1、背景
goldengate的基本原理是：按最简单的单项复制结构为例
- 由3个模块构成：extract抽取、pump投递、replicat复制；2个文件trail：主机本地队列、备机远程队列。
- 在主机侧：extract抽取redo中的事务，放到本地trail中；然后pump把本地trail投递到远端备机的远程trail。
- 在备机侧：replicat读取远程trail，构造sql，在备机执行。完成复制。
进程的失败无非是事务的失败。 因此学会查看流程中的事务信息，对维护极有帮助。
2、从日志查看失败或当前的事务位置
2.1 从错误日志ggserr.log查看
2020-06-03 16:56:47 WARNING OGG-01396 Oracle GoldenGate Delivery for Oracle, rept01.prm:
A complete after image is not available in BDE.TXT_KONFIG at rba 30971697 in file /home/oracle/ggate/dirdat/r1000198, while inserting a row into BDE.TXT_KONFIG due to missing target row for a key update operation.
NOCOMPRESSUPDATES or FETCHOPTIONS FETCHPKUPDATECOLS may be specified in the EXTRACT parameter file to include a complete image for key update operations.
注意，事务的位置是：trail文件r1000198，rba事务号30971697。
2.2 从discard文件查看
Current time: 2020-06-04 09:27:21
OCI Error ORA-02291: integrity constraint (ML.CS_LOG_FOR_CS_CHARGE_ID) violated - parent key not found (status = 2291). INSERT INTO "ML"."CS_LOG" ("KAP_ID","CHARGE_NR","AS_NR","KOMP_INDEX","CS_CHARGE_ID","CS_KAP_STATE_ID","POSITION","ANLAUF","AUSLAUF","CREATED","USER_ID","STAMP","KOMPONENTE_ID","SNR_MAT_ID","KOMP_ANZ","BOOK_QTY","BOOK_QTY_TOTAL","BOOK_STAMP","BOOK_TRANSFERED","PARENT_ID","FLOOR_LIFETIME_REMAIN","SNR_MAT_MSL_SEQ_NO","CS_LOG_ID","DRYPACK_EXPIREDATE","PRE_SETUP_ITEM_SECTION_ID","ORDER_CLUSTER_ID") VALUES (:a0,:a1,:a2,:a3,:a4,:a5,:a6,:a7,:a8,:a9,:a10,:a11,:a12,:a13,:a14,:a15,:a16,:a17,:a18,:a19,:a20,:a21,:a22,:a23,:a24,:a25)
Operation failed at seqno 265 rba 25881602
Discarding record on action DISCARD on error 2291
Problem replicating ML.CS_LOG to ML.CS_LOG
Mapping problem with insert record (target format)...
*
KAP_ID = 301
CHARGE_NR = 4050
AS_NR = 10
KOMP_INDEX = 0
CS_CHARGE_ID = 645234
CS_KAP_STATE_ID = 58598984
POSITION = L304NXT1_6_10
ANLAUF = 2020-06-04 09:33:02.696000000
AUSLAUF = 3000-12-31 23:59:59.999000000
CREATED = 2020-06-04 09:33:07.356000000
USER_ID = -5
STAMP = 2020-06-04 09:33:07.356000000
KOMPONENTE_ID = 110151
SNR_MAT_ID = 1761945
KOMP_ANZ = 4.0000
BOOK_QTY = 0
BOOK_QTY_TOTAL = 0
BOOK_STAMP = NULL
BOOK_TRANSFERED = 1970-01-01 00:00:00.000000000
PARENT_ID = 0
FLOOR_LIFETIME_REMAIN = NULL
SNR_MAT_MSL_SEQ_NO = NULL
CS_LOG_ID = 57287147
DRYPACK_EXPIREDATE = NULL
PRE_SETUP_ITEM_SECTION_ID = NULL
ORDER_CLUSTER_ID = NULL
*
注意，事务位置队列trail位置：seqno 265 rba 25881602
3、用logdump工具查看队列里的事务
logdump是OGG自带的工具，在OGG根目录下。
3.1 打开trail文件
trail文件名对应seqno，如seqno 265–>dirdat/s1000265
Logdump 60 >open dirdat/r1000198
Current LogTrail is /home/oracle/ggate/dirdat/r1000198
3.2 移动到事务位置
事务位置就是rba
Logdump 61 >pos 30971697
Reading forward from RBA 30971697
显示当前事务信息
Logdump 62 >n
2020/06/03 16:56:44.147.117 GGSPKUpdate Len 128 RBA 30971697
Name: BDE.TXT_KONFIG
After Image: Partition 4 G m
002b 0000 000a 0000 0000 0000 0000 00ad 0001 000a | .+..................
0000 0000 0000 0000 5209 0003 000b 0000 0007 3137 | ........R.........17
3633 3534 3500 0000 0a00 0000 0000 0000 0000 ad00 | 63545...............
0100 0a00 0000 0000 0000 0052 0900 0300 0500 0000 | ...........R........
0130 0006 001f 0000 3230 3230 2d30 362d 3033 3a31 | .0......2020-06-03:1
373a 3032 3a32 382e 3235 3430 3030 3030 3000 0700 | 7:02:28.254000000...
0700 0000 0331 3733 | .....173
4、使用场景
- 跳过指定的事务：有时进程的延迟lag很大，看是哪些事务导致的，可以用logdump查看trail的事务内容。
- 查看进程hang住的事务：有时进程hang住，info all查看正常的running状态，错误日志、丢弃日志都没有输出，但就是不同步。可以查看到底是什么事务导致。