What we've figured out so far is that "duplicate" files occur during
create and write operation of new file. Chances that file will be
"duplicated" are max 1 file per milion. The problem is that write fails
(but content is written properly)
And as I wrote in previous mail find is showing same as ls.
Gorazd
On Wed, 2006-01-18 at 11:22 +0530, Shailendra Tripathi wrote:
> Hi,
> Can you try some other command instead of 'ls'. I suspect that the 'ls
> -lia' filter is doing something strange. I wish to rule out this
> possibility.
>
> Can you just do the following as well, meanwhile.
>
> find <path> -name 'f4407c24c1f31d2304c8f8d4825*'
>
> Regards,
> Shailendra
>
> Gorazd Golob wrote:
> >>Can you send xfs_info output?
> >>
> >
> > This one is on 2.6.11 - x86 - partiton is on 72gb scsi drive:
> > meta-data=/xxxxx isize=256 agcount=16, agsize=4480124 blks
> > = sectsz=512
> > data = bsize=1024 blocks=71681984,
> > imaxpct=25
> > = sunit=0 swidth=0 blks, unwritten=1
> > naming =version 2 bsize=4096
> > log =internal bsize=1024 blocks=35000, version=1
> > = sectsz=512 sunit=0 blks
> > realtime =none extsz=65536 blocks=0, rtextents=0
> >
> > with xfsprogs-2.6.13 compiled for i486
> >
> > And 2.6.12-6 - x86_64 - partition is on MD raid1 - but don't have
> > xfs_info on system - I'll have it tommorow - I guess ;) I have installed
> > xfsprogs-2.6.36 compiled for Intel's cpus with EM64T (nocona).
> >
> > Hardware for both systems is not the same.
> >
> >
> >>>- some (very rarerly but they does) files appear like this on file
> >>>system:
> >>>
> >>>$ ls -ila f4407c24c1f31d2304c8f8d4825*
> >>>237047122 -rw-r--r-- 1 xxx yyy 6107 2006-01-16 23:33
> >>>f4407c24c1f31d2304c8f8d48258.dat.gz
> >>>237047122 -rw-r--r-- 1 xxx yyy 6107 2006-01-16 23:33
> >>>f4407c24c1f31d2304c8f8d48258.dat.gz
> >>>
> >>>Same problem appear on kernels 2.6.11 and 2.6.12.6 (x86 and x86_x64).
> >>>Have anyone already seen something like this?
> >>
> >>Nope. Any ideas on how to reproduce it? Is NFS being used?
> >>
> >
> > Yep - we are reproducing every day - but not intentialy. No NFS used -
> > just writting and reading files from partition (with o_direct and
> > without it). Maybe is just that we have about 5 milion files on those
> > partitions.
> >
> >
> >
> >>cheers.
> >>
> >
> >
> >
|