什么是checkpoint？

在Oracle 数据库中checkpoint的意思是将内存中的脏数据强制写入到磁盘的事件,其作用是保持内存中的数据与磁盘上的数据一致。SCN是用来描述该事件发生的准确的时间点。

而GoldenGate中出现的checkpoint有着不同的含义。简单的说就是position，位置的意思。它记录了Extract进程在抽取事务时的进度。
使用INFO showch 命令可以查看当前正在运行的Extract进程中所记录的checkpoint。

当GoldenGate被计划内或者是计划外中断了，为了保证事务的连续性我们从哪里续传呢？答案是：哪里中断，哪里继续。在一个繁忙的数据库中用时间戳来衡量OGG抽取进度是不准确的，因为任何一个时间点都有可能截断了正在处理中的事务。准确的记录的方法是参考该事物在发起时所在trail文件的字节位置（RBA）。

startup checkpoint:

Extract进程启动时的信息，用多个参数描述了一个确切的时间点。包括thread,sequence,timestamp,RBA,SCN,redo log
sequence表示redo log、或者archive log的序列号。
RBA表示改时间点处于该sequence号日志文件中的字节地址。
SCN表示Oracle内部时钟

recovery checkpoint:

到目前为止还没结束的并且耗时最长的事务信息。包括thread,sequence,timestamp,RBA,SCN,redo log。
如果源端的抽取进程中断了，那么断点续传的位置就是这里了。

current checkpoint:

当前正在读取的事务信息。包括thread,sequence,timestamp,RBA,SCN,redo log

write checkpoint：
当前正在写入的trail文件信息。包括队列号（sequence）,RBA,时间戳，队列文件
sequence表示trail文件的序号，有别于数据库中的同名概念。例如：trail文件ex000001的sequence为1

RBA表示当前信息在trail文件中的位置，有别于数据库中事务日志的RBA。通常这个RBA的数值等于当前正在写的trail文件的大小。
可以通过ls -al ./dirdat/ex000001来查看该文件的字节数来对比write checkpoint的RBA


下面通过一个屏幕输出来对应一下刚才提到的概念。在实验开始时，还需要再次执行该命令来获取最新的检查点信息。

查看抽取进程ex1的检查点：
GGSCI (node3) 9> info ex1 showch

EXTRACT    EX1       Last Started 2014-11-03 13:23   Status RUNNING  << 进程开始抽取事务的时间，ADD EXTRACT之后就生效了
Checkpoint Lag       00:00:03 (updated 00:00:07 ago)
Log Read Checkpoint  Oracle Redo Logs
                     2014-11-03 15:13:04  Thread 1, Seqno 21, RBA 26012672  << 2个节点的RAC就有2个Checkpoint
                     SCN 0.1614343 (1614343)
Log Read Checkpoint  Oracle Redo Logs
                     2014-11-03 15:13:02  Thread 2, Seqno 29, RBA 6584832
                     SCN 0.1614345 (1614345)

Current Checkpoint Detail:

Read Checkpoint #1  << 抽取进程的读检查点 节点1

  Oracle Threaded Redo Log

  Startup Checkpoint (starting position in the data source):
    Thread #: 1
    Sequence #: 11
    RBA: 6084096
    Timestamp: 2014-11-02 17:41:26.000000
    SCN: Not available
    Redo File:

Recovery Checkpoint是我们续传时关注的位置  
Recovery Checkpoint (position of oldest unprocessed transaction in the data source):
    Thread #: 1
    Sequence #: 21  << 所在队列序号
    RBA: 26012176   << 所在队列中的具体位置
    Timestamp: 2014-11-03 15:13:04.000000  << 事务发起的时间戳
    SCN: 0.1614343 (1614343)
    Redo File: +DATA/prod/onlinelog/group_1.299.859509355

  Current Checkpoint (position of last record read in the data source):
    Thread #: 1
    Sequence #: 21
    RBA: 26012672
    Timestamp: 2014-11-03 15:13:04.000000
    SCN: 0.1614343 (1614343)
    Redo File: +DATA/prod/onlinelog/group_1.299.859509355

Read Checkpoint #2  << 抽取进程的读检查点 节点2

  Oracle Threaded Redo Log
     
  Startup Checkpoint (starting position in the data source):
    Thread #: 2
    Sequence #: 11
    RBA: 4263936
    Timestamp: 2014-11-02 17:41:26.000000
    SCN: Not available
    Redo File:

  Recovery Checkpoint (position of oldest unprocessed transaction in the data source):
    Thread #: 2
    Sequence #: 29
    RBA: 6584336
    Timestamp: 2014-11-03 15:13:02.000000
    SCN: 0.1614345 (1614345)
    Redo File: +DATA/prod/onlinelog/group_3.301.859509357

  Current Checkpoint (position of last record read in the data source):
    Thread #: 2
    Sequence #: 29
    RBA: 6584832
    Timestamp: 2014-11-03 15:13:02.000000
    SCN: 0.1614345 (1614345)
    Redo File: +DATA/prod/onlinelog/group_3.301.859509357

Write Checkpoint #1

  GGS Log Trail

  Current Checkpoint (current write position):
    Sequence #: 47
    RBA: 1124760
    Timestamp: 2014-11-03 15:13:07.093886
    Extract Trail: /goldengate/dirdat/ex

CSN state information:
  CRC: E4-D6-3B-B4
  CSN: Not available

Header:
  Version = 2
  Record Source = A
  Type = 11
  # Input Checkpoints = 2
  # Output Checkpoints = 1

File Information:
  Block Size = 2048
  Max Blocks = 100
  Record Length = 4096
  Current Offset = 0

Configuration:
  Data Source = 3
  Transaction Integrity = 1
  Task Type = 0

Status:
  Start Time = 2014-11-03 13:23:05
  Last Update Time = 2014-11-03 15:13:07
  Stop Status = A
  Last Result = 0


当前源端有100万条记录
SYS@PROD2 > select count(*) from snow.t1;

  COUNT(*)
----------
   1000000

系统时间为2014-11-03 15:37:08
SYS@PROD2 > alter session set nls_date_format=‘yyyy-mm-dd hh24:mi:ss‘;

Session altered.

SYS@PROD2 > select sysdate from dual;

SYSDATE
-------------------
2014-11-03 15:37:08

源端插入一条新数据，不提交
SYS@PROD2 > insert into snow.t1 values(1000001,‘Snow‘);

1 row created.

SYS@PROD2 >

关闭源端所有Extract进程
GGSCI (node3) 4> stop *

Sending STOP request to EXTRACT DP1 ...
Request processed.

Sending STOP request to EXTRACT EX1 ...
Request processed.

观察最新的checkpoint信息
GGSCI (node3) 14> info ex1 showch

EXTRACT    EX1       Last Started 2014-11-03 15:41   Status RUNNING
Checkpoint Lag       00:00:02 (updated 00:00:01 ago)
Log Read Checkpoint  Oracle Redo Logs
                     2014-11-03 15:41:48  Thread 1, Seqno 21, RBA 27887616
                     SCN 0.1620285 (1620285)
Log Read Checkpoint  Oracle Redo Logs
                     2014-11-03 15:41:46  Thread 2, Seqno 29, RBA 7889920
                     SCN 0.1620284 (1620284)

Current Checkpoint Detail:

Read Checkpoint #1

  Oracle Threaded Redo Log

  Startup Checkpoint (starting position in the data source):
    Thread #: 1
    Sequence #: 21
    RBA: 27644432
    Timestamp: 2014-11-03 15:39:18.000000
    SCN: Not available
    Redo File: +DATA/prod/onlinelog/group_3.301.859509357

  Recovery Checkpoint (position of oldest unprocessed transaction in the data source):
    Thread #: 1
    Sequence #: 21
    RBA: 27887120
    Timestamp: 2014-11-03 15:41:48.000000
    SCN: 0.1620285 (1620285)
    Redo File: +DATA/prod/onlinelog/group_1.299.859509355

  Current Checkpoint (position of last record read in the data source):
    Thread #: 1
    Sequence #: 21
    RBA: 27887616
    Timestamp: 2014-11-03 15:41:48.000000
    SCN: 0.1620285 (1620285)
    Redo File: +DATA/prod/onlinelog/group_1.299.859509355

Read Checkpoint #2

  Oracle Threaded Redo Log

  Startup Checkpoint (starting position in the data source):
    Thread #: 2
    Sequence #: 29
    RBA: 7750160
    Timestamp: 2014-11-03 15:39:16.000000
    SCN: Not available
    Redo File: +DATA/prod/onlinelog/group_3.301.859509357

  Recovery Checkpoint (position of oldest unprocessed transaction in the data source):
    Thread #: 2
    Sequence #: 29
    RBA: 7750160
    Timestamp: 2014-11-03 15:37:40.000000
    SCN: 0.1619666 (1619666)
    Redo File: +DATA/prod/onlinelog/group_3.301.859509357

  Current Checkpoint (position of last record read in the data source):
    Thread #: 2
    Sequence #: 29
    RBA: 7889920
    Timestamp: 2014-11-03 15:41:46.000000
    SCN: 0.1620284 (1620284)
    Redo File: +DATA/prod/onlinelog/group_3.301.859509357

Write Checkpoint #1

  GGS Log Trail

  Current Checkpoint (current write position):
    Sequence #: 48
    RBA: 1076
    Timestamp: 2014-11-03 15:41:50.718150
    Extract Trail: /goldengate/dirdat/ex

CSN state information:
  CRC: CE-DB-CE-78
  CSN: Not available

Header:
  Version = 2
  Record Source = A
  Type = 11
  # Input Checkpoints = 2
  # Output Checkpoints = 1

File Information:
  Block Size = 2048
  Max Blocks = 100
  Record Length = 4096
  Current Offset = 0

Configuration:
  Data Source = 3
  Transaction Integrity = 1
  Task Type = 0

Status:
  Start Time = 2014-11-03 15:41:29
  Last Update Time = 2014-11-03 15:41:50
  Stop Status = A
  Last Result = 0


修改抽取进程启动位置，两个进程
GGSCI (node3) 7> ALTER EXTRACT ex1, EXTSEQNO 21, EXTRBA 27644432 THREAD 1
EXTRACT altered.

GGSCI (node3) 8> ALTER EXTRACT ex1, EXTSEQNO 29, EXTRBA 7750160 THREAD 2
EXTRACT altered.


重启进程后源端发起提交命名，结束该事务。
SYS@PROD2 > commit;

Commit complete.

第1000001条数据插入完成
SYS@PROD2 > select count(*) from snow.t1;

  COUNT(*)
----------
   1000001

目标端第1000001条数据复制完成！ 
SQL> select count(*) from snow.t1;

  COUNT(*)
----------
   1000001

实验结束

http://blog.itpub.net/29047826/viewspace-1328761/

关于Oracle GoldenGate中Extract的checkpoint的理解 转载