RE: Latency of "direct path reads"

From: Hameed, Amir <Amir.Hameed_at_xerox.com>
Date: Tue, 20 Aug 2013 02:14:17 +0000
Message-ID: <AF02C941134B1A4AB5F61A726D08DCED0DD6E6DB_at_USA7109MB012.na.xerox.net>



Hi Austin,
I really appreciate your feedback. I did go through the document that you have mentioned below a few times in the past but most of the patches listed at that time were either released for Linux or Exadata for Solaris; we are SPARC Solaris. After we provided pstack dumps to Oracle today, Oracle confirmed that we are hitting the same bug that you have mentioned below. But since we are at 11.2.0.3.6 and the patch for Solaris is available for 11.2.0.3.5, Oracle is in the process of back-porting the patch.

You are also right about the 32k chunk size because of the rsize and wsize mount options and I confirmed it by running truss. How has your experience been with dNFS so far?

Thanks
-----Original Message-----

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Austin Hackett Sent: Monday, August 19, 2013 5:39 PM
To: oracle-l digest users
Subject: Re: Latency of "direct path reads"

Hi Amir

I've recently deployed a 5 node 11.2.0.3.2 RAC with dNFS and a NetApp FAS6290 into production.

Are you familiar with MOS Doc ID 1495104.1, which has a list of Recommended Patches for Direct NFS? If you don't have these patches applied, I would definitely suggest getting them installed. For example, during my initial lab tests I kept hitting bug 15987992 under heavy sequential read load. I/O would completely hang for a seemingly indefinite period. Admittedly, this doesn't sound like what you're seeing. However, there are other bugs mentioned which sound like they could be of interest. You may need to request back ports of the patches for your exact PSU-level. I did.

Apologies if this is old news.

Regarding your question about sending > 32K chunks of data, the Oracle recommended value for the rsize mount option is 32K, so I think a 1MB multi block read would be split into 32K chunks under KNFS anyway. I don't know if DNFS uses a larger transfer buffer size?

Austin

--

http://www.freelists.org/webpage/oracle-l

--

http://www.freelists.org/webpage/oracle-l Received on Tue Aug 20 2013 - 04:14:17 CEST

Original text of this message