Oracle 日志文件Warning: log write elapsed time 664ms, size 1KB

Warning: log write elapsed time 664ms, size 1KB

突然才发现一个数据库的LGWR进程的跟踪文件scm2_lgwr_5690.trc有大量的告警信息，如下所示：

*** 2017-08-07 17:05:03.132
Warning: log write elapsed time 664ms, size 1KB
kcrfw_update_adaptive_sync_mode: post->poll long#=1 sync#=6 sync=302 poll=1079 rw=322 ack=0 min_sleep=1079

*** 2017-08-07 17:42:56.997
Log file sync switching to polling
Current scheduling delay is 1 usec
Current approximate redo synch write rate is 2 per sec

*** 2017-08-08 09:00:03.923
Warning: log write elapsed time 1445ms, size 1KB

*** 2017-08-08 10:00:20.162
Warning: log write elapsed time 1190ms, size 87KB

*** 2017-08-08 17:05:02.924
Warning: log write elapsed time 1213ms, size 1KB

*** 2017-08-09 08:36:15.179
Warning: log write elapsed time 3171ms, size 1KB
kcrfw_update_adaptive_sync_mode: poll->post current_sched_delay=0 switch_sched_delay=1 current_sync_count_delta=0 switch_sync_count_delta=6

*** 2017-08-09 18:39:36.593
Log file sync switching to post/wait
Current approximate redo synch write rate is 0 per sec

搜索了Metalink上关于该告警的阐述Metalink ID 601316.1,具体内容如下所示（英文部分）：

CHANGES

The problem surfaced after upgrading to 10.2.0.4.

CAUSE

The above warning messages has been introduced in 10.2.0.4 patchset.

The warning message will be generated only if the log write time is more than 500 ms and it will be written to the lgwr trace file.

SOLUTION

These messages are expected in a 10.2.0.4 database in case the log write is more than 500 ms.

This is a warning which means that the write process is not as fast as it intented to be.

So, probably you need to check if the disk is slow or not or for any potential OS causes.

If everything looks fine at the hardware level or OS level i.e if enviroment is unable to deliver a faster service because of its own nature,then you can ignore this message.then you can safely ignore these messages. The trace file can easily be deleted or truncated.

Also, according to Bug:7559549 , these trace can be disabled by setting event 10468 level 4.

刚好我们数据库版本也是10.2.0.4.0，于是我检查了另外其它几台数据库，发现基本上都有这类错误，根据官方文档，如果磁盘没有什么问题，可以忽略这个告警信息。

发现了对这个告警现象的记录、描述.应该可以忽略这个告警。另外，也可以通过下面命令取消该trace命令，如下所示：

ALTER SYSTEM SET EVENTS '10468 trace name context off';


另外要注意oracle数据库不要放在RAID5上面，效率低。