Date: Fri, 10 Jan 2014 07:47:06 +0100 From: Attila Nagy <bra@fsn.hu> To: Adrian Chadd <adrian@freebsd.org>, Alexander Motin <mav@freebsd.org> Cc: "svn-src-head@freebsd.org" <svn-src-head@freebsd.org>, svn-src-all@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r260486 - head/etc/defaults Message-ID: <52CF976A.3070501@fsn.hu> In-Reply-To: <CAJ-Vmon7=Dtm--PN_YrXRY3hBGbvYP3reQAENn0OPVXgne26vg@mail.gmail.com> References: <201401091555.s09Fttju004938@svn.freebsd.org> <CAJ-VmokAfJVwHqdV=GojLj2EeDgU-njgVedWUBM6-Zv1mDgffw@mail.gmail.com> <52CEC79F.2090708@FreeBSD.org> <CAJ-Vmo=exgGkzq6sFrgBzgZB0TjNJRvZ3eo6=Fs%2BU1Rp354pvg@mail.gmail.com> <52CECE57.4040002@FreeBSD.org> <CAJ-Vmon7=Dtm--PN_YrXRY3hBGbvYP3reQAENn0OPVXgne26vg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
We have 2 and 4 hw.ncpu NFS servers with 70+ disks, so there may be cases, where even the default maximum of 256 threads is not enough to feed the -otherwise slow- disks. I guess the real solution here is to change the nfsd worker model to async. (adapting to top poster) On 01/09/14 18:14, Adrian Chadd wrote: > If it's one request per nfsd then we should likely scale it separate to > ncpu. Otherwise the default for two or core boxes may not be enough. > > Adrian > On Jan 9, 2014 11:29 AM, "Alexander Motin" <mav@freebsd.org> wrote: > >> On 09.01.2014 18:18, Adrian Chadd wrote: >> >>> Depends if you're thinking locally or globally. >>> >>> Locally - for nfs? not a big deal. >>> >>> Globally - NFS, ZFS, GELI, geom/cam, NIC, etc.. suddenly your machine >>> could default to having a couple thousand worker threads just for a >>> HBA and a 10GE NIC. That's a little nuts. >>> >> So, what is your point? Each NFS thread (unlike GEOM or CAM) executes only >> _one_ request at a time. Would you like your 128-core/many-spindle system >> executed only 4 synchronous requests at a time? >> >> On 9 January 2014 08:00, Alexander Motin <mav@freebsd.org> wrote: >>>> On 09.01.2014 17:57, Adrian Chadd wrote: >>>> >>>>> .. so with say, 128 core boxes showing up, is this really a good >>>>> default? >>>>> >>>> >>>> And what is the price? 16K+ of KVA per thread for thread stack, etc? 4 >>>> threads is probably much worse default there. May be nfsd's default >>>> could be >>>> tuned, but obviously it should not be hardcoded value. >>>> >>>> >>>> On 9 January 2014 07:55, Alexander Motin <mav@freebsd.org> wrote: >>>>>> Author: mav >>>>>> Date: Thu Jan 9 15:55:55 2014 >>>>>> New Revision: 260486 >>>>>> URL: http://svnweb.freebsd.org/changeset/base/260486 >>>>>> >>>>>> Log: >>>>>> Remove very low default limit of 4 nfsd threads. nfsd's own >>>>>> default >>>>>> is >>>>>> 8 * hw.ncpu, that sounds more appropriate for these SMP/NCQ/... >>>>>> days. >>>>>> >>>>>> Modified: >>>>>> head/etc/defaults/rc.conf >>>>>> >>>>>> Modified: head/etc/defaults/rc.conf >>>>>> >>>>>> ============================================================ >>>>>> ================== >>>>>> --- head/etc/defaults/rc.conf Thu Jan 9 15:38:28 2014 >>>>>> (r260485) >>>>>> +++ head/etc/defaults/rc.conf Thu Jan 9 15:55:55 2014 >>>>>> (r260486) >>>>>> @@ -311,7 +311,7 @@ nfs_client_enable="NO" # This host is >>>>>> a >>>>>> nfs_access_cache="60" # Client cache timeout in seconds >>>>>> nfs_server_enable="NO" # This host is an NFS server (or NO). >>>>>> oldnfs_server_enable="NO" # Run the old NFS server (YES/NO). >>>>>> -nfs_server_flags="-u -t -n 4" # Flags to nfsd (if enabled). >>>>>> +nfs_server_flags="-u -t" # Flags to nfsd (if enabled). >>>>>> mountd_enable="NO" # Run mountd (or NO). >>>>>> mountd_flags="-r" # Flags to mountd (if NFS server >>>>>> enabled). >>>>>> weak_mountd_authentication="NO" # Allow non-root mount >>>>>> requests >>>>>> to be served. >>>>>> >>>> >>>> -- >>>> Alexander Motin >>>> >> -- >> Alexander Motin >> > _______________________________________________ > svn-src-all@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/svn-src-all > To unsubscribe, send any mail to "svn-src-all-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?52CF976A.3070501>