----- Original Message ----- > From: "Phil Pennock" <[email protected]> > To: [email protected] > Sent: Wednesday, August 17, 2011 11:48:25 PM > Subject: Re: [lopsa-tech] Poor NFS performace from CentOS5 client to > Oracle (Sun) Unified Storage NAS
> On 2011-08-17 at 17:14 -0500, Patrick M. Landry wrote: > > The main difference with the poorly performing share is that there > > are over 7000 subdirectories in the directory which is heavily > > referenced > > by the application. Also, about 2-3 months ago is a fair estimate > > of when the number of subdirectories increased substantially. > Is this a constant set of sub-directories, or are they created and > destroyed? Is that being done from another host? The NFS share is being accessed from several hosts, any of which can create directories. But once the directories are created they stay around for a long time. And they are not created very often (a few a day). > Every time the list of entries in a directory is changed, the > directory > is invalidated on the clients and they need to re-fetch the entire > directory. There's no concept of delta notification in NFS v2 and v3. > NFSv4 fixes this, both with CB_NOTIFY notifications of added and > removed > entries, and with directory delegation to clients. > If only one host at a time is accessing the share, it shouldn't make > a > difference, but if some worker or admin tool on another host is doing > this, then it's an avenue to investigate. Thanks for that. Probably another reason for me to investigate NFSv4. -- patrick
_______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
