From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mark Nelson Subject: Re: ceph from poc to production Date: Thu, 17 Jan 2013 10:29:57 -0600 Message-ID: <50F82705.50502@inktank.com> References: <50F6F471.7030108@profihost.ag> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from mail-ia0-f169.google.com ([209.85.210.169]:64589 "EHLO mail-ia0-f169.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757052Ab3AQQ3z (ORCPT ); Thu, 17 Jan 2013 11:29:55 -0500 Received: by mail-ia0-f169.google.com with SMTP id j5so571054iaf.0 for ; Thu, 17 Jan 2013 08:29:54 -0800 (PST) In-Reply-To: Sender: ceph-devel-owner@vger.kernel.org List-ID: To: =?windows-1252?Q?S=E9bastien_Han?= Cc: Stefan Priebe , "ceph-devel@vger.kernel.org" On 01/17/2013 10:24 AM, S=E9bastien Han wrote: > Hi Stephan, > >> - Increase the osdmax value > > Well actually this doesn't seem to be necessary because when you issu= e > a "ceph osd create" the osdmax value is automatically incremented. It > just that a while ago I saw on the ML or IRC that it was a good > practice to do this. And this documentation > http://ceph.com/docs/master/rados/operations/control/ says: > > "Set the max_osd parameter in the osd map. This is necessary when > expanding the storage cluster." > > Maybe the doc is outdated on this option, I don't know. > >> - Set =93filestore flusher=94 option to false, in your [osd] section > > During my benchmarks sessions I noticed that disabling this option > brought way better performance, at least for old hardware. Fo recent > hardware I haven't seen any difference, but I kept the habit to > disable it. There are some discussions about this option on the ML. A= s > far as I can remember, it's harmless to disable it. > From the doc here > http://ceph.com/docs/master/rados/configuration/filestore-config-ref/= : I actually have been doing parametric sweeps over the various ceph=20 tunables to try to see what affects performance. I'll have an article=20 coming out soonish documenting some of the findings. Typically=20 disabling the flusher seems to be beneficial on XFS and BTRFS, but it=20 seems to be the opposite on EXT4. > > "The filestore flusher forces data from large writes to be written ou= t > using sync file range before the sync in order to (hopefully) reduce > the cost of the eventual sync. In practice, disabling =91filestore > flusher=92 seems to improve performance in some cases." > > Cheers, > > -- > Regards, > S=E9bastien Han. > > > On Wed, Jan 16, 2013 at 7:41 PM, Stefan Priebe wrote: >> >> Hello Sebastien, >> hello list, >> >> first nice article sebastien ;-) thanks for contributing. >> >> I've two question regarding the tips: >> >> Can somebody explain me why i should do this?: >> - Increase the osdmax value >> - Set =93filestore flusher=94 option to false, in your [osd] section >> >> Greets, >> Stefan >> -- >> To unsubscribe from this list: send the line "unsubscribe ceph-devel= " in >> the body of a message to majordomo@vger.kernel.org >> More majordomo info at http://vger.kernel.org/majordomo-info.html > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel"= in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html