OGG的extract进程checkpoint时间点回到1988-01-01 00:00:00故障处理
1、故障现象
Extract进程（SEXTR01）状态为running，但是Lag at Chkpt却达到5个多小时，且时间一直在增长，根本就不抽取新日志，状态信息如下：
|
GGSCI (calladgdb) 21> info all
Program Status Group Lag at Chkpt Time Since Chkpt
MANAGER RUNNING EXTRACT RUNNING DPEYWGL 00:00:00 00:00:04 EXTRACT RUNNING SEXTR01 05:14:58 00:00:03 |
2、故障分析
2.1 debug一下进程状态信息
|
GGSCI (calladgdb) 23> info sextr01, showch debug
EXTRACT SEXTR01 Last Started 2015-01-25 22:51 Status RUNNING Checkpoint Lag 05:21:26 (updated 00:00:04 ago) Log Read Checkpoint Oracle Redo Logs 2015-01-25 18:05:16 Thread 1, Seqno 5554, RBA 611373536 SCN 3126.136656462 (13426204423758) Log Read Checkpoint Oracle Redo Logs 1988-01-01 00:00:00 Thread 2, Seqno 4016, RBA 1024 SCN 44527.110828288 (191242119617280)
Current Checkpoint Detail:
Read Checkpoint #1 ……
Read Checkpoint #2
Oracle Threaded Redo Log
Startup Checkpoint (starting position in the data source): Thread #: 2 Sequence #: 4015 RBA: 640984080 Timestamp: 2015-01-25 18:03:51.000000 SCN: 3126.136602577 (13426204369873) Redo File: +DGROUP1/calldb/onlinelog/group_15.267.799673791
Recovery Checkpoint (position of oldest unprocessed transaction in the data source): Thread #: 2 Sequence #: 4015 RBA: 640984080 Timestamp: 2015-01-25 18:03:51.000000 SCN: 3126.136602577 (13426204369873) Redo File: +DGROUP1/calldb/onlinelog/group_15.267.799673791
Current Checkpoint (position of last record read in the data source): Thread #: 2 Sequence #: 4016 RBA: 1024 Timestamp: 1988-01-01 00:00:00.000000 SCN: 44527.110828288 (191242119617280) Redo File: +DGROUP1/calldb/onlinelog/group_16.266.799673793 …… |
从这里发现一个不可思议的问题：
Thread #: 2的current checkpoint 的timestamp竟然是1988-01-0100:00:00.000000，SCN的十进制数，竟然为191242119617280，远远大于当前数据库的SCN，多出一位数。
时间倒退，SCN号反而大涨，时间还那么有个性。这个问题挺有意思。
2.2 view extract进程的report
使用view report sextr01查看进程的报告，提示有ERROR，thread#2 4016 archive log 不能打开。（由于笔者没有将信息复制下来，所以当时的状态只能这么描述了），但是4016号archivelog实际物理文件是存在的，而且权限正常。
2.3 将进程停止下来，showch其状态
|
GGSCI (calladgdb) 24>stop sextr01 GGSCI (calladgdb) 25> info SEXTR01, showch EXTRACT SEXTR01 Last Started 2015-01-26 01:00 Status STOPPED Checkpoint Lag 06:56:55 (updated 00:00:58 ago) Log Read Checkpoint Oracle Redo Logs 2015-01-25 18:05:16 Thread 1, Seqno 5554, RBA 611373536 SCN 3126.136656462 (13426204423758) Log Read Checkpoint Oracle Redo Logs 1988-01-01 00:00:00 Thread 2, Seqno 4016, RBA 1024 SCN 44527.110828288 (191242119617280)
Current Checkpoint Detail:
Read Checkpoint #1 ……
Read Checkpoint #2
Oracle Threaded Redo Log
Startup Checkpoint (starting position in the data source): Thread #: 2 Sequence #: 4015 RBA: 640984080 Timestamp: 2015-01-25 18:03:51.000000 SCN: 3126.136602577 (13426204369873) Redo File: +DGROUP1/calldb/onlinelog/group_15.267.799673791
Recovery Checkpoint (position of oldest unprocessed transaction in the data source): Thread #: 2 Sequence #: 4015 RBA: 640984080 Timestamp: 2015-01-25 18:03:51.000000 SCN: 3126.136602577 (13426204369873) Redo File: +DGROUP1/calldb/onlinelog/group_15.267.799673791
Current Checkpoint (position of last record read in the data source): Thread #: 2 Sequence #: 4016 RBA: 1024 Timestamp: 1988-01-01 00:00:00.000000 SCN: 44527.110828288 (191242119617280) Redo File: +DGROUP1/calldb/onlinelog/group_16.266.799673793 …… |
3 分析原因
进程的thread#2的checkpoint time竟然是1988-01-01 00:00:00.000000，肯定是不正常，结合error信息为找不到4016归档日志文件，以及进程的current checkpoint thread#2的sequence#:4016，而RBA却为1024。
既然RBA都已经运行取1024了，就不应该找不到archivelog的情况，如果4016号归档日志文件还没有开始抽取，RBA号就应该是0。
根据此分析，决定手动的将extract进程的chkeckpoint恢得到recover chkeckpoint点，以及next extseqno调整到4016的0号RBA试一下，看能否解决。
4、解决办法
4.1 调整netxt sequno为4016，RBA为0
GGSCI (calladgdb)2> alter SEXTR01, thread 2, extseqno 4016, extrba 0
4.2 调整ioextseqno与ioextrba到进程的rechvercheckpoint状态
GGSCI (calladgdb)3> alter SEXTR01, thread 2, ioextseqno 4015, ioextrba 640984080
4.3 验证调整结果
GGSCI (calladgdb)4> info sextr01, showch
|
EXTRACT SEXTR01 Initialized 2015-01-26 01:48 Status STOPPED Checkpoint Lag 07:43:41 (updated 00:00:51 ago) Log Read Checkpoint Oracle Redo Logs 2015-01-25 18:05:16 Thread 1, Seqno 5554, RBA 611373536 SCN 3126.136656462 (13426204423758) Log Read Checkpoint Oracle Redo Logs 1988-01-01 00:00:00 Thread 2, Seqno 4016, RBA 0 SCN 0.0 (0)
Current Checkpoint Detail:
Read Checkpoint #1 ……
Read Checkpoint #2
Oracle Threaded Redo Log
Startup Checkpoint (starting position in the data source): Thread #: 2 Sequence #: 4016 RBA: 0 Timestamp: 1988-01-01 00:00:00.000000 SCN: Not available Redo File: +DGROUP1/calldb/onlinelog/group_16.266.799673793
Recovery Checkpoint (position of oldest unprocessed transaction in the data source): Thread #: 2 Sequence #: 4015 RBA: 640984080 Timestamp: SCN: Not available Redo File:
Current Checkpoint (position of last record read in the data source): Thread #: 2 Sequence #: 4016 RBA: 0 Timestamp: 1988-01-01 00:00:00.000000 SCN: Not available Redo File: +DGROUP1/calldb/onlinelog/group_16.266.799673793 …… |
Start checkpoint和current chkeckpoint的sequence#都变成了4016，RBA号都变成了0
4.4 启动extract进程sextr01，并做brreset操作
GGSCI (calladgdb)6> startSEXTR01, brreset
|
GGSCI (calladgdb) 6> start SEXTR01, brreset
Sending START request to MANAGER ... EXTRACT SEXTR01 starting
GGSCI (calladgdb) 7> info all
Program Status Group Lag at Chkpt Time Since Chkpt
MANAGER RUNNING EXTRACT RUNNING DPEYWGL 00:00:00 00:00:01 EXTRACT STARTING SEXTR01 07:43:41 00:03:47 |
4.5 多做几次进程状态信息显示，验证延迟下降效果
|
GGSCI (calladgdb) 10> info all
Program Status Group Lag at Chkpt Time Since Chkpt
MANAGER RUNNING EXTRACT RUNNING DPEYWGL 04:46:20 00:00:01 EXTRACT RUNNING SEXTR01 04:23:13 00:00:01
GGSCI (calladgdb) 11> info all
Program Status Group Lag at Chkpt Time Since Chkpt
MANAGER RUNNING EXTRACT RUNNING DPEYWGL 01:53:22 00:00:09 EXTRACT RUNNING SEXTR01 01:27:25 00:00:08
GGSCI (calladgdb) 12> info all
Program Status Group Lag at Chkpt Time Since Chkpt
MANAGER RUNNING EXTRACT RUNNING DPEYWGL 00:00:00 00:00:01 EXTRACT RUNNING SEXTR01 00:00:02 00:00:00 |
通过几次info all查看进程状态，看到lag at chkpt时间快速下降。问题解决。
本文作者：黎俊杰（网名：踩点），从事”系统架构、操作系统、存储设备、数据库、中间件、应用程序“六个层面系统性的性能优化工作
欢迎加入 系统性能优化专业群，共同探讨性能优化技术。群号：258187244郑重声明：本站内容如果来自互联网及其他传播媒体，其版权均属原媒体及文章作者所有。转载目的在于传递更多信息及用于网络分享，并不代表本站赞同其观点和对其真实性负责，也不构成任何其他建议。
