【Oracle性能优化】从10.2.0.5升级到11.2出现LOG FILE SYNCS等待事件显著增长的性能问题

如果 你遇到从10.2.0.5升级到11.2出现LOG FILE SYNCS等待事件显著增长的性能问题,那么有必要读一下这篇文章了。


在以往的经验中如果遇到这种场景 ,那么 优先考虑设置 “_use_adaptive_log_file_sync”=false, adaptive log file sync是 11.2中提出的一个优化重做日志写的新特性, 在11.2.0.3以后默认为TRUE。

有客户在将”_use_adaptive_log_file_sync”=false后,log file sync等待事件的平均等待时间从10ms 下降到 1~2ms的案例。


_use_adaptive_log_file_sync造成性能下降的原因可能是其导致LGWR使用了polling 方式来取代 post/wait,并且polling的间隔是10ms,这个间隔是在代码里写死的。


此外如果使用了Veritas/symantec 的ODM的话也需要特别注意:你可能遇到了Bug 13551402  High “log file parallel write” and “log file sync” after upgrading 11.2 with Veritas/Symantec ODM,这个BUG已经确认在11.2.0.3和11.2.0.2上存在。

对于该bug的内部讨论最后确认是由于 11.2中lgwr的 IO使用了一种批量同步I/O接口,导致当配合Veritas/symantec 的ODM一起使用时会导致性能下降。



high logfile sync after upgrade







  1. RELEASE NOTES: Increase in log file parallel write times when Veritas ODM is used. INTERNAL PROBLEM DESCRIPTION: In 11.2 lgwr I/O’s are issued using a batched synch I/O interface. This causes performance regressin in Veritas ODM implementation. INTERNAL FIX DESCRIPTION: Submit lgwr I/O’s asynchronously and immediately wait for them. Though this results in extra code path and an increase in number of system calls this showed improvement due to deficiencies in Veritas ODM stack. BACKPORT FEASIBLE: Yes FORWARD MERGE REQUIRED: No (merged to main branch) REDISCOVERY INFORMATION: Check if there is an increase in log file parallel write time after upgrading to 11.2 in Symantec env.

  2. vmcd says:

    good article

  3. vmcd says:

    good article

Speak Your Mind

沪公网安备 31010802001379号

TEL/電話+86 13764045638
Email service@parnassusdata.com
QQ 47079569