Goldengate Integrated Capture hungs when reading from Oracle 19c

一、OGG集成抽取19c db hang

在找其它资料时，翻阅到了这个MOS，觉得比较有意思就翻译发一篇博客,有遇到的小伙伴可以借鉴一下。

问题现象：

Goldengate Integrated Capture hungs when reading from Oracle 19c (Doc ID 2693598.1)
APPLIES TO:
Oracle GoldenGate - Version 19.0.0.0.0 and later
Information in this document applies to any platform.
SYMPTOMS
Integrated extract is hung . 

In database alert log :

LOGMINER: skipped redo. Thread 2, RBA 0x00xxx.xxx.xxx, nCV 5
LOGMINER: session# 4 op 4.9 (Transaction Block)
就是说使用OGG19版本,对19C db集成模式抽取时,rba卡着不动了，db alert存在skipped redo的信息！！！

二、问题处理

CAUSE
The problem is introduced when Goldengate extract is configured to read from Oracle 19c.
Database Bug 30414679 - GoldenGate Extract or other LogMiner clients hang.
GoldenGate or other LogMiner clients hang when mining through redo that contains more than 4 change vectors.
The alert log indicates a message "LOGMINER: skipped redo..." prior to the hang.
SOLUTION
 1) Apply the database patch .
 2) As a workaround ,  use  dbms_logmnr_internal.skip_rba() to skip the RBA , here is the example :
LOGMINER: skipped redo. Thread 2, RBA 0x001e5d.0026a808.0138, nCV 5
LOGMINER: session# 1 op 4.9 (Transaction Block)
Get output of below query :
select session#,resetlogs_change#,RESET_TIMESTAMP from  system.logmnr_session$;
 SESSION# RESETLOGS_CHANGE# RESET_TIMESTAMP
---------- ----------------- ---------------
        1                 1       1000000000
1st number, the thread number from the message, here it is 2
2nd number, the sequence number of the redo log, in this case, it is the first number in the RBA field,
 0x001e5d in hex, which is 7773 in decimal.
3rd number, the block number of the redo log block, in this case, it is the second number in the RBA field,
 0026a808 in hex, which is 2533384 in decimal.
4th number, the offset within the block, in this case, is the third number in the RBA field, 0138 in hex,
 which is 312 in decimal.
5th number, <RESETLOGS_CHANGE#> value from the above command output.
6th number, <RESET_TIMESTAMP> value from the above command output.
7th number, <SESSION#> value from the above command output.
The command should be :


SQL> exec dbms_logmnr_internal.skip_rba(2,7773,2533384,312,1,1000000000,1);
Note :  Make sure the extract is stopped before running above command.
Then restart the extract process.

Oracle提供两种解决方法：
1.出现这个Bug,对DB安装补丁;
2.临时性跳过,通过上述的规则进行跳过Rba,但是跳过这个rba,这个rba代表的日志能否正常抽取呢？ OGG没说,那样是否存在数据丢失？
3.个人建议，如果是抽取最前期的阶段,可以尝试删除重来

NEW_OGGPRO=ecc
NEW_OGG_FILE=ecc/ec
ggsci <<EOF
dblogin USERIDALIAS targetogg
delete ${NEW_OGGPRO}
unregister EXTRACT ${NEW_OGGPRO} database
add extract ${NEW_OGGPRO},INTEGRATED TRANLOG,BEGIN NOW
add exttrail ./dirdat/${NEW_OGG_FILE},extract ${NEW_OGGPRO}, megabytes 100
register extract ${NEW_OGGPRO} database CONTAINER(FPROD)
START ${NEW_OGGPRO}
exit
EOF