Date: Tue, 5 Mar 2013 01:25:48 -0800 (PST) From: Don Lewis <truckman@FreeBSD.org> To: phk@phk.freebsd.dk Cc: kostikbel@gmail.com, deeptech71@gmail.com, freebsd-current@FreeBSD.org, peter@rulingia.com, ian@FreeBSD.org Subject: Re: access to hard drives is "blocked" by writes to a flash drive Message-ID: <201303050925.r259Pm7o012790@gw.catspoiler.org> In-Reply-To: <65342.1362473108@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On 5 Mar, Poul-Henning Kamp wrote: > In message <201303050840.r258egAG012697@gw.catspoiler.org>, Don Lewis writes: > >>For composite devices such as mirrors, using the first underlying device >>is probably a reasonable choice. For more complicated cases, or to >>override the default, the syncer thread could be specified as a mount >>option. > > I doubt that will be any better than what we have today. I'm thinking of the case where ada0s1a and ada1s1a are mirrored, ada0s2b and ada1s2b are mirrored, etc. Or ada0 and ada1 are mirrored and then the mirror is partitioned. We don't want two syncer threads simultaneously trying to fsync files on the outer tracks and the inner tracks. If we just create a syncer thread for ada0, it will fsync the files on the outer tracks first and then go on to the inner tracks, avoiding lots of long seeks. > I think it is a much better idea to have the syncer monitor bio write > latency and adjust accordingly. That's a good idea as well, but that's also something that is global to each "device".
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201303050925.r259Pm7o012790>