使用 SQLNET.EXPIRE_TIME 清除僵死连接

本文涉及的产品
传统型负载均衡 CLB,每月750个小时 15LCU
EMR Serverless StarRocks,5000CU*H 48000GB*H
应用型负载均衡 ALB,每月750个小时 15LCU
简介:     数据库连接的客户端异常断开后,其占有的相应并没有被释放,如从v$session视图中依旧可以看到对应的session处于inactive,且对应的服务器进程也没有释放,导致资源长时间地被占用,对于这种情形开该如何处理呢?SQLNET.EXPIRE_TIME对于这个问题我们提供了解决方案,专门用于清理那些异常断开的情形,如网络异常中断,客户端异常掉电,异常重启等。

    数据库连接的客户端异常断开后,其占有的相应并没有被释放,如从v$session视图中依旧可以看到对应的session处于inactive,且对应的服务器进程也没有释放,导致资源长时间地被占用,对于这种情形开该如何处理呢?SQLNET.EXPIRE_TIME对于这个问题我们提供了解决方案,专门用于清理那些异常断开的情形,如网络异常中断,客户端异常掉电,异常重启等。本文描述了设置SQLNET.EXPIRE_TIME参数以及演示死连接以及资源被释放的情形。

 

1、理解SQLNET.EXPIRE_TIME参数
   Use parameter SQLNET.EXPIRE_TIME to specify a the time interval, in minutes, to send a probe to verify that client/server
   connections are active.
   Setting a value greater than 0 ensures that connections are not left open indefinitely, due to an abnormal client termination.
   If the probe finds a terminated connection, or a connection that is no longer in use, it returns an error, causing the
   server process to exit.
   This parameter is primarily intended for the database server,which typically handles multiple connections at any one time.
   
   通过设定参数为非零值(分钟)来发送探测包以检查客户端的异常断开。一旦探测包找到了异常的连接将返回错误,清除对应的server process
   下面是参数使用的一些限制。(缺省值为0,最小值0,建议值10。SQLNET.EXPIRE_TIME=10)
   Limitations on using this terminated connection detection feature are:
   
      It is not allowed on bequeathed connections.
      Though very small, a probe packet generates additional traffic that may downgrade network performance.
      Depending on which operating system is in use, the server may need to perform additional processing to distinguish
      the connection probing event from other events that occur. This can also result in degraded network performance.

 

2、Dead Connection Detection (DCD)与Inactive Sessions

Dead connections:
   These are previously valid connections with the database but the connection between the client and server processes has
   terminated abnormally.
   Examples of a dead connection:
   - A user reboots/turns-off their machine without logging off or disconnecting from the database.
   - A network problem prevents communication between the client and the server.
   
   In these cases, the shadow process running on the server and the session in the database may not terminate.
   
   Implemented by
         * adding SQLNET.EXPIRE_TIME = <MINUTES> to the sqlnet.ora file
   
   With DCD is enabled, the Server-side process sends a small 10-byte packet to the client process after the duration of
   the time interval specified in minutes by the SQLNET.EXPIRE_TIME parameter.
   
   If the client side connection is still connected and responsive, the client sends a response packet back to the database
   server, resetting the timer..and another packet will be sent when next interval expires (assuming no other activity on
   the connection).
   
   If the client fails to respond to the DCD probe packet
        * the Server side process is marked as a dead connection and
        * PMON performs the clean up of the database processes / resources
        * The client OS processes are terminated
   
   NOTE: SQLNET.RECV_TIMEOUT can be set on the SERVER side sqlnet.ora file. This will set a timeout for the server process
         to wait for data from the client process.

Inactive Sessions:
   These are sessions that remain connected to the database with a status in v$session of INACTIVE.
   Example of an INACTIVE session:
   - A user starts a program/session, then leaves it running and idle for an extended period of time.

 

3、配置SQLNET.EXPIRE_TIME

#对于SQLNET.EXPIRE_TIME的配置,需要修改sqlnet.ora,然后添加SQLNET.EXPIRE_TIME项
[oracle@orasrv admin]$ more sqlnet.ora
sqlnet.expire_time = 1     #仅仅需要配置此项,后面的各项仅仅是为了生成跟踪日志,可省略
TRACE_LEVEL_SERVER = 16 
TRACE_FILE_SERVER = SERVER
TRACE_DIRECTORY_SERVER= /u01/app/oracle/network/trace 
TRACE_TIMESTAMP_ SERVER = ON 
TRACE_UNIQUE_SERVER = ON
DIAG_ADR_ENABLED=OFF

4、模拟及测试DCD连接

C:\Users\robinson.cheng>sqlplus scott/tiger@ora11g    --->从windows客户端发起连接

SQL*Plus: Release 11.2.0.1.0 Production on Tue Jun 25 09:57:59 2013

Copyright (c) 1982, 2010, Oracle.  All rights reserved.

Connected to:
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production
With the Partitioning, OLAP, Data Mining and Real Application Testing options

---Issued the sql to hold a lock
SQL> update emp set sal=sal*1.1 where deptno=20;   

5 rows updated.

--disabled the network adapter in VM setting
SQL> select * from dual;
select * from dual
       *
ERROR at line 1:
ORA-03113: end-of-file on communication channel
Process ID: 29522
Session ID: 15 Serial number: 447

--服务器端环境   
SQL> select * from v$version where rownum<2;  
  
BANNER  
--------------------------------------------------------------------------------   
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production  

--在服务器端查看session的情况,SCOTT用户的session状态为INACTIVE
SQL> @comm_sess_users;

+----------------------------------------------------+
| User Sessions (All)                                |
+----------------------------------------------------+
Instance     SID Serial ID    Status Oracle User     O/S User O/S PID Session Program              Terminal       Machine
--------- ------ --------- --------- ----------- ------------ ------- -------------------------- ---------- -------------
ora11g        15       447  INACTIVE       SCOTT Robinson.Che   29522 sqlplus.exe                      PC39  TRADESZ\PC39
             125         5  INACTIVE         SYS       oracle    4734 sqlplus@orasrv.com (TNS V1      pts/0    orasrv.com
             139         9    ACTIVE         SYS       oracle   29447 sqlplus@orasrv.com (TNS V1      pts/4    orasrv.com

--Get the spid for user scott by SID
SQL> @my_spid_from_sid
Enter value for input_sid: 15
old   4: AND s.sid = &input_sid
new   4: AND s.sid = 15

   SID    SERIAL# SPID
------ ---------- ------------------------
    15        447 29522

--To find the locked object
SQL> @lock_obj

OBJECT_NAME||''||LOCKED_MODE||''||CTIME||''||C.SID||''||SERIAL#
------------------------------------------------------------------
EMP   3   14   15  447
EMP   3   83   15  447

--The trace file exists
SQL> ho ls -hltr /u01/app/oracle/network/trace/s*29522*
-rw-r----- 1 oracle oinstall 241K Jun 25 09:59 /u01/app/oracle/network/trace/server_29522.trc

--->try to issue another sql. the sql is blocked
SQL> set time on;
10:03:46 SQL> delete scott.emp where deptno=20;  
delete scott.emp where deptno=20
             *
ERROR at line 1:
ORA-01013: user requested cancel of current operation

--Check the server process for scott 
10:04:37 SQL> ho ps -ef | grep 29522 | grep -v grep
oracle   29522     1  0 09:58 ?        00:00:00 oracleora11g (LOCAL=NO)

--Could not reach to client from server.
10:06:51 SQL> ho ping 192.168.7.133
PING 192.168.7.133 (192.168.7.133) 56(84) bytes of data.
From 192.168.7.40 icmp_seq=2 Destination Host Unreachable
From 192.168.7.40 icmp_seq=3 Destination Host Unreachable
From 192.168.7.40 icmp_seq=4 Destination Host Unreachable
From 192.168.7.40 icmp_seq=6 Destination Host Unreachable
From 192.168.7.40 icmp_seq=7 Destination Host Unreachable
From 192.168.7.40 icmp_seq=8 Destination Host Unreachable

--此时总进程数为27个
10:15:08 SQL> select count(*) from v$process;

  COUNT(*)
----------
        27

--从09:58进程启动开始到10:17:59进程依旧没有被释放
10:17:59 SQL> ho ps -ef | grep 29522 | grep -v grep
oracle   29522     1  0 09:58 ?        00:00:00 oracleora11g (LOCAL=NO)

-->At this time the server process was released
10:18:08 SQL> ho ps -ef | grep 29522 | grep -v grep

--进程释放后此时进程总数变为26个
10:19:45 SQL> select count(*) from v$process;

  COUNT(*)
----------
        26

-->the lock was released
10:19:54 SQL> @lock_obj

no rows selected

--Author : Robinson
--Blog   : http://blog.csdn.net/robinson_0612

--scott用户的session已经从v$session中被移除
10:20:03 SQL> @comm_sess_users;

+----------------------------------------------------+
| User Sessions (All)                                |
+----------------------------------------------------+

Instance     SID Serial ID    Status    Oracle User     O/S User O/S PID Session Program            Terminal    Machine
--------- ------ --------- --------- -------------- ------------ ------- -------------------------- -------- ----------
ora11g       125         5  INACTIVE            SYS       oracle    4734 sqlplus@orasrv.com (TNS V1    pts/0 orasrv.com
             139         9    ACTIVE            SYS       oracle   29447 sqlplus@orasrv.com (TNS V1    pts/4 orasrv.com

5、查看SQLNET.EXPIRE_TIME是否启用

#下面对跟踪日志过滤,可以看到09:58:02:853中提示开启dead connection detection
[oracle@orasrv trace]$ cat -n server_29522.trc |grep dead
    78  [25-JUN-2013 09:58:02:853] niotns: Enabling dead connection detection (1 min)

#下面的查询中,在09:58:03 timer被启动,10:18:26后,连接被彻底关闭(包括server process)    
[oracle@orasrv trace]$ cat -n server_29522.trc |grep timer
   447  [25-JUN-2013 09:58:03:050] nstimstart: starting timer at 25-JUN-2013 09:58:03
   451  [25-JUN-2013 09:58:03:051] nsconbrok: timer created for connection
  4092  [25-JUN-2013 10:18:26:173] nstimarmed: timer is armed, with value 3833

#下面是starting timer的详细信息  
[oracle@orasrv trace]$ head -451 server_29522.trc | tail -5
[25-JUN-2013 09:58:03:050] nstimstart: starting timer at 25-JUN-2013 09:58:03
[25-JUN-2013 09:58:03:051] nstimset: entry
[25-JUN-2013 09:58:03:051] nstimset: normal exit
[25-JUN-2013 09:58:03:051] nstimstart: normal exit
[25-JUN-2013 09:58:03:051] nsconbrok: timer created for connection 

#下面是timer被清除后的详细信息nstimclear: normal exit
[oracle@orasrv trace]$ head -4097 server_29522.trc | tail -7
[25-JUN-2013 10:18:26:173] nstimarmed: entry
[25-JUN-2013 10:18:26:173] nstimarmed: timer is armed, with value 3833
[25-JUN-2013 10:18:26:173] nstimarmed: normal exit
[25-JUN-2013 10:18:26:173] nstimclear: entry
[25-JUN-2013 10:18:26:173] nstimclear: normal exit
[25-JUN-2013 10:18:26:173] nttctl: entry
[25-JUN-2013 10:18:26:173] nttctl: entry 

6、小结
a、DCD连接通常指用户没有正常断开连接而重启客户端,关机以及网络问题导致客户端无法与服务器正常通信所致的连接
b、相对于DCD连接,INACTIVE session则是用户建立连接之后,尚未执行任何操作或操作已经完成但没有断开,等同于与处于idle状态
c、无论是DCD连接,还是出于idle状态的INACTIVE session,在v$session视图呈现的都是INACTIVE状态
d、对于使用resource_limit及profile配置后用户session超出idle_time的情形,在v$session视图呈现sniped状态
e、当在sqlnet.ora配置文件中设置了SQLNET.EXPIRE_TIME参数为非零值时,僵死连接在EXPIRE_TIME指定的时间后被清除
f、演示中仅仅设定EXPIRE_TIME为1分钟,而实际的释放时间接近20分钟左右,什么原因尚不清楚,有待进一步测试
g、设定SQLNET.EXPIRE_TIME为非零值之后,系统需要产生而外的开销以及带来网络性能的下降
h、对于需要及时释放OS及DB资源的情形,Oracle建议使用resource_limit 及 profile 限制用户连接的同时并设定SQLNET.EXPIRE_TIME为非零值
i、Reference: [ID 206007.1] [ID 395505.1] [ID 601605.1] [ID 151972.1]

 

Oracle&nbsp;牛鹏社

 

更多参考

有关Oracle RAC请参考
     使用crs_setperm修改RAC资源的所有者及权限
     使用crs_profile管理RAC资源配置文件
     RAC 数据库的启动与关闭
     再说 Oracle RAC services
     Services in Oracle Database 10g
     Migrate datbase from single instance to Oracle RAC
     Oracle RAC 连接到指定实例
     Oracle RAC 负载均衡测试(结合服务器端与客户端)
     Oracle RAC 服务器端连接负载均衡(Load Balance)
     Oracle RAC 客户端连接负载均衡(Load Balance)
     ORACLE RAC 下非缺省端口监听配置(listener.ora tnsnames.ora)
     ORACLE RAC 监听配置 (listener.ora tnsnames.ora)
     配置 RAC 负载均衡与故障转移
     CRS-1006 , CRS-0215 故障一例 
     基于Linux (RHEL 5.5) 安装Oracle 10g RAC
     使用 runcluvfy 校验Oracle RAC安装环境

有关Oracle 网络配置相关基础以及概念性的问题请参考:
     配置非默认端口的动态服务注册
     配置sqlnet.ora限制IP访问Oracle
     Oracle 监听器日志配置与管理
     设置 Oracle 监听器密码(LISTENER)
     配置ORACLE 客户端连接到数据库

有关基于用户管理的备份和备份恢复的概念请参考
     Oracle 冷备份
     Oracle 热备份
     Oracle 备份恢复概念
     Oracle 实例恢复
     Oracle 基于用户管理恢复的处理
     SYSTEM 表空间管理及备份恢复
     SYSAUX表空间管理及恢复
     Oracle 基于备份控制文件的恢复(unsing backup controlfile)

有关RMAN的备份恢复与管理请参考
     RMAN 概述及其体系结构
     RMAN 配置、监控与管理
     RMAN 备份详解
     RMAN 还原与恢复
     RMAN catalog 的创建和使用
     基于catalog 创建RMAN存储脚本
     基于catalog 的RMAN 备份与恢复
     RMAN 备份路径困惑
     使用RMAN实现异机备份恢复(WIN平台)
     使用RMAN迁移文件系统数据库到ASM
     linux 下RMAN备份shell脚本
     使用RMAN迁移数据库到异机

有关ORACLE体系结构请参考
     Oracle 表空间与数据文件
     Oracle 密码文件
     Oracle 参数文件
     Oracle 联机重做日志文件(ONLINE LOG FILE)
     Oracle 控制文件(CONTROLFILE)
     Oracle 归档日志
     Oracle 回滚(ROLLBACK)和撤销(UNDO)
     Oracle 数据库实例启动关闭过程
     Oracle 10g SGA 的自动化管理
     Oracle 实例和Oracle数据库(Oracle体系结构) 

相关实践学习
SLB负载均衡实践
本场景通过使用阿里云负载均衡 SLB 以及对负载均衡 SLB 后端服务器 ECS 的权重进行修改,快速解决服务器响应速度慢的问题
负载均衡入门与产品使用指南
负载均衡(Server Load Balancer)是对多台云服务器进行流量分发的负载均衡服务,可以通过流量分发扩展应用系统对外的服务能力,通过消除单点故障提升应用系统的可用性。 本课程主要介绍负载均衡的相关技术以及阿里云负载均衡产品的使用方法。
目录
相关文章
|
11月前
|
SQL druid 关系型数据库
MySQL连接超时时间wait_timeout导致间歇性报错:communication link failure
MySQL连接超时时间wait_timeout导致间歇性报错:communication link failure
687 1
PG14中的idle_session_timeout
PG14中的idle_session_timeout
114 0
|
关系型数据库 MySQL
二进制日志的保存时间参数binlog_expire_logs_seconds和expire_logs_days的设置
在MySQL 8.0中,默认的二进制日志的保存时间参数binlog_expire_logs_seconds和expire_logs_days的设置如下
274 0
如何使用time_expire绝对超时时间-参数解读系列
说明:    time_expire,绝对超时时间,格式为yyyy-MM-dd HH:mm。注:1)以支付宝系统时间为准;2)如果和timeout_express参数同时传入,以time_expire为准。
2180 12
|
机器学习/深度学习 网络协议 测试技术
[20180124]测试SQLNET.EXPIRE_TIME参数3
[20180124]测试SQLNET.EXPIRE_TIME参数3.txt --//昨天测试SQLNET.EXPIRE_TIME参数时,链接如下: http://blog.
1237 0
|
机器学习/深度学习 网络协议 Oracle
[20180123]测试SQLNET.EXPIRE_TIME参数2
[20180123]测试SQLNET.EXPIRE_TIME参数2.txt --//曾经写过一篇linux内核网络参数测试tcp_keepalive,链接http://blog.itpub.net/267265/viewspace-2138391/ --//测试服务端会定时发起连接监测与client的连接状态. 参数解析: /proc/sys/net/ipv4/tcp_keepalive_time    当keepalive起用的时候,TCP发送keepalive消息的频度。
1033 0
|
机器学习/深度学习 网络协议 Oracle
[20180123]测试SQLNET.EXPIRE_TIME参数.txt
[20180123]测试SQLNET.EXPIRE_TIME参数.txt --//曾经写过一篇linux内核网络参数测试tcp_keepalive,链接http://blog.itpub.net/267265/viewspace-2138391/ --//测试服务端会定时发起连接监测与client的连接状态. 参数解析: /proc/sys/net/ipv4/tcp_keepalive_time    当keepalive起用的时候,TCP发送keepalive消息的频度。
965 0
|
网络协议 关系型数据库 MySQL