Date: Wed, 29 Jul 2009 11:52:11 +0200 From: Arnaud Houdelette <arnaud.houdelette@tzim.net> To: Mark Stapper <stark@mapper.nl> Cc: FreeBSD <freebsd-stable@freebsd.org> Subject: Re: Cannot rm files when ZFS is full Message-ID: <4A701BCB.4030805@tzim.net> In-Reply-To: <4A7008F2.6090107@mapper.nl> References: <d2e731a10907282029rd17ca1ag892c3c9615e76140@mail.gmail.com> <14989d6e0907290045j94bab06wb5d0114fe2af2331@mail.gmail.com> <4A7008F2.6090107@mapper.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
Mark Stapper a écrit : > Christian Walther wrote: > >> 2009/7/29 grarpamp <grarpamp@gmail.com>: >> >> >>> One week old build... >>> >>> # df -i . >>> Filesystem 1K-blocks Used Avail Capacity iused ifree %iused Mounted on >>> ram01/mnt1 239465344 239465344 0 100% 13163 0 100% /mnt1 >>> # ls -aliT zero >>> 20797 -rw-r--r-- 1 user user 43515904 Jul 28 23:20:57 2009 zero >>> # rm -f zero >>> rm: zero: No space left on device >>> # :> zero >>> cannot create zero: File exists >>> # cp /dev/null zero >>> overwrite zero? (y/n [n]) y >>> # ls -aliT zero >>> 20797 -rw-rw-rw- 1 root wheel 0 Jul 28 23:25:17 2009 zero >>> # rm -f zero >>> [gone] >>> >>> >> this is a known problem with the current version of ZFS. Due to the >> way ZFS handles access to the data it stores, even a rm causes a >> write, which requires some additional disk space in the beginning: >> Instead of simply unlinking what should be removed ZFS creates another >> tree without the removed data. Only if this new tree has been entirely >> written to disk the old information is removed. This is a rather rough >> explanation and probably not entirely correct, but I hope it suffices. >> Only hope: Make sure that not all disk space is used. >> >> Christian >> >> > Indeed, if by some coincident (like a growing logfile) every single byte > is used, even the copy action might fail... > To prevent this you could set the maximum size of all partitions in your > pool so that the sum of them is smaller then the size of your pool. > Just a thought. > Greetz, > Mark > > Wouldn't that be counterproductive ? One avantage of ZFS is that you don't need to manage space, and that free space is shared. Setting size as this would reintroduce the hassle of free space management. Wouldn't the simplest be to set a reservation on an unused (or known as non-growing) filesystem, and/or quotas to filesystem susceptible of growing out of control ? Arnaud
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4A701BCB.4030805>