Python Postgres psycopg2 ThreadedConnectionPool耗尽

Question

我已经在这里调查了几个“太多客户”的相关话题，但仍然无法解决我的问题，所以我不得不再次问这个问题，为我的具体情况。

基本上，我设置了本地Postgres服务器，需要执行数万个查询，所以我使用了Python psycopg2包。以下是我的代码：

import psycopg2
import pandas as pd
import numpy as np
from flashtext import KeywordProcessor
from psycopg2.pool import ThreadedConnectionPool
from concurrent.futures import ThreadPoolExecutor

df = pd.DataFrame({'S':['California', 'Ohio', 'Texas'], 'T':['Dispatcher', 'Zookeeper', 'Mechanics']})
# df = pd.concat([df]*10000) # repeat df 10000 times

DSN = "postgresql://User:password@localhost/db"
tcp = ThreadedConnectionPool(1, 800, DSN)

def do_one_query(inputS, inputT):
    conn = tcp.getconn()
    c = conn.cursor()

    q = r"SELECT * from eridata where "State" = 'California' and "Title" = 'Dispatcher' limit 1;"   

    c.execute(q)
    all_results = c.fetchall()
    for row in all_results:
        return row
    tcp.putconn(conn, close=True)

cnt=0
for idx, row in df.iterrows():

    cnt+=1
    with ThreadPoolExecutor(max_workers=1) as pool:
        ret = pool.submit(do_one_query,  row["S"], row["T"])
        print ret.result()
    print cnt

代码在一个小的df下运行良好。如果我重复df 10000次，就会收到一条错误消息，说连接池已耗尽 . 我想我用过的连接已经被这条线切断了：

tcp.putconn（conn，close=True） 但我想事实上他们没有关门？我怎样才能避开这个问题？

Answer 1

你的问题是，你实际上没有返回到池的连接，而是用

tcp.putconn(conn, close=True)

请参阅此处的文档http://initd.org/psycopg/docs/pool.html

If close is True, discard the connection from the pool.

因此，如果将800个连接放入池中，在801个循环之后，会出现“耗尽错误”，因为连接池大小为零。

Answer 2

我一直在努力寻找有关ThreadedConnectionPool如何工作的真正详细信息。https://bbengfort.github.io/observations/2017/12/06/psycopg2-transactions.html还不错，但事实证明，getconn在连接可用之前阻塞的声明是不正确的。检查代码时，所有ThreadedConnectionPool添加的都是围绕AbstractConnectionPool方法的锁，以防止竞争条件。如果在任何一点尝试使用超过maxconn连接，将引发连接池耗尽池错误。

如果您想要比the accepted answer简单一点的东西，那么进一步将方法包装在一个提供阻塞的信号量中，直到连接可用为止，就可以做到这一点：

from psycopg2.pool import ThreadedConnectionPool
from threading import Semaphore

class ReallyThreadedConnectionPool(ThreadedConnectionPool):
    def __init__(self, minconn, maxconn, *args, **kwargs):
        self._semaphore = Semaphore(maxconn)
        super().__init__(minconn, maxconn, *args, **kwargs)

    def getconn(self, *args, **kwargs):
        self._semaphore.acquire()
        return super().getconn(*args, **kwargs)

    def putconn(self, *args, **kwargs):
        super().putconn(*args, **kwargs)
        self._semaphore.release()

Answer 3

你需要在游泳池上面排队。

类似于以下的方法应该有效：

import gevent, sys, random, psycopg2, logging
from contextlib import contextmanager
from gevent.queue import Queue
from gevent.socket import wait_read, wait_write
from psycopg2.pool import ThreadedConnectionPool
from psycopg2 import extensions, OperationalError
import sys
logger = logging.getLogger(__name__)

poolsize = 100  #number of max connections
pdsn = '' # put your dsn here

if sys.version_info[0] >= 3:
    integer_types = (int,)
else:
    import __builtin__
    integer_types = (int, __builtin__.long)


class ConnectorError(Exception):
    """ This is a base class for all CONNECTOR related exceptions """
    pass


#singleton connection pool, gets reset if a connection is bad or drops
_pgpool = None
def pgpool():
    global _pgpool
    if not _pgpool:
        try:
            _pgpool = PostgresConnectionPool(maxsize=poolsize)
        except psycopg2.OperationalError as exc:
            _pgpool = None
    return _pgpool

class Pcursor(object):

    def __init__(self, **kwargs):
        #in case of a lost connection lets sit and wait till it's online
        global _pgpool
        if not _pgpool:
            while not _pgpool:
                try:
                    pgpool()
                except:
                    logger.debug('Attempting Connection To Postgres...')
                    gevent.sleep(1)

    def fetchone(self, PSQL, *args):
        with _pgpool.cursor() as cursor:
            try:
                cursor.execute(PSQL, args)
            except TypeError:
                cursor.execute(PSQL, args[0])
            except Exception as exc:
                print(sys._getframe().f_back.f_code)
                print(sys._getframe().f_back.f_code.co_name)
                logger.warning(str(exc))
            logger.debug(cursor.query)
            return cursor.fetchone()

    def fetchall(self, PSQL, *args):
        with _pgpool.cursor() as cursor:
            try:
                cursor.execute(PSQL, args)
            except TypeError:
                cursor.execute(PSQL, args[0])
            except Exception as exc:
                print(sys._getframe().f_back.f_code)
                print(sys._getframe().f_back.f_code.co_name)
                logger.warning(str(exc))
            logger.debug(cursor.query)
            return cursor.fetchall()

    def execute(self, PSQL, *args):
        with _pgpool.cursor() as cursor:
            try:
                cursor.execute(PSQL, args)
            except TypeError:
                cursor.execute(PSQL, args[0])
            except Exception as exc:
                print(sys._getframe().f_back.f_code)
                print(sys._getframe().f_back.f_code.co_name)
                logger.warning(str(exc))
            finally:
                logger.debug(cursor.query)
                return cursor.query

    def fetchmany(self, PSQL, *args):
        with _pgpool.cursor() as cursor:
            try:
                cursor.execute(PSQL, args)
            except TypeError:
                cursor.execute(PSQL, args[0])
            while 1:
                items = cursor.fetchmany()
                if not items:
                    break
                for item in items:
                    yield item

class AbstractDatabaseConnectionPool(object):

    def __init__(self, maxsize=poolsize):
        if not isinstance(maxsize, integer_types):
            raise TypeError('Expected integer, got %r' % (maxsize, ))
        self.maxsize = maxsize
        self.pool = Queue()
        self.size = 0

    def create_connection(self):
        #overridden by PostgresConnectionPool
        raise NotImplementedError()

    def get(self):
        pool = self.pool
        if self.size >= self.maxsize or pool.qsize():
            return pool.get()

        self.size += 1
        try:
            new_item = self.create_connection()
        except:
            self.size -= 1
            raise
        return new_item

    def put(self, item):
        self.pool.put(item)

    def closeall(self):
        while not self.pool.empty():
            conn = self.pool.get_nowait()
            try:
                conn.close()
            except Exception:
                pass

    @contextmanager
    def connection(self, isolation_level=None):
        conn = self.get()
        try:
            if isolation_level is not None:
                if conn.isolation_level == isolation_level:
                    isolation_level = None
                else:
                    conn.set_isolation_level(isolation_level)
            yield conn
        except:
            if conn.closed:
                conn = None
                self.closeall()
            raise
        else:
            if conn.closed:
                raise OperationalError("Cannot commit because connection was closed: %r" % (conn, ))
        finally:
            if conn is not None and not conn.closed:
                if isolation_level is not None:
                    conn.set_isolation_level(isolation_level)
                self.put(conn)

    @contextmanager
    def cursor(self, *args, **kwargs):
        isolation_level = kwargs.pop('isolation_level', None)
        with self.connection(isolation_level) as conn:
            try:
                yield conn.cursor(*args, **kwargs)
            except:
                global _pgpool
                _pgpool = None
                del(self)


class PostgresConnectionPool(AbstractDatabaseConnectionPool):
    def __init__(self,**kwargs):
        try:
            self.pconnect = ThreadedConnectionPool(1, poolsize, dsn=pdsn)
        except:
            global _pgpool
            _pgpool = None
            raise ConnectorError('Database Connection Failed')
        maxsize = kwargs.pop('maxsize', None)
        self.kwargs = kwargs
        AbstractDatabaseConnectionPool.__init__(self, maxsize)

    def create_connection(self):
        self.conn = self.pconnect.getconn()
        self.conn.autocommit = True
        return self.conn


def gevent_wait_callback(conn, timeout=None):
    """A wait callback useful to allow gevent to work with Psycopg."""
    while 1:
        state = conn.poll()
        if state == extensions.POLL_OK:
            break
        elif state == extensions.POLL_READ:
            wait_read(conn.fileno(), timeout=timeout)
        elif state == extensions.POLL_WRITE:
            wait_write(conn.fileno(), timeout=timeout)
        else:
            raise ConnectorError("Bad result from poll: %r" % state)

extensions.set_wait_callback(gevent_wait_callback)

然后您可以通过以下方式拨打您的连接：

import db
db.Pcursor().execute(PSQL, arg1, arg2, arg3)

基本上，我借用了异步postgres的gevent示例，并对其进行了修改，以通过pyscopg2支持线程池。

https://github.com/gevent/gevent/blob/master/examples/psycopg2_pool.py

我在模块中添加了psycogreen的功能，所以您只需要导入并调用类。对类的每次调用都会在队列上堆叠一个新查询，但只使用特定大小的池。这样你就不会失去联系。这基本上类似于PGBouncer所做的，我认为这也可以消除你的问题。

https://pgbouncer.github.io/