不论MySQL 还是pg 数据库都通过监听某个ip/端口, 或者某个socket 来实现通讯. 
这里涉及到一个问题,就是这个监听队列的长度问题. 

mysql  是自己实现的,  在my.cnf 里有个配置选项  back_log   这就是设置监听队列的长度的. 


PG 数据库的监听队列的长度, 似乎没有地方可以设置. 

在做一个pgbench 的高并发压力测试的时候,似乎出现这个问题. 

命令:
pgbench -n -r -c 250 -j 250 -T 2 -f update_smallrange.sql


错误消息：
Connection to database "" failed:
could not connect to server: Resource temporarily unavailable
Is the server running locally and accepting
connections on Unix domain socket "/tmp/.s.PGSQL.5432"?


但是从上面的“Resource temporarily unavailable”看不出是哪个资源出问题了。
经过调查，找到了下面一个链接
http://www.postgresql.org/message-id/20130617141622.GH5875@alap2.anarazel.de

[code]
From:Andres Freund To:pgsql-hackers(at)postgresql(dot)orgSubject:PQConnectPoll, connect(2), EWOULDBLOCK and somaxconnDate:2013-06-17 14:16:22Message-ID:20130617141622.GH5875@alap2.anarazel.de (view raw, whole thread or download thread mbox)Thread: 2013-06-17 14:16:22 from Andres Freund   2013-06-26 11:22:58 from Andres Freund    2013-06-26 16:07:54 from Tom Lane     2013-06-26 18:12:00 from Andres Freund      2013-06-27 00:07:40 from Tom Lane       2013-06-27 06:17:57 from Andres Freund        2013-06-27 13:48:25 from Tom Lane        2013-06-27 16:42:47 from Tom Lane      Lists:pgsql-hackersHi,

When postgres on linux receives connection on a high rate client
connections sometimes error out with:
could not send data to server: Transport endpoint is not connected
could not send startup packet: Transport endpoint is not connected

To reproduce start something like on a server with sufficiently high
max_connections:
pgbench -h /tmp -p 5440 -T 10 -c 400 -j 400 -n -f /tmp/simplequery.sql

Now that's strange since that error should happen at connect(2) time,
not when sending the startup packet. Some investigation led me to
fe-secure.c's PQConnectPoll:

if (connect(conn->sock, addr_cur->ai_addr,
                        addr_cur->ai_addrlen) < 0)
{
    if (SOCK_ERRNO == EINPROGRESS ||
        SOCK_ERRNO == EWOULDBLOCK ||
        SOCK_ERRNO == EINTR ||
        SOCK_ERRNO == 0)
    {
        /*
         * This is fine - we're in non-blocking mode, and
         * the connection is in progress.  Tell caller to
         * wait for write-ready on socket.
         */
        conn->status = CONNECTION_STARTED;
        return PGRES_POLLING_WRITING;
    }
    /* otherwise, trouble */
}

So, we're accepting EWOULDBLOCK as a valid return value for
connect(2). Which it isn't. EAGAIN in contrast is on some BSDs and on
linux. Unfortunately POSIX allows those two to share the same value...

My manpage tells me:
EAGAIN No more free local ports or insufficient entries in the routing cache.  For
       AF_INET see the description of
       /proc/sys/net/ipv4/ip_local_port_range ip(7)
       for information on how to increase the number of local
       ports.

So, the problem is that we took a failed connection as having been
initially successfull but in progress.

Not accepting EWOULDBLOCK in the above if() results in:
could not connect to server: Resource temporarily unavailable
      Is the server running locally and accepting
      connections on Unix domain socket "/tmp/.s.PGSQL.5440"?

which makes more sense.

Trivial patch attached.

Now, the question is why we cannot complete connections on unix sockets?
Some code reading reading shows net/unix/af_unix.c:unix_stream_connect()
shows:
        if (unix_recvq_full(other)) {
                err = -EAGAIN;
                if (!timeo)
                        goto out_unlock;
So, if we're in nonblocking mode - which we are - and the receive queue
is full we return EAGAIN. The receive queue for unix sockets is defined
as
static inline int unix_recvq_full(struct sock const *sk)
{
        return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog;
}
Where sk_max_ack_backlog is whatever has been passed to the
listen(backlog) on the listening side.

Question: But postgres does listen(fd, MaxBackends * 2), how can that be
a problem?
Answer:
       If the backlog argument is greater than the value in /proc/sys/net/core/somaxconn,
       then  it  is  silently  truncated to that value; the default value in this file is
       128.  In kernels before 2.4.25, this limit was a hard coded value, SOMAXCONN, with
       the value 128.

Setting somaxconn to something higher indeed makes the problem go away.

I'd guess that pretty much the same holds true for tcp connections,
although I didn't verify that which would explain some previous reports
on the lists.

TLDR: Increase /proc/sys/net/core/somaxconn

Greetings,

Andres Freund

-- 
Andres Freund                           http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services


[/code]


原来是PG服务端的listen backlog(受内核参数somaxconn限制)不够用了，somaxconn的默认值是128，调大后，重启PG再测就OK了。




       /proc/sys/net/core/somaxconn
              This  file  defines a ceiling value for the backlog argument of listen(2); see the listen(2) manual page
              for details.



到这里解决方案就很明了了,  

当前名称：PG数据库库监听队列的长度问题
转载来源：http://hbruida.cn/article/jegjss.html