[Gluster-devel] bug with TLA 313?

Brent A Nelson brent at phys.ufl.edu
Tue Jul 17 14:01:37 UTC 2007


Just a quick note that this doesn't seem to be any sort of corruption 
issue.  I completely emptied all my shares (even removing lost+found) and 
my namespace and rsynced the corresponding AFR shares and namespace.  The 
only thing different between the AFRs would be ctimes.

I restarted everything, and did:
ls -al /beast
ls: /beast: File exists
ls: /beast/.: File exists
total 8
drwxr-xr-x  2 root root 4096 2007-07-17 09:27 .
drwxr-xr-x 27 root root 4096 2007-07-02 10:18 ..

I also tried disabling readahead and writebehind (my only performance 
translators).  It didn't help.  Changing the unify from alu to rr also 
didn't help.

I then tried "glusterfs -f /etc/glusterfs/beast -n mirror0 /beast" to 
mount a single AFR, no unify.  It STILL produces the same messages.

I then tried "glusterfs -f /etc/glusterfs/beast -n share0-0 /beast" to 
mount a simple, single share used as half of an AFR.  Same issue.

I then stripped down a server to serve out one single storage/posix share, 
with no posix locks (I wasn't using any other translators on the server 
side, apart from protocol/server, of course).  I mounted that share as in 
the previous attempt.  No difference!

So, this issue occurs even with just protocol/client, protocol/server, and 
storage/posix in use.  As barebones as you can get.  Almost.

One more try.  No glusterfsd, and glusterfs accesses a single 
storage/posix directly:

ls -al /beast
ls: /beast: File exists
ls: /beast/.: File exists
total 8
drwxr-xr-x  2 root root 4096 2007-07-17 09:27 .
drwxr-xr-x 27 root root 4096 2007-07-02 10:18 ..

No difference, even with just glusterfs directly accessing a single, local 
storage/posix, with no other translators.  Spec is simply:

volume share0
   type storage/posix                   # POSIX FS translator
   option directory /share0             # Export this directory
end-volume

Ubuntu Feisty, Fuse 2.6.3.

Any ideas?

Thanks,

Brent


On Sat, 14 Jul 2007, Brent A Nelson wrote:

> It's the same spec I was using previously (AFRed namespace cache, unified 
> AFRs spread across four servers, posix-locks, readahead, and writebehind). 
> It's not just the top-level directory; it's everywhere.
>
> Thanks,
>
> Brent
>
> On Sat, 14 Jul 2007, Anand Avati wrote:
>
>> Brent,
>> this is strange, we are having patch-313 work pretty smooth so far. are
>> there any changes in your spec? is this behaviour seen only in this
>> particular directory or 'anywhere' in general? please attach your spec so
>> that we can try to reproduce it in our labs.
>> 
>> thanks,
>> avati
>> 
>> 2007/7/14, Brent A Nelson <brent at phys.ufl.edu>:
>>> 
>>> Updating to the latest TLA patch, I got odd issues just with "ls":
>>> 
>>> Example:
>>> 
>>> ls -al /beast/
>>> ls: /beast/: No such file or directory
>>> ls: /beast/.: No such file or directory
>>> ls: /beast/lost+found: No such file or directory
>>> ls: /beast/usr0: No such file or directory
>>> ls: /beast/usr: No such file or directory
>>> total 32
>>> drwxr-xr-x  5 root root  4096 2007-07-13 16:18 .
>>> drwxr-xr-x 27 root root  4096 2007-06-25 18:34 ..
>>> drwx------  2 root root 16384 2007-06-25 17:08 lost+found
>>> drwxr-xr-x 10 root root  4096 2007-06-18 13:31 usr
>>> drwxr-xr-x 10 root root  4096 2007-06-18 13:31 usr0
>>> 
>>> I have one machine that is no longer returning from an "ls".  I get other
>>> messages sometimes, not just "No such file or directory", but also "Bad
>>> file descriptor" or even "File exists".  These extraneous messages are
>>> also occurring when copying from the GlusterFS to the GlusterFS.  The
>>> files and directories mentioned do, in fact, exist, no matter what the
>>> extraneous error message says.
>>> 
>>> Is there a known issue with the current patchset?
>>> 
>>> Thanks,
>>> 
>>> Brent
>>> 
>>> 
>>> _______________________________________________
>>> Gluster-devel mailing list
>>> Gluster-devel at nongnu.org
>>> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>>> 
>> 
>> 
>> 
>> -- 
>> Anand V. Avati
>> 
>





More information about the Gluster-devel mailing list