操作系统：windows server 2008R2

数据库版本：oracle 11.2.0.3 单实例

今天，一开发同事找来，说他的存储过程在龙岩地市执行报错ORA-12828，说代码执行到如下代码块的时候报错退出：

EXECUTE IMMEDIATE 'ALTER TABLE FACT_COSTS TRUNCATE PARTITION P_FACT_COSTS_'||PRM_MONTH||' update global indexes';

查看oracle联机文档对ORA-12828的解释：

[oracle@se31 ~]$ oerr ora 12828

12828, 00000, "Can't start parallel transaction at a remote site"

// *Cause: PDML transaction cannot be started because we are not in the

//         coordinator site of the distributed transaction.

// *Action: Do not use PDML at remote sites.

[oracle@se31 ~]$

查询oracle support知识库，有篇文档(ID 1535660.1)与ORA-12828相关

APPLIES TO:

Oracle Database - Enterprise Edition - Version 11.2.0.1.0 to 11.2.0.4 [Release 11.2]

Information in this document applies to any platform.

SYMPTOMS

The remote parallel index rebuild runs serial even if enabling parallel replication propagation at the database link level. For example if a database link is created for a particular destination database, and enabling parallel propagation for a database link, then Oracle uses multiple parallel slave processes to replicate to the corresponding destination. But it doesn't work while parallel index rebuild remotely through a procedure. In detail:

CAUSE

As theBug 14684769states, this works as expected and it is not supported to parallelize the index by using the dblink.

SOLUTION

The dblink is not supported to parallelize the rebuild index.

As per the fix for Bug 2166879, create index will not be compiled for parallel execution if it is being executed in a distributed txn.

REFERENCES

BUG:14684769- REMOTE PARALLEL INDEX REBUILD RUNS SERIAL

BUG:2166879- ORA-12828 WHEN INDEX CREATED IN PARALLEL MODE THROUGH REMOTE PROCEDURE

查看开发同事提供的代码并没有使用到dblink，根据oracle官方(ID 1535660.1)提示，ORA12828与索引重建有关，

让开发同事尝试去掉update global indexes子句重新执行，可成功执行。其实更新表数据同步更新索引，是防止分区表分区索引失效。

于是，建议他将索引重建子句去除，另外在存储过程中添加专门的索引重建语句(经他测试，可以执行)。

另外，根据他提供的信息该代码并不是在所有地市都报错，只有目前的龙岩windows server 2008R2 oracle 11.2.0.3单实例下。