截屏视频帮助

Optimal mount options for VxFs (603) For Oracle DB's

创建时间: 21 11 月 2013 • Updated: 30 12 月 2013 | 2 条评论
此问题已解决。 请查看解决方案。

I know the recommended practice is to use ODM, but the presence of bug 13551402 is preventing us from using it as this time. This bug causes an increase in the DB commit elasped time. The fix is to upgrade to 11204.

We know we can mount the vxfs filesystem with mincache=direct/convsync=direct to enable direct IO. But how do we get async IO. My understanding is that enabling DIO will disable AIO.

Any feedback will be appreciated.

操作系统:
讨论 归类至以下社区:

评论 条评论跳转至最新评论

SpikeB 的图片

Hi,

  The presence of mincache=direct and for that matter, convosync=direct wil not disable asynchronous IO.  So if you put these mount options in, AIO will still work.  The caveat is that the OS must support AIO as well.  In reference to bugid 13551402, is that a Oracle bug as I  cannot find it in our tracking system.

Best Regards,

Spike

cnewtonne 的图片

Thanks. Is there an official document you direct me to that states this clearly. I reviwed the VSF UG's, but found none.

As for the bug, yes it is an Oracle bug.

Bug 13551402 - High "log file parallel write" and "log file sync" after upgrading 11.2 with Veritas/Symantec ODM (Doc ID 13551402.8).

解决方案