MySQL参数max_connect_errors分析释疑

最近一MySQL服务器，由于一些特殊因素遇到“ERROR 1129 (00000): Host ‘xxx’ is blocked because of many connection errors. Unblock with ‘mysqladmin flush-hosts’”，在问题解决后，在详细了解参数max_connect_errors的过程中，有些不同网络资料的矛盾描述确实让我有点迷惑和混淆（关于这个错误，本质原因是因为同一个IP在短时间内产生太多中断的数据库连接（超过max_connect_errors的最大值）而导致的），下面介绍我的探索问题、分析问题、释疑的一个过程。

首先，我在网上搜索了一些资料，不少资料信誓旦旦的介绍，密码输入错误的尝试次数超过max_connect_errors变量，MySQL就会阻塞这个客户端登录，然后我找到了官方资料关于max_connect_errors的介绍，如下所示，MySQL 5.6/5.7的介绍一致

If more than this many successive connection requests from a host are interrupted without a successful connection, the server blocks that host from further connections. You can unblock blocked hosts by flushing the host cache. To do so, issue a FLUSH HOSTS statement or execute a mysqladmin flush-hosts command. If a connection is established successfully within fewer than max_connect_errors attempts after a previous connection was interrupted, the error count for the host is cleared to zero. However, once a host is blocked, flushing the host cache is the only way to unblock it. The default is 100.

如上所示，翻译出来的话，大致如下：如果MySQL服务器连续接收到了来自于同一个主机的请求，而且这些连续的请求全部都没有成功的建立连接就被中断了，当这些连续的请求的累计值大于max_connect_errors的设定值时，MySQL服务器就会阻止这台主机后续的所有请求。相信一开始你看到这些资料，也会被“many successive connection requests from a host are interrupted without a successful connection”给弄懵，其实这个就是因为由于网络异常而中止数据库连接。网上搜索到这么一个资料：

There seems to be confusion around that variable. It does not really block hosts for repeated invalid passwords but for aborted connections due to network errors.

好吧，那么我们自己动手实验验证一下，就能弄明白到底那个是正确的。在MySQL数据库里面创建一个test账号，然后我们将max_connect_errors变量设置为3.

mysql> show variables like ‘%max_connect_errors%’;
±-------------------±------+
| Variable_name | Value |
±-------------------±------+
| max_connect_errors | 100 |
±-------------------±------+
1 row in set (0.00 sec)

mysql> set global max_connect_errors=3;
Query OK, 0 rows affected (0.00 sec)

mysql> show variables like ‘%max_connect_error%’;
±-------------------±------+
| Variable_name | Value |
±-------------------±------+
| max_connect_errors | 3 |
±-------------------±------+
1 row in set (0.00 sec)

然后我们在另外一台测试机器，以错误的密码去连接这个MySQL数据库，如下所示，即使前面输入了三次错误密码，第四次输入是也没有碰到上面错误。那么可以排除这个变量与密码错误输入有关系。

[root@mytestlnx02 tmp]# mysql -h10.20.57.24 -utest -p

Enter password:

ERROR 1045 (28000): Access denied for user ‘test’@‘mytestlnx02’ (using password: YES)

[root@mytestlnx02 tmp]# mysql -h10.20.57.24 -utest -p

Enter password:

ERROR 1045 (28000): Access denied for user ‘test’@‘mytestlnx02’ (using password: YES)

[root@mytestlnx02 tmp]# mysql -h10.20.57.24 -utest -p

Enter password:

ERROR 1045 (28000): Access denied for user ‘test’@‘mytestlnx02’ (using password: YES)

[root@mytestlnx02 tmp]# mysql -h10.20.57.24 -utest -p

Enter password:

ERROR 1045 (28000): Access denied for user ‘test’@‘mytestlnx02’ (using password: YES)

[root@mytestlnx02 tmp]#

其实，关于某个IP输入了错误密码，MySQL会在performance_schema数据库下的host_cache表中记录。它会累计记录在COUNT_AUTHENTICATION_ERRORS字段，如下所示：

mysql> use performance_schema;
Database changed
mysql> select * from host_cache\G;
*************************** 1. row ***************************
IP: 192.168.27.180
HOST: gettestlnx02
HOST_VALIDATED: YES
SUM_CONNECT_ERRORS: 0
COUNT_HOST_BLOCKED_ERRORS: 0
COUNT_NAMEINFO_TRANSIENT_ERRORS: 0
COUNT_NAMEINFO_PERMANENT_ERRORS: 0
COUNT_FORMAT_ERRORS: 0
COUNT_ADDRINFO_TRANSIENT_ERRORS: 0
COUNT_ADDRINFO_PERMANENT_ERRORS: 0
COUNT_FCRDNS_ERRORS: 0
COUNT_HOST_ACL_ERRORS: 0
COUNT_NO_AUTH_PLUGIN_ERRORS: 0
COUNT_AUTH_PLUGIN_ERRORS: 0
COUNT_HANDSHAKE_ERRORS: 0
COUNT_PROXY_USER_ERRORS: 0
COUNT_PROXY_USER_ACL_ERRORS: 0
COUNT_AUTHENTICATION_ERRORS: 4
COUNT_SSL_ERRORS: 0
COUNT_MAX_USER_CONNECTIONS_ERRORS: 0
COUNT_MAX_USER_CONNECTIONS_PER_HOUR_ERRORS: 0
COUNT_DEFAULT_DATABASE_ERRORS: 0
COUNT_INIT_CONNECT_ERRORS: 0
COUNT_LOCAL_ERRORS: 0
COUNT_UNKNOWN_ERRORS: 0
FIRST_SEEN: 2018-01-31 16:28:19
LAST_SEEN: 2018-01-31 16:28:26
FIRST_ERROR_SEEN: 2018-01-31 16:28:19
LAST_ERROR_SEEN: 2018-01-31 16:28:26
1 row in set (0.00 sec)

ERROR:
No query specified

clip_image001

官方资料介绍，host_cache的字段是统计被视为“阻塞”的连接错误的数量（根据max_connect_errors系统变量进行评估）。 只计算协议握手错误，并且仅用于通过验证的主机（HOST_VALIDATED = YES）。

SUM_CONNECT_ERRORS

The number of connection errors that are deemed “blocking” (assessed against the max_connect_errors system variable). Only protocol handshake errors are counted, and only for hosts that passed validation (HOST_VALIDATED = YES).

MySQL客户端与数据库建立连接需要发起三次握手协议，正常情况下，这个时间非常短，但是一旦网络异常，网络超时等因素出现，就会导致这个握手协议无法完成，MySQL有个参数connect_timeout，它是MySQL服务端进程mysqld等待连接建立完成的时间，单位为秒。如果超过connect_timeout时间范围内，仍然无法完成协议握手话，MySQL客户端会收到异常，异常消息类似于: Lost connection to MySQL server at ‘XXX’, system error: errno，该变量默认是10秒：

mysql> show variables like ‘connect_timeout’;
±----------------±------+
| Variable_name | Value |
±----------------±------+
| connect_timeout | 10 |
±----------------±------+
1 row in set (0.00 sec)

mysql>

那么我们就构造一个网络超时引起的数据库连接被中断案例吧，我们用Linux下的netem与tc命令模拟构造出复杂环境下的网络传输延时案例，如下设置后，此时从测试服务器去访问MySQL服务器，都会出现延时11秒：

[root@gettestlnx02 ~]# ping 10.20.57.24

PING 10.20.57.24 (10.20.57.24) 56(84) bytes of data.

64 bytes from 10.20.57.24: icmp_seq=1 ttl=62 time=0.251 ms

64 bytes from 10.20.57.24: icmp_seq=2 ttl=62 time=0.330 ms

64 bytes from 10.20.57.24: icmp_seq=3 ttl=62 time=0.362 ms

64 bytes from 10.20.57.24: icmp_seq=4 ttl=62 time=0.316 ms

64 bytes from 10.20.57.24: icmp_seq=5 ttl=62 time=0.281 ms

64 bytes from 10.20.57.24: icmp_seq=6 ttl=62 time=0.377 ms

^C

— 10.20.57.24 ping statistics —

6 packets transmitted, 6 received, 0% packet loss, time 5716ms

rtt min/avg/max/mdev = 0.251/0.319/0.377/0.047 ms

[root@gettestlnx02 ~]# tc qdisc add dev eth0 root netem delay 11000ms

[root@gettestlnx02 ~]# ping 10.20.57.24

PING 10.20.57.24 (10.20.57.24) 56(84) bytes of data.

64 bytes from 10.20.57.24: icmp_seq=1 ttl=62 time=11000 ms

64 bytes from 10.20.57.24: icmp_seq=2 ttl=62 time=11000 ms

64 bytes from 10.20.57.24: icmp_seq=3 ttl=62 time=11000 ms

64 bytes from 10.20.57.24: icmp_seq=4 ttl=62 time=11000 ms

64 bytes from 10.20.57.24: icmp_seq=5 ttl=62 time=11000 ms

64 bytes from 10.20.57.24: icmp_seq=6 ttl=62 time=11000 ms

64 bytes from 10.20.57.24: icmp_seq=7 ttl=62 time=11000 ms

clip_image002

我们在测试服务器gettestlnx02连接MySQL数据库，如下所示（注意，如果你是在通过ssh连接这台服务器的话，此时在gettestlnx02上操作会相当慢。当然你也可以在MySQL服务器模拟网络延时，或者你将connect_timeout和网络延时都设小一点）

[root@gettestlnx02 ~]# mysql -h10.20.57.24 -utest -p

Enter password:

ERROR 2013 (HY000): Lost connection to MySQL server at ‘reading authorization packet’, system error: 0

[root@gettestlnx02 ~]#

如上所示，由于网络延时超过10秒，导致连接MySQL失败，此时，你在MySQL服务器上查询host_cache表时,那么你就会看到SUM_CONNECT_ERRORS变成1了，COUNT_HANDSHAKE_ERRORS也变成了1.

clip_image003

那么我们反复这样折腾三次，那么你会看到SUM_CONNECT_ERRORS变成3了，COUNT_HANDSHAKE_ERRORS也变成了3了.

mysql> select * from host_cache\G;
*************************** 1. row ***************************
IP: 192.168.27.180
HOST: gettestlnx02
HOST_VALIDATED: YES
SUM_CONNECT_ERRORS: 3
COUNT_HOST_BLOCKED_ERRORS: 1
COUNT_NAMEINFO_TRANSIENT_ERRORS: 0
COUNT_NAMEINFO_PERMANENT_ERRORS: 0
COUNT_FORMAT_ERRORS: 0
COUNT_ADDRINFO_TRANSIENT_ERRORS: 0
COUNT_ADDRINFO_PERMANENT_ERRORS: 0
COUNT_FCRDNS_ERRORS: 0
COUNT_HOST_ACL_ERRORS: 0
COUNT_NO_AUTH_PLUGIN_ERRORS: 0
COUNT_AUTH_PLUGIN_ERRORS: 0
COUNT_HANDSHAKE_ERRORS: 3
COUNT_PROXY_USER_ERRORS: 0
COUNT_PROXY_USER_ACL_ERRORS: 0
COUNT_AUTHENTICATION_ERRORS: 4
COUNT_SSL_ERRORS: 0
COUNT_MAX_USER_CONNECTIONS_ERRORS: 0
COUNT_MAX_USER_CONNECTIONS_PER_HOUR_ERRORS: 0
COUNT_DEFAULT_DATABASE_ERRORS: 0
COUNT_INIT_CONNECT_ERRORS: 0
COUNT_LOCAL_ERRORS: 0
COUNT_UNKNOWN_ERRORS: 0
FIRST_SEEN: 2018-01-31 16:28:19
LAST_SEEN: 2018-01-31 17:02:10
FIRST_ERROR_SEEN: 2018-01-31 16:28:19
LAST_ERROR_SEEN: 2018-01-31 17:02:10
1 row in set (0.00 sec)

ERROR:
No query specified

然后我们用netem与tc 命令在测试服务器上取消网络延迟模拟，然后去测试连接MySQL数据库，如下测试所示：

[root@gettestlnx02 ~]# tc qdisc del dev eth0 root netem delay 11000ms

[root@gettestlnx02 ~]# mysql -h10.20.57.24 -utest -p

Enter password:

ERROR 1129 (HY000): Host ‘192.168.27.180’ is blocked because of many connection errors; unblock with ‘mysqladmin flush-hosts’

[root@gettestlnx02 ~]#

clip_image004

此时就能构造出“ERROR 1129 (HY000): Host ‘192.168.27.180’ is blocked because of many connection errors; unblock with ‘mysqladmin flush-hosts’”错误了。

解决方案

解决ERROR 1129 (00000): Host ‘xxx’ is blocked because of many connection errors. Unblock with 'mysqladmin flush-hosts’这个错误的方法比较多，不过有些方案都是临时方案。临时方案是指标不治本。关键还是需要解决网络错误（这个往往需要求助网络管理人员或系统管理员）

解决方法：

1、将变量max_connection_errors的值设置为一个更大的值

mysql> show variables like ‘max_connect_errors’;
±-------------------±------+
| Variable_name | Value |
±-------------------±------+
| max_connect_errors | 3 |
±-------------------±------+
1 row in set (0.01 sec)

mysql> set global max_connect_errors=150;
Query OK, 0 rows affected (0.00 sec)

mysql>

这个临时方案只是延迟触发IP被禁止访问的条件而已，而且在复杂情况或高并发的情况下，需要设置一个很大的值，否则很容易就会再次被触发。另外，变量只对当前环境生效，如果重启就会失效，如果需要永久有效，可以在my.cnf配置文件里面配置。

2：使用flush hosts

mysql> flush hosts;

Query OK, 0 rows affected (0.00 sec)

mysql> select * from performance_schema.host_cache;

Empty set (0.00 sec)

mysql>

当然你也可以mysqladmin flush-hosts 命令清理一下hosts cache信息

[root@DB-Server ~]# mysqladmin --port=3306 -uroot -p flush-host

Enter password:

那么host cache是什么呢？ 官方介绍如下：

The MySQL server maintains a host cache in memory that contains information about clients: IP address, host name, and error information. The server uses this cache for nonlocal TCP connections. It does not use the cache for TCP connections established using a loopback interface address (127.0.0.1 or ::1), or for connections established using a Unix socket file, named pipe, or shared memory.

简单来说，就是MySQL服务器在内存中维护一个包含客户端信息的缓存：IP地址，主机名和错误信息等。 服务器会将非本地TCP连接信息缓存起来。它不会缓存使用环回接口地址（127.0.0.1或者:: 1）建立的TCP连接，或者使用Unix套接字文件，命名管道或共享内存建立的连接。host cache信息可以通过performance_schema数据库下的host_cache表查询。

3：将变量host_cache_size设置为0

其实我想说这是一个最不靠谱的解决方法，只是让MySQL服务器不记录host cache信息而已。完全可以忽略这个方法。

mysql> show variables like ‘%host_cache_size%’;
±----------------±------+
| Variable_name | Value |
±----------------±------+
| host_cache_size | 279 |
±----------------±------+
1 row in set (0.00 sec)

mysql> set global host_cache_size=0;
Query OK, 0 rows affected (0.00 sec)

mysql> select * from performance_schema.host_cache;
Empty set (0.00 sec)

mysql>