如果TCP/IP KeepAlive 设置过长，那么在Pirmary意外down机之后，Standby上的partially filled archived log 无法被正常register，会报ORA-332错误。
修改操作系统的TCP/IP KeepAlive值就可以了。
以下总结了几种操作系统中修改TCP/IP KeepAlive值的方法。
Problem Description:
====================
How to implement Solaris TCP/IP KeepAlive mechanism.
To change the keepalive interval on Solaris (SPARC) servers you have to use the maintenance command ndd.
It is not possible to configure this information in the /etc/system file.
To find out the current value of the TCP/IP keepalive interval use the following command:
ndd /dev/tcp tcp_keepalive_interval
The value returned is in units of milliseconds, eg.
% ndd /dev/tcp tcp_keepalive_interval
7200000
Which in this case works out to 2 hours.
To change this value you use ndd with the '-set' flag, eg:
# ndd -set /dev/tcp tcp_keepalive_interval 1200000
Problem Description:
====================
How to implement Windows NT TCP/IP KeepAlive mechanism.
Problem Explanation:
====================
It is important to implement the Windows NT TCP/IP KeepAlive mechanism to prevent dead sessions from accumulating when SQL*Net/Net8 Dead Connection Detection [SQLNET.EXPIRE_TIME] is set to zero or unset in the SQLNET.ORA file.
Solution Description:
=====================
To implement the Windows NT TCP/IP KeepAlive mechanism, click:
Start > Run > Regedit.exe
In the Registry,navigate to:
HKEY_LOCAL_MACHINE/SYSTEM/CurrentControlSet/Services/Tcpip/Parameters
Edit - Add value: KeepAliveTime (REG_DWORD)
Decimal -> 180000 (for example = 3 minutes)
The parameter will not be present the first time you modify the parameter,
so you will need to add it. The default value is 2 hours.
Problem Description:
====================
How to implement Linux TCP/IP KeepAlive mechanism.
修改/proc/sys/net/ipv4/tcp_keepalive_time
# echo 120 > /proc/sys/net/ipv4/tcp_keepalive_time
