[Gluster-users] Random directory/files gets unavailable after sometime

[Fernando Morgenstern]

Em 07/01/2010, às 14:27, Anand Avati escreveu:

> On Thu, Jan 7, 2010 at 6:22 PM, Fernando Morgenstern
> <fernando at consultorpc.com> wrote:
>> Hello,
>> 
>> I am using glusterfs v3.0.0 and having some problems with random directory/files.
>> 
>> They work fine for some time ( hours ) and them suddenly gets unavailable:
>> 
>> # ls -lh
>> ls: cannot access MyDir: No such file or directory
>> total 107M
>> d????????? ? ?      ?         ?            ? MyDir
>> ( long dir list, intentionally hidden )
>> 
>> At the logs i get a lot of messages like those ones:
>> 
>> [2010-01-07 13:36:16] W [fuse-bridge.c:793:fuse_getattr] glusterfs-fuse: 270708: GETATTR 3057375160 (fuse_loc_fill() failed)
>> [2010-01-07 13:36:16] W [fuse-bridge.c:793:fuse_getattr] glusterfs-fuse: 270735: GETATTR 3057375160 (fuse_loc_fill() failed)
>> [2010-01-07 13:36:26] W [fuse-bridge.c:793:fuse_getattr] glusterfs-fuse: 270841: GETATTR 3057375160 (fuse_loc_fill() failed)
>> [2010-01-07 13:36:26] W [fuse-bridge.c:793:fuse_getattr] glusterfs-fuse: 270844: GETATTR 3057375160 (fuse_loc_fill() failed)
>> [2010-01-07 13:36:26] W [fuse-bridge.c:793:fuse_getattr] glusterfs-fuse: 270847: GETATTR 3057375160 (fuse_loc_fill() failed)
>> [2010-01-07 13:36:26] W [fuse-bridge.c:793:fuse_getattr] glusterfs-fuse: 270874: GETATTR 3057375160 (fuse_loc_fill() failed)
>> 
>> Any ideas of what is causing this issue?
>> 
> 
> Can you describe what kind of IO is performed over the mountpoint?
> What applications and tools use the mountpoint? Also, posting the full
> log file (including the config on top) will help us analyze better.
> 
> Avati

Hello,

Mountpoint is used by a tomcat cluster. It is not in production yet, so we are only deploying apps and web accesses are only done by developers testing applications.

The log is available at http://pastebin.ca/1741574. I haven't posted it entirely, because it has more than 2MB and after the fuser_loc_fill() errors start, the only thing that change is the numbers ( error message remains the same ).

Thanks,
---

Fernando Marcelo
www.consultorpc.com
fernando at consultorpc.com