ORA-609 错误分析及解决方法

某个客户数据库在巡检的时候发现alert日志里不定期会出现ORA-609错误，大致内容如下：

***********************************************************************

Fatal NI connect error 12537, connecting to:

 (LOCAL=NO)

  VERSION INFORMATION:

TNS for HPUX: Version 11.2.0.3.0 - Production

Oracle Bequeath NT Protocol Adapter for HPUX: Version 11.2.0.3.0 - Production

TCP/IP NT Protocol Adapter for HPUX: Version 11.2.0.3.0 - Production

  Time: 19-OCT-2014 20:24:16

  Tracing not turned on.

  Tns error struct:

    ns main err code: 12537

    

TNS-12537: TNS:connection closed

    ns secondary err code: 12560

    nt main err code: 0

    nt secondary err code: 0

    nt OS err code: 0

opiodr aborting process unknown ospid (2734) as a result of ORA-609

Sun Oct 19 21:27:24 2014

***********************************************************************

由于ORA-609的缘故，ospid(xxxx)进程被aborting了，同时还伴随着TNS-12537的错误，连接关闭

去MOS搜了一圈，正好有篇文档是针对这个错误的，下面是描述：

适用于：

Oracle Net Services - Version 11.2.0.1 to 11.2.0.3 [Release 11.2]
Information in this document applies to any platform.

症状：

alert日志出现以上类似的内容（略）

变化：

Changes in database server load, client connect descriptor, changes in network infrastructure (firewall configuration).

原因：

首先，这个“opiodr aborting process unknown ospid (2734) as a result of ORA-609”消息仅仅是说明了由于ORA-609，使Oracle数据库专用进程被关闭了

来看一段描述：

ORA-609 means  "could not attach to incoming connection" so the database process was 'aborted' (closed) because it couldn't attach to the incoming connection passed to it by the listener.

ORA-609意味着不能通过监听把它附加到即将到来的连接上，因此服务器进程被终止（关闭）

The reason for this is found in the sqlnet error stack, in our case is:
   TNS-12537: TNS:connection closed.
Basically the dedicated process didn't have a client connection anymore to work with.

客户端连接有6个步骤:

1. Client initiates a connection to the database so it connects to the listener
2. Listener starts (fork) a dedicated database process that will receive this connection (session)
3. After this dedicated process is started, the listener passes the connection from the client to this process
4. The server process takes the connection from the listener to continue the handshake with the client
5. Server process and client exchange information required for establishing a session (ASO, Two Task Common, User logon)
6. Session is opened
   
   
   

In the case of the above error the connection from the client was closed somewhere between 3. and 4. So when the dedicated process tries to communicate with the client it finds that connection closed.

鉴于以上的错误，在第3步与第4步之间时， 客户端连接就关闭了 ，此时当专有进程尝试与客户端连接时，发现连接已经关闭了

To determine the client which hit this problem we can try to match the timestamp of the error from alert log with an entry in listener.log, but this might be difficult in case of a loaded listener with many incoming connections per second.
Server sqlnet trace will not provide any information about the client.

去确定碰到问题的client，我们可以尝试去匹配alert日志中错误发生的时间戳并且在监听日志中也有相应的条目，但当 加载的监听 每秒有许多连接的时候是非常困难去判断的，服务器sqlnet的trace不会提供任何该客户端的信息

We can enable sqlnet server trace to catch the error (the match is done based on the ospid found in sqlnet server trace file name and the line with ORA-609 error):

还可以启用sqlnet server的trace中抓取到ORA-609错误，匹配成功基于sqlnet server trace文件名和ORA-609错误信息中的ospid

nscon: doing connect handshake...
    nscon: recving a packet
    nsprecv: entry
    nsprecv: reading from transport...
    nttrd: entry
    nttrd: exit
    ntt2err: entry
    ntt2err: Read unexpected EOF ERROR on 15    <<<<<<< error
    ntt2err: exit
    nsprecv: error exit
    nserror: entry
    nserror: nsres: id=0, op=68, ns=12537, ns2=12560; nt[0]=507, nt[1]=0, nt[2]=0; ora[0]=0, ora[1]=0, ora[2]=0
    nscon: error exit
    nsdo: nsctxrnk=0
    nsdo: error exit
    nsinh_hoff: error recving request

可能引起问原因：

Several possible situations can cause this to happen:

•     client changed its mind and closed the connection immediately after initiating it
•     client crashed
•     firewall kills the connection
•     some oracle timeout set on client

解决方案：

Because the entry from listener.log contains only CONNECT_DATA and CID related information we need to check the client configuration for any sqlnet  timeouts:

• possible timeouts in sqlnet.ora in client oracle home:

    sqlnet.outbound_connect_time
    sqlnet.recv_timeout
    sqlnet.send_timeout
    tcp_connect_timeout

检查客户端目录中sqlnet.ora的超时设置，通常是这个引起的

• possible timeout in client connect descriptor (hardcoded in client application or in client tnsnames.ora):

    connect_timeout

检查客户端应用或客户单tnsnames.ora中的超时参数