Date: Mon, 6 Jul 2009 23:01:10 +0300 From: Kostik Belousov <kostikbel@gmail.com> To: Dmitry Morozovsky <marck@rinet.ru> Cc: Ralf Folkerts <ralf.folkerts@gmx.de>, "Marat N.Afanasyev" <amarat@ksu.ru>, FreeBSD-STABLE Mailing List <freebsd-stable@freebsd.org> Subject: Re: bug in ufs? Message-ID: <20090706200110.GX2884@deviant.kiev.zoral.com.ua> In-Reply-To: <alpine.BSF.2.00.0907062351100.12926@woozle.rinet.ru> References: <4A50E947.9020608@ksu.ru> <4A523518.7050008@gmx.de> <4A523849.1070001@ksu.ru> <20090706193653.GU2884@deviant.kiev.zoral.com.ua> <alpine.BSF.2.00.0907062351100.12926@woozle.rinet.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
--7aBBD6ClYIPKzb+y Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Mon, Jul 06, 2009 at 11:54:48PM +0400, Dmitry Morozovsky wrote: > On Mon, 6 Jul 2009, Kostik Belousov wrote: >=20 > KB> > >>i have a strange problem with writing data to my ufs2+su filesyst= em. > KB> > >> > KB> > >>1. i made a 1T gpt partition on my storage server, and formatted = it: > KB> > >>newfs -U -m 0 -o time -i 32768 /dev/da1p3a > KB> > >> > KB> > >>2. i tried to move data from other servers to this filesystem, to= tal=20 > KB> > >>size of files is slightly less than 1T > KB> > >> > KB> > >>3. i encountered a 'No space left on device' while i still have 1= 1G of=20 > KB> > >>free space and about 13 million free inodes on the filesystem: > KB> > >> > KB> > >>#df -ih > KB> > >>Filesystem Size Used Avail Capacity iused ifree %ius= ed=20 > KB> > >>Mounted on > KB> > >>/dev/da1p3a 1.0T 1.0T 11G 99% 20397465 13363173 60= %=20 > KB> > >>/mnt/45_114 > KB> > >> > KB> > >>all i want to know is whether this is a bug or a feature? and if = such=20 > KB> > >>a behavior is well-known, where can i read about it? > KB> > >Hi Marat, > KB> > > > KB> > >just a guess: Are there sparse Files on the Source System that are= not=20 > KB> > >being detected by the Tool you used to restore the data? If you us= ed=20 > KB> > >(bsd)tar, did you try -S? > KB> > > > KB> > >A while ago I ran into a similar Problem when copying=20 > KB> > >Oracle-Database-Files (on Linux, though) and the -S - Option came = to=20 > KB> > >rescue. > KB> > > > KB> > >Cheers, > KB> > >_ralf_ > KB> >=20 > KB> > i have a huge amount of small files on the source systems, as you c= an=20 > KB> > see they have about 20 million files and almost each of them is jpe= g or=20 > KB> > gif. afaik, there are no sparse files at all. > KB> >=20 > KB> > i still cannot figure out what is it: a free space leak in ufs2+su = or=20 > KB> > bug in statfs(3), that is used in df, or something else. > KB>=20 > KB> My guess that it is due to fragmentation. > KB> As an experiment, try to create 1-byte file. Does it work on the file= system > KB> in described state ? >=20 > Doesn't UFS store one-byte (or several-bytes, like PID file) file entirel= y in=20 > the inode, not consuming any data blocks? No, 1-byte file is stored in fragment. Long write tries to allocate whole block. This is why I wrote about fragmentation. Answering your question, it is short symlinks that are stored in inode in the data block pointers. --7aBBD6ClYIPKzb+y Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (FreeBSD) iEYEARECAAYFAkpSWAYACgkQC3+MBN1Mb4gtNQCgsI5MyodW0usa9NS3sjb/EGbw q8wAoJDnyBexdckyU3u4g8HKUej+ShFz =K+wz -----END PGP SIGNATURE----- --7aBBD6ClYIPKzb+y--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20090706200110.GX2884>