Date: Fri, 8 Jul 2011 04:08:29 -0400 From: Jason Hellenthal <jhell@DataIX.net> To: Volodymyr Kostyrko <c.kworr@gmail.com> Cc: freebsd-fs@freebsd.org Subject: Re: everlasting log device Message-ID: <20110708080828.GA49288@DataIX.net> In-Reply-To: <4E169107.6040109@gmail.com> References: <iv581p$8bv$1@dough.gmane.org> <20110707220019.GA79464@DataIX.net> <4E169107.6040109@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Jul 08, 2011 at 08:09:27AM +0300, Volodymyr Kostyrko wrote: > 08.07.2011 01:00, Jason Hellenthal wrote: >=20 > >> System claims that this log device has 148K data. Is this the size of > >> unwritten data? The number is still the same when booting into single > >> user mode and doesn't change at all. > >> > >> Can I remove this log device? Should I recreate the pool to get rid of > >> this behavior? > >> > > > > If you have the possibility to re-create the pool then Id definately > > suggest it. > > > > If you remove this device (physically) your pool will not be operable > > unfortunately there is still somehting missing to allow SLOGs to be > > removed from a running pool yet, what that might be is beyond me at this > > time. You might try to export the pool then boot into single user mode > > and reimport the pool and try the removal procedure but I raelly dont > > think that will help you. >=20 > Well I have already removed device physically and the pool continues to= =20 > work despite yelling about missing log device. I already tried booting=20 > from flash drive and importing/exporting pool - this doesn't work. I=20 > also tried to replace this log device with another one. Replacing works= =20 > and I can remove old device but the new one will now be unkillable. >=20 > Am I right that working without external log device is almost the same=20 > as working with disabled ZIL? >=20 What was the reason or solution for which you added the slog in the first place that you were trying to solve. I would assume it would be for speed but if that's the case and your workloads are not fsync() high amounts of data then there is absolutely no point other than debugging to make FreeBSD better! ;) I don't know what size your device was that you allotted to the slog but hope you didn't go far beyond 256MB yes that's meant to say 'MB'. There is some long discussions about slog/ZIL devices in earlier threads on these lists so you'll have to refer to them to understand what I am talking about here. Cache devices can be removed any time, and depending on the case can improve or decrease performance. ZFS is a large beast to tackle with a lot of different ways you can configure a system and a slew of options at accomplishing that. So in the least words there was a lot of documentation that was put out about it that can aid in determining how YOU should configure your pool and give you somewhat of a description of what things do and how they interact with other system processes. Besides that there is a slew of other information by some well known people on the freebsd threads that will also come in handy, check out the following. Google "site:lists.freebsd.org zfs, slog, zil, cache" Other than that it seems one layer of the system is not talking well to ZFS when its told to do something and it would be really swell if this could be found and exterminated with extreme prejudice.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20110708080828.GA49288>