[Oracle工程师手记] 记一次 transport lag 的解析

返回：Oracle 索引页

客户联络说他观察到，每天的某个固定时刻，OEM会报告比较突出的 transport lag ，平时的 transport lag，几乎是没有的。

观察这个时间段的主库的 alert log，可以看出些问题：

比如，在问题尚未发生的 4/5 05:45 左右，开始生成 sequence# = 21595 的 archive log：

Web Apr 05 05:45:15 2021
Beginning log switch checkpoint up to RBA [0x86bf.2.10], SCN: 215805932158
Thread 1 advanced to log sequence 21595 (LGWR switch) 　***
  Current log# 4 seq# 21595 mem# 0: +DATA1/orclp/onlinelog/group_3.3725.845417343
......

 而下一个 sequence 的 archive log，是什么时候开始生成的呢？

Web Apr 05 10:02:07 2021
Beginning log switch checkpoint up to RBA [0x76c1.2.10], SCN: 215811522963
Thread 1 advanced to log sequence 21596 (LGWR switch) ***
  Current log# 1 seq# 21596 mem# 0: +DATA1/orclp/onlinelog/group_4.3703.845417337

......
Web Apr 05 10:02:30 2021
LNS: Completed archiving log 4 thread 1 sequence 21595 (orclp_1)
LNS: Closing remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (orclp_1)
LNS: Creating remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (thread 1 sequence 21596) (orclp_1)　***
LNS: Standby redo logfile selected for thread 1 sequence 21596 for destination LOG_ARCHIVE_DEST_2
LNS: Standby redo logfile selected for thread 1 sequence 21596 for destination LOG_ARCHIVE_DEST_2
LNS: Beginning to archive log 1 thread 1 sequence 21596 (orclp_1)
......

 已经是几个小时之后了。这正是客户反映开始出现 transport lag 的时刻。这之后，生成新的 sequence# 的 archive log 比较频繁了。

Sequence#= 21597 的 REDO ：

Web Apr 05 10:05:22 2021
Beginning log switch checkpoint up to RBA [0x76c3.2.10], SCN: 215811692247
Thread 1 advanced to log sequence 21597 (LGWR switch) ***
  Current log# 2 seq# 21597 mem# 0: +DATA1/orclp/onlinelog/group_1.3709.845417339
......
Web Apr 05 10:10:40 2021
LNS: Completed archiving log 1 thread 1 sequence 21596 (orclp_1)
LNS: Closing remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (orclp_1)　***
LNS: Creating remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (thread 1 sequence 21597) (orclp_1)　***
LNS: Standby redo logfile selected for thread 1 sequence 21597 for destination LOG_ARCHIVE_DEST_2
LNS: Standby redo logfile selected for thread 1 sequence 21597 for destination LOG_ARCHIVE_DEST_2
LNS: Beginning to archive log 2 thread 1 sequence 21597 (orclp_1)

 

Sequence#=21598 的 REDO：

Web Apr 05 10:12:34 2021
Beginning log switch checkpoint up to RBA [0x87d3.2.10], SCN: 215811845211
Thread 1 advanced to log sequence 21598 (LGWR switch) ***
  Current log# 3 seq# 21598 mem# 0: +DATA1/orclp/onlinelog/group_3.3713.845417341
......
Web Apr 05 10:16:17 2021
Incremental checkpoint up to RBA [0x76c3.2940d5.0], current log tail at RBA [0x87d3.2654a5.0]
Web Apr 05 10:16:20 2021
LNS: Completed archiving log 2 thread 1 sequence 21597 (orclp_1)
LNS: Closing remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (orclp_1)
LNS: Creating remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (thread 1 sequence 21598) (orclp_1)　***
LNS: Standby redo logfile selected for thread 1 sequence 21598 for destination LOG_ARCHIVE_DEST_2
LNS: Standby redo logfile selected for thread 1 sequence 21598 for destination LOG_ARCHIVE_DEST_2
LNS: Beginning to archive log 3 thread 1 sequence 21598 (orclp_1)

 

Sequence#=21599 的 REDO：

Web Apr 05 10:17:46 2021
Beginning log switch checkpoint up to RBA [0x95d2.2.10], SCN: 215812104538
Thread 1 advanced to log sequence 21599 (LGWR switch) ***
  Current log# 4 seq# 21599 mem# 0: +DATA1/orclp/onlinelog/group_3.3725.845417343
......
Web Apr 05 10:22:44 2021
Completed checkpoint up to RBA [0x95d2.2.10], SCN: 215812104538
Web Apr 05 10:22:53 2021
LNS: Completed archiving log 3 thread 1 sequence 21598 (orclp_1)
LNS: Closing remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (orclp_1)
LNS: Creating remote archive destination LOG_ARCHIVE_DEST_2: 'DG_STY' (thread 1 sequence 21599) (orclp_1)　***
LNS: Standby redo logfile selected for thread 1 sequence 21599 for destination LOG_ARCHIVE_DEST_2
LNS: Standby redo logfile selected for thread 1 sequence 21599 for destination LOG_ARCHIVE_DEST_2
LNS: Beginning to archive log 4 thread 1 sequence 21599 (orclp_1)

 

在主库端快速生成大量REDO 的时候，备库端需要开启RFS进程，利用 Standby REDO 来接收这些 REDO数据，一时之间接收的速度赶不上产生REDO的数据，就会出现 transport lag。

返回：Oracle 索引页