Hbase区域服务器中的异常

我不断看到这些日志打印在hbase区域服务器日志中。 10.0.5.133是hbase java客户端所在的机器

它不断地断开和连接。为什么会这样？ 客户端库/服务器版本是否不匹配？还有其他原因吗？我试图用谷歌搜索这个，但找不到任何有用的链接

2015-12-25 20:24:03,210 DEBUG [RpcServer.reader=9,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:54538 because read count=-1. Number of active connections: 30
2015-12-25 20:24:03,213 DEBUG [RpcServer.reader=2,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:54564 because read count=-1. Number of active connections: 29
2015-12-25 20:24:08,350 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:54642; # active connections: 26
2015-12-25 20:24:18,358 DEBUG [RpcServer.reader=8,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:54642 because read count=-1. Number of active connections: 29
2015-12-25 20:24:30,543 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:54799; # active connections: 28
2015-12-25 20:24:40,547 DEBUG [RpcServer.reader=0,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:54799 because read count=-1. Number of active connections: 29
2015-12-25 20:24:41,870 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:54864; # active connections: 29
2015-12-25 20:24:51,877 DEBUG [RpcServer.reader=8,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:54864 because read count=-1. Number of active connections: 28
2015-12-25 20:25:00,634 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:54982; # active connections: 27
2015-12-25 20:25:10,645 DEBUG [RpcServer.reader=6,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:54982 because read count=-1. Number of active connections: 30
2015-12-25 20:26:30,169 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:55511; # active connections: 26
2015-12-25 20:26:40,176 DEBUG [RpcServer.reader=3,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:55511 because read count=-1. Number of active connections: 32
2015-12-25 20:26:41,117 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:55594; # active connections: 30
2015-12-25 20:26:44,607 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:55623; # active connections: 31
2015-12-25 20:26:51,129 DEBUG [RpcServer.reader=7,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:55594 because read count=-1. Number of active connections: 26
2015-12-25 20:26:54,614 DEBUG [RpcServer.reader=2,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:55623 because read count=-1. Number of active connections: 23
2015-12-25 20:27:01,926 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:55731; # active connections: 31
2015-12-25 20:27:11,938 DEBUG [RpcServer.reader=5,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:55731 because read count=-1. Number of active connections: 29
2015-12-25 20:27:14,711 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:55776; # active connections: 28
2015-12-25 20:27:24,719 DEBUG [RpcServer.reader=3,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:55776 because read count=-1. Number of active connections: 28
2015-12-25 20:27:36,791 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:55938; # active connections: 24
2015-12-25 20:27:45,008 DEBUG [RpcServer.listener,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: connection from 10.0.5.133:56013; # active connections: 27
2015-12-25 20:27:46,799 DEBUG [RpcServer.reader=4,port=60020] ipc.RpcServer: RpcServer.listener,port=60020: DISCONNECTING client 10.0.5.133:55938 because read count=-1. Number of active connections: 29

---

看起来客户端连接负载远远超过默认的最大客户端连接限制30。你可能想考虑一下这个配置。

以下是更多信息：

hbase.zookeeper.property.maxClientCnxns

动物园管理员的财产 config zoo.cfg。对并发连接数的限制（在 套接字级别），由IP地址标识的单个客户端可以 致动物园管理员团的一名成员。设置高以避免zk 独立运行和伪分布式运行的连接问题

默认值：30

---

hbase中hbase.zookeeper.property.maxClientCnxns属性的值这些是到hbase区域服务器的连接，它们与来自clientZookeeper的zookeeper连接的关系如何zookeeper是在hbase中管理客户端连接的内容。这正是您看到该配置的原因。Zookeeper将在任何新连接达到设置的限制后断开连接。区域服务器是zookeeper集成的一部分。此限制是针对每个客户端IP的，对吗？不允许连接到特定区域服务器的总连接数。如果是这种情况，那么我看不出30个数字有任何问题，但是如果它是区域服务器的总限制，那么我肯定需要增加这个数字。您是对的，限制是每个ip。现在，根据部署和客户端负载，您需要调整此配置。如果您的仲裁只有1或2个RS，并且您每秒收到数千个读取请求，那么您会看到客户端连接数量的增加，但zookeeper正在限制它。这些都有意义吗？