[CentOS] CentOS6 gvfs disk thrashing

Mon Dec 1 09:47:38 UTC 2014
isdtor <isdtor at gmail.com>

On 25/11/2014, isdtor <isdtor at gmail.com> wrote:
> We recently started upgrading users' workstations to CentOS6. Now
> we've come across an interesting issue:
> ~user/.local/share/gvfs-metadata is filling up with tons of files.
> Once there are around 69270 of them, no more are being created, and
> the NetApp filer the user's home is on logs:
>
> Directory ..../user/.local/share/gvfs-metadata/ reached the maxdirsize
> limit. Reduce the number of files or use the vol options command to
> increase this limit..
>
> Nearly all of the files are exactly 32k in size and they are named
> .openXXXXXX. All of them are being created within the same range of
> timestamps, with a time that's in the past. I.e. even when deleted,
> they are being recreated with these timestamps.
>
> drwx------     2 user group 5353472 Nov 15 17:39 ./
> drwxr-xr-x     5 user group    4096 Oct 29 11:05 ../
> -rw-r--r-- 25699 user group   32768 Nov 11 17:11 .open0018OX
> -rw-r--r-- 42962 user group   32768 Nov 11 17:10 .open0019OX
> -rw-r--r-- 42962 user group   32768 Nov 11 17:10 .open002EPX
> -rw-r--r-- 42962 user group   32768 Nov 11 17:10 .open002GPX
> -rw-r--r-- 25699 user group   32768 Nov 11 17:11 .open002OPX
> -rw-r--r-- 42962 user group   32768 Nov 11 17:10 .open0037OX
> -rw-r--r-- 42962 user group   32768 Nov 11 17:10 .open003MPX
> -rw-r--r-- 42962 user group   32768 Nov 11 17:10 .open004NPX
> -rw-r--r-- 25699 user group   32768 Nov 11 17:11 .open0067OX
>
> According to lsof, the only processes operating in this directory are
> a few instances of nautilus and /usr/libexec/gvfsd-metadata. There's a
> whole bunch of gvfs* processes running
>
> user     21209     1  0 Nov11 ?        00:00:00 /usr/libexec/gvfsd
> user     21549     1  0 Nov11 ?        00:03:17
> /usr/libexec/gvfsd-trash --spawner :1.2 /org/gtk/gvfs/exec_spaw/0
> user     21788     1  0 Nov11 ?        00:00:06
> /usr/libexec/gvfs-gdu-volume-monitor
> user     21829     1  0 Nov11 ?        00:00:00
> /usr/libexec/gvfs-gphoto2-volume-monitor
> user     21849     1  0 Nov11 ?        00:00:25
> /usr/libexec/gvfs-afc-volume-monitor
> user     22150     1  0 Nov11 ?        00:00:00 /usr/libexec/gvfsd-metadata
> user     22323     1  0 Nov11 ?        00:00:00
> /usr/libexec/gvfsd-burn --spawner :1.2 /org/gtk/gvfs/exec_spaw/1
>
> Does anyone have an idea what's going on here? And how to stop it?

The more machines we're moving to CentOS6, the more urgent this
problems becomes. Is there any suggestion what could be causing this
behavior?