Date: Fri, 10 Aug 2018 17:54:03 +0300 From: Paul <devgs@ukr.net> To: Konstantin Belousov <kostikbel@gmail.com> Cc: freebsd-fs@freebsd.org Subject: Re[2]: Question regarding relevance of syncer(4) in the context of ZFS Message-ID: <1533912779.505892598.lhqtiahk@frv33.fwdcdn.com> In-Reply-To: <20180810143940.GC2649@kib.kiev.ua> References: <1533910747.525373107.k9z2n7hj@frv33.fwdcdn.com> <20180810143940.GC2649@kib.kiev.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
10 August 2018, 17:39:52, by "Konstantin Belousov" <kostikbel@gmail.com>: > On Fri, Aug 10, 2018 at 05:28:11PM +0300, Paul wrote: > > Hello team, > > > > > > If my understanding is correct then ZFS does not need to be pushed around and being told when to sync data to devices. > > It is perfectly capable of keeping data consistent and synchronized according to configured options. > > > > We even disable 'sync' option of file system that we use. But unfortunately we see a constant and periodical spikes > > of load on our servers that are directly related to wake-ups of 'syncer' kernel daemon. Is it safe to assume that > > 'syncer' is not necessary on configurations that only use ZFS, and no other file systems? And hence, is it safe > > to assume that setting all of 'kern.filedelay', 'kern.dirdelay' and 'kern.metadelay' to some large values will not > > compromise system's durability? > > One of the job of the syncer is to convert dirty user-mapped pages into > the filesystem-specific write requests. Without syncer touching the mount > point, corresponding pages could linger forever, or at least until a memory > pressure causes pagedaemon to clean them. Thanks a lot for clarification! I didn't think about dirty user-mapped pages... Best regards, Paul
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1533912779.505892598.lhqtiahk>