version 2.0.

attribute sqlalchemy.pool.ManagesConnection.dbapi_connection: DBAPIConnection | None

A reference to the actual DBAPI connection being tracked.

This is a PEP 249-compliant object that for traditional sync-style dialects is provided by the third-party DBAPI implementation in use. For asyncio dialects, the implementation is typically an adapter object provided by the SQLAlchemy dialect itself; the underlying asyncio object is available via the ManagesConnection.driver_connection attribute.

SQLAlchemy’s interface for the DBAPI connection is based on the DBAPIConnection protocol object

attribute sqlalchemy.pool.ManagesConnection.driver_connection: Any | None

The “driver level” connection object as used by the Python DBAPI or database driver.

For traditional PEP 249 DBAPI implementations, this object will be the same object as that of ManagesConnection.dbapi_connection. For an asyncio database driver, this will be the ultimate “connection” object used by that driver, such as the asyncpg.Connection object which will not have standard pep-249 methods.

New in version 1.4.24.

attribute sqlalchemy.pool.ManagesConnection.info

Info dictionary associated with the underlying DBAPI connection referred to by this ManagesConnection instance, allowing user-defined data to be associated with the connection.

The data in this dictionary is persistent for the lifespan of the DBAPI connection itself, including across pool checkins and checkouts. When the connection is invalidated and replaced with a new one, this dictionary is cleared.

For a PoolProxiedConnection instance that’s not associated with a ConnectionPoolEntry, such as if it were detached, the attribute returns a dictionary that is local to that ConnectionPoolEntry. Therefore the ManagesConnection.info attribute will always provide a Python dictionary.

method sqlalchemy.pool.ManagesConnection.invalidate(e: BaseException | None = None, soft: bool = False) None

Mark the managed connection as invalidated.

Parameters:
  • e – an exception object indicating a reason for the invalidation.

  • soft – if True, the connection isn’t closed; instead, this connection will be recycled on next checkout.

attribute sqlalchemy.pool.ManagesConnection.record_info

Persistent info dictionary associated with this ManagesConnection.

Unlike the ManagesConnection.info dictionary, the lifespan of this dictionary is that of the ConnectionPoolEntry which owns it; therefore this dictionary will persist across reconnects and connection invalidation for a particular entry in the connection pool.

For a PoolProxiedConnection instance that’s not associated with a ConnectionPoolEntry, such as if it were detached, the attribute returns None. Contrast to the ManagesConnection.info dictionary which is never None.

class sqlalchemy.pool.ConnectionPoolEntry

Interface for the object that maintains an individual database connection on behalf of a Pool instance.

The ConnectionPoolEntry object represents the long term maintainance of a particular connection for a pool, including expiring or invalidating that connection to have it replaced with a new one, which will continue to be maintained by that same ConnectionPoolEntry instance. Compared to PoolProxiedConnection, which is the short-term, per-checkout connection manager, this object lasts for the lifespan of a particular “slot” within a connection pool.

The ConnectionPoolEntry object is mostly visible to public-facing API code when it is delivered to connection pool event hooks, such as PoolEvents.connect() and PoolEvents.checkout().

New in version 2.0: ConnectionPoolEntry provides the public facing interface for the _ConnectionRecord internal class.

method sqlalchemy.pool.ConnectionPoolEntry.close() None

Close the DBAPI connection managed by this connection pool entry.

attribute sqlalchemy.pool.ConnectionPoolEntry.dbapi_connection: DBAPIConnection | None

A reference to the actual DBAPI connection being tracked.

This is a PEP 249-compliant object that for traditional sync-style dialects is provided by the third-party DBAPI implementation in use. For asyncio dialects, the implementation is typically an adapter object provided by the SQLAlchemy dialect itself; the underlying asyncio object is available via the ManagesConnection.driver_connection attribute.

SQLAlchemy’s interface for the DBAPI connection is based on the DBAPIConnection protocol object

attribute sqlalchemy.pool.ConnectionPoolEntry.driver_connection: Any | None

The “driver level” connection object as used by the Python DBAPI or database driver.

For traditional PEP 249 DBAPI implementations, this object will be the same object as that of ManagesConnection.dbapi_connection. For an asyncio database driver, this will be the ultimate “connection” object used by that driver, such as the asyncpg.Connection object which will not have standard pep-249 methods.

New in version 1.4.24.

attribute sqlalchemy.pool.ConnectionPoolEntry.in_use

Return True the connection is currently checked out

attribute sqlalchemy.pool.ConnectionPoolEntry.info

inherited from the ManagesConnection.info attribute of ManagesConnection

Info dictionary associated with the underlying DBAPI connection referred to by this ManagesConnection instance, allowing user-defined data to be associated with the connection.

The data in this dictionary is persistent for the lifespan of the DBAPI connection itself, including across pool checkins and checkouts. When the connection is invalidated and replaced with a new one, this dictionary is cleared.

For a PoolProxiedConnection instance that’s not associated with a ConnectionPoolEntry, such as if it were detached, the attribute returns a dictionary that is local to that ConnectionPoolEntry. Therefore the ManagesConnection.info attribute will always provide a Python dictionary.

method sqlalchemy.pool.ConnectionPoolEntry.invalidate(e: BaseException | None = None, soft: bool = False) None

inherited from the ManagesConnection.invalidate() method of ManagesConnection

Mark the managed connection as invalidated.

Parameters:
  • e – an exception object indicating a reason for the invalidation.

  • soft – if True, the connection isn’t closed; instead, this connection will be recycled on next checkout.

attribute sqlalchemy.pool.ConnectionPoolEntry.record_info

inherited from the ManagesConnection.record_info attribute of ManagesConnection

Persistent info dictionary associated with this ManagesConnection.

Unlike the ManagesConnection.info dictionary, the lifespan of this dictionary is that of the ConnectionPoolEntry which owns it; therefore this dictionary will persist across reconnects and connection invalidation for a particular entry in the connection pool.

For a PoolProxiedConnection instance that’s not associated with a ConnectionPoolEntry, such as if it were detached, the attribute returns None. Contrast to the ManagesConnection.info dictionary which is never None.

class sqlalchemy.pool.PoolProxiedConnection

A connection-like adapter for a PEP 249 DBAPI connection, which includes additional methods specific to the Pool implementation.

PoolProxiedConnection is the public-facing interface for the internal _ConnectionFairy implementation object; users familiar with _ConnectionFairy can consider this object to be equivalent.

New in version 2.0: PoolProxiedConnection provides the public- facing interface for the _ConnectionFairy internal class.

method sqlalchemy.pool.PoolProxiedConnection.close() None

Release this connection back to the pool.

The PoolProxiedConnection.close() method shadows the PEP 249 .close() method, altering its behavior to instead release the proxied connection back to the connection pool.

Upon release to the pool, whether the connection stays “opened” and pooled in the Python process, versus actually closed out and removed from the Python process, is based on the pool implementation in use and its configuration and current state.

attribute sqlalchemy.pool.PoolProxiedConnection.dbapi_connection: DBAPIConnection | None

A reference to the actual DBAPI connection being tracked.

This is a PEP 249-compliant object that for traditional sync-style dialects is provided by the third-party DBAPI implementation in use. For asyncio dialects, the implementation is typically an adapter object provided by the SQLAlchemy dialect itself; the underlying asyncio object is available via the ManagesConnection.driver_connection attribute.

SQLAlchemy’s interface for the DBAPI connection is based on the DBAPIConnection protocol object

method sqlalchemy.pool.PoolProxiedConnection.detach() None

Separate this connection from its Pool.

This means that the connection will no longer be returned to the pool when closed, and will instead be literally closed. The associated ConnectionPoolEntry is de-associated from this DBAPI connection.

Note that any overall connection limiting constraints imposed by a Pool implementation may be violated after a detach, as the detached connection is removed from the pool’s knowledge and control.

attribute sqlalchemy.pool.PoolProxiedConnection.driver_connection: Any | None

The “driver level” connection object as used by the Python DBAPI or database driver.

For traditional PEP 249 DBAPI implementations, this object will be the same object as that of ManagesConnection.dbapi_connection. For an asyncio database driver, this will be the ultimate “connection” object used by that driver, such as the asyncpg.Connection object which will not have standard pep-249 methods.

New in version 1.4.24.

attribute sqlalchemy.pool.PoolProxiedConnection.info

inherited from the ManagesConnection.info attribute of ManagesConnection

Info dictionary associated with the underlying DBAPI connection referred to by this ManagesConnection instance, allowing user-defined data to be associated with the connection.

The data in this dictionary is persistent for the lifespan of the DBAPI connection itself, including across pool checkins and checkouts. When the connection is invalidated and replaced with a new one, this dictionary is cleared.

For a PoolProxiedConnection instance that’s not associated with a ConnectionPoolEntry, such as if it were detached, the attribute returns a dictionary that is local to that ConnectionPoolEntry. Therefore the ManagesConnection.info attribute will always provide a Python dictionary.

method sqlalchemy.pool.PoolProxiedConnection.invalidate(e: BaseException | None = None, soft: bool = False) None

inherited from the ManagesConnection.invalidate() method of ManagesConnection

Mark the managed connection as invalidated.

Parameters:
  • e – an exception object indicating a reason for the invalidation.

  • soft – if True, the connection isn’t closed; instead, this connection will be recycled on next checkout.

attribute sqlalchemy.pool.PoolProxiedConnection.is_detached

Return True if this PoolProxiedConnection is detached from its pool.

attribute sqlalchemy.pool.PoolProxiedConnection.is_valid

Return True if this PoolProxiedConnection still refers to an active DBAPI connection.

attribute sqlalchemy.pool.PoolProxiedConnection.record_info

inherited from the ManagesConnection.record_info attribute of ManagesConnection

Persistent info dictionary associated with this ManagesConnection.

Unlike the ManagesConnection.info dictionary, the lifespan of this dictionary is that of the ConnectionPoolEntry which owns it; therefore this dictionary will persist across reconnects and connection invalidation for a particular entry in the connection pool.

For a PoolProxiedConnection instance that’s not associated with a ConnectionPoolEntry, such as if it were detached, the attribute returns None. Contrast to the ManagesConnection.info dictionary which is never None.

class sqlalchemy.pool._ConnectionFairy

Proxies a DBAPI connection and provides return-on-dereference support.

This is an internal object used by the Pool implementation to provide context management to a DBAPI connection delivered by that Pool. The public facing interface for this class is described by the PoolProxiedConnection class. See that class for public API details.

The name “fairy” is inspired by the fact that the _ConnectionFairy object’s lifespan is transitory, as it lasts only for the length of a specific DBAPI connection being checked out from the pool, and additionally that as a transparent proxy, it is mostly invisible.

Class signature

class sqlalchemy.pool._ConnectionFairy (sqlalchemy.pool.base.PoolProxiedConnection)

class sqlalchemy.pool._ConnectionRecord

Maintains a position in a connection pool which references a pooled connection.

This is an internal object used by the Pool implementation to provide context management to a DBAPI connection maintained by that Pool. The public facing interface for this class is described by the ConnectionPoolEntry class. See that class for public API details.

Class signature

class sqlalchemy.pool._ConnectionRecord (sqlalchemy.pool.base.ConnectionPoolEntry)