SAP系统 SYSFAIL 失败TCPSBUILD 案例一则

客户的SAP系统，遇到大量的qRFC事务失败，内部显示SYSFAIL，在操作系统日志看到大量如下错误：
Spool / print: Database error: TemSe->RTAB-S/G(8)->1 for table TCPSBUILD key 

对于这个错误，在SAP Notes 864267得到了一些信息。这个Note描述的是进行Offline Backup之后，系统codepage转换出现错误：

After database offline backup, system runs into codepage converter erros.
Syslog (SM21):
=============

The syslog (SM21) contains messages like
SYSLOG : BV4 Work process is in reconnect status.

followed by
SYSLOG : F28 Character set converter cannot get semaphore (SemRq-> 1)
SYSLOG : R32 001 units of semaphore 012 locked

or

SYSLOG : CPR No Code Page Conversion 'xxxx' -> 'yyyy': CCC->CCC
SYSLOG : F6H Database error: TemSe->RTAB-S/G(8)->1 for table TCPSBUILD key zzzz

or in ST22: short dump CONVERSION_CODEPAGE_UNKNOWN caused by an import statemen, f.g.  import xxx to yyy

进一步的解释：
Reason and Prerequisites
If a  work process is in reconnect state and calls the codepage converter, and if the  converter has to load a codepage from database into the converter cache, the codepage converter will return an internal error without releasing the converter semaphore (semaphore 12). Afterwards, when the WP is reconnected, reorganization of the converter cache fails because the WP's cannot get the semaphore.  The WP which has the semaphore does not start any further attempts to  reorganize again.

由于是WP进程重试连接，不能成功获得信号量导致，我们首先尝试重新启动应用和数据库，完成这些工作后，所有新创建的WP进程能够正确处理连接和转换，以上的问题就不再出现。