[Gluster-users] Hundreds of duplicate files

Vijay Bellur vbellur at redhat.com
Sat Dec 27 17:12:52 UTC 2014


On 12/27/2014 01:11 PM, tbenzvi at 3vgeomatics.com wrote:
> Thanks for your continued help Joe.
> A demonstration of the problem, in this case I was able to open the file
> in vim (a text file) without any issues, however sometimes duplicated
> text files open in vim as one line consisting of @ characters, and
> binary data files can also not be opened correctly for reading.
> However the duplicate listing is still an issue. Note that Dec 13 was
> the date of a server crash.
>
> [root at jongoo ~]# ll /sar/complete/vancouver/refdem/tif2flt.pro*
> -rw-rw-r-T 1 parwant users 1712 Dec 13 19:02 tif2flt.pro
> -rw-rw-r-- 1 parwant users 1712 Jun 17 2010 tif2flt.pro
>
> A few minutes later doing the same listing.. sticky bit disappeared and
> modification date changed
>
> [root at jongoo ~]# ll /sar/complete/vancouver/refdem/tif2flt.pro*
> -rw-rw-r-- 1 parwant users 1712 Jun 17 2010
> /sar/complete/vancouver/refdem/tif2flt.pro
> -rw-rw-r-- 1 parwant users 1712 Jun 17 2010
> /sar/complete/vancouver/refdem/tif2flt.pro
>
> [root at jongoo ~]# getfattr -m . -d -e hex
> /data/glusterfs/safari/brick00/brick/complete/vancouver/refdem/tif2flt.pro
> getfattr: Removing leading '/' from absolute path names
> # file:
> data/glusterfs/safari/brick00/brick/complete/vancouver/refdem/tif2flt.pro
> system.posix_acl_access=0x0200000001000600ffffffff04000600ffffffff10000600ffffffff20000400ffffffff
> trusted.SGI_ACL_FILE=0x0000000400000001ffffffff0006000000000004ffffffff0006000000000010ffffffff0006000000000020ffffffff00040000
> trusted.gfid=0xdfe13dc088bf4a779488ef72f0a879cd
> trusted.glusterfs.dht.linkto=0x7361666172692d636c69656e742d3300
>
> [root at ndovu ~]# getfattr -m . -d -e hex
> /data/glusterfs/safari/brick03/brick/complete/vancouver/refdem/tif2flt.pro
> getfattr: Removing leading '/' from absolute path names
> # file:
> data/glusterfs/safari/brick03/brick/complete/vancouver/refdem/tif2flt.pro
> system.posix_acl_access=0x0200000001000600ffffffff04000600ffffffff10000600ffffffff20000400ffffffff
> trusted.SGI_ACL_FILE=0x0000000400000001ffffffff0006000000000004ffffffff0006000000000010ffffffff0006000000000020ffffffff00040000
> trusted.gfid=0xdfe13dc088bf4a779488ef72f0a879cd
>

Is rebalance running on this volume right now? If not, can you please 
move out the file copy with "trusted.glusterfs.dht.linkto" attribute out 
of the brick directory 
(/data/glusterfs/safari/brick00/brick/complete/vancouver/refdem/tif2flt.pro) 
to an alternate location & check the behavior?

Thanks,
Vijay



More information about the Gluster-users mailing list