filename too long (pjd-fstest on centos5 with ceph 0.20.2)

filename too long (pjd-fstest on centos5 with ceph 0.20.2)

[Thomas Mueller]

hi 

i'm doing some testing. i've ran pjd-fstest. on ceph (kclient) 
mountpoints there are many "file name too long" messages. 

example error message:
mkdir: `_xxxxxx-veryveryveryverylongpath/x': File name too long


fstest results:

Failed Test                   Stat Wstat Total Fail  Failed  List of Failed
-------------------------------------------------------------------------------
/tmp/fstest/tests/mkdir/00.t                36    5  13.89%  30-34
/tmp/fstest/tests/mkfifo/00.t               36    5  13.89%  30-34
/tmp/fstest/tests/open/00.t                 47    5  10.64%  30-34
Failed 3/184 test scripts, 98.37% okay. 15/1957 subtests failed, 99.23% 
okay.


if I run fstest on the local filesystem it completes without errors - so 
the "filename too long" seems to be a ceph specific error.

- Thomas

Env:
- 4x osd, 1x mds, 1x mon, 1x kclient
- ceph 0.20.2, ceph kclient 2.6.34, Cento5, pjd-fstest-20080816

Re: filename too long (pjd-fstest on centos5 with ceph 0.20.2)

[Gregory Farnum]

Thomas:
The "file name too long" errors are actually supposed to occur in a
number of places in pjd-fstest -- it's part of the correctness
testing!

That said, I don't think the kernel client shouldn't be failing any of
the tests, so we'll look at this in the next day or two using the
failed test numbers you gave us. Thanks for the bug report!
-Greg

On Mon, May 31, 2010 at 3:35 AM, Thomas Mueller <thomas@chaschperli.ch> wrote:
> hi
>
> i'm doing some testing. i've ran pjd-fstest. on ceph (kclient)
> mountpoints there are many "file name too long" messages.
>
> example error message:
> mkdir: `_xxxxxx-veryveryveryverylongpath/x': File name too long
>
>
> fstest results:
>
> Failed Test                   Stat Wstat Total Fail  Failed  List of Failed
> -------------------------------------------------------------------------------
> /tmp/fstest/tests/mkdir/00.t                36    5  13.89%  30-34
> /tmp/fstest/tests/mkfifo/00.t               36    5  13.89%  30-34
> /tmp/fstest/tests/open/00.t                 47    5  10.64%  30-34
> Failed 3/184 test scripts, 98.37% okay. 15/1957 subtests failed, 99.23%
> okay.
>
>
> if I run fstest on the local filesystem it completes without errors - so
> the "filename too long" seems to be a ceph specific error.
>
> - Thomas
>
> Env:
> - 4x osd, 1x mds, 1x mon, 1x kclient
> - ceph 0.20.2, ceph kclient 2.6.34, Cento5, pjd-fstest-20080816
>
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: filename too long (pjd-fstest on centos5 with ceph 0.20.2)

[Thomas Mueller]

Am Mon, 31 May 2010 03:42:33 -0700 schrieb Gregory Farnum:

> Thomas:
> The "file name too long" errors are actually supposed to occur in a
> number of places in pjd-fstest -- it's part of the correctness testing!

I can't see these error messages if I run the test on the local fs. 

- Thomas

Re: filename too long (pjd-fstest on centos5 with ceph 0.20.2)

[Gregory Farnum]

Thomas:
We've checked into this a bit here, and we aren't seeing any failures
from pjd ourselves. The tests you reference as failing are all
mtime/atime/ctime checks, so our best guess is that your client and
server don't have synchronized clocks, which could cause this
behavior.

As for the "mkdir: `_xxxxxx-veryveryveryverylongpath/x': File name too
long" issue, we think that may be a result of the shell complaining --
Ceph itself doesn't have any filename size limits so we just set the
max length to be the max path length. We're looking into it
separately, but that definitely isn't causing any pjd failures as pjd
expects those filenames to not be accepted. :)

On Mon, May 31, 2010 at 4:31 AM, Thomas Mueller <thomas@chaschperli.ch> wrote:
> Am Mon, 31 May 2010 03:42:33 -0700 schrieb Gregory Farnum:
>
>> Thomas:
>> The "file name too long" errors are actually supposed to occur in a
>> number of places in pjd-fstest -- it's part of the correctness testing!
>
> I can't see these error messages if I run the test on the local fs.
>
> - Thomas
>
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: filename too long (pjd-fstest on centos5 with ceph 0.20.2)

[Sage Weil]

On Mon, 31 May 2010, Thomas Mueller wrote:
> Am Mon, 31 May 2010 03:42:33 -0700 schrieb Gregory Farnum:
> 
> > Thomas:
> > The "file name too long" errors are actually supposed to occur in a
> > number of places in pjd-fstest -- it's part of the correctness testing!
> 
> I can't see these error messages if I run the test on the local fs. 

Fixed this.  The problem was that we were reporting PATH_MAX instead of 
NAME_MAX as the max file name length in statfs.  The pjd tests actually 
all succeeded, but the script would spit out those errors when trying to 
clean up.

Thanks for the heads up!
sage

Re: filename too long (pjd-fstest on centos5 with ceph 0.20.2)

[Thomas Mueller]

Am Tue, 01 Jun 2010 12:51:45 -0700 schrieb Sage Weil:

> On Mon, 31 May 2010, Thomas Mueller wrote:
>> Am Mon, 31 May 2010 03:42:33 -0700 schrieb Gregory Farnum:
>> 
>> > Thomas:
>> > The "file name too long" errors are actually supposed to occur in a
>> > number of places in pjd-fstest -- it's part of the correctness
>> > testing!
>> 
>> I can't see these error messages if I run the test on the local fs.
> 
> Fixed this.  The problem was that we were reporting PATH_MAX instead of
> NAME_MAX as the max file name length in statfs.  The pjd tests actually
> all succeeded, but the script would spit out those errors when trying to
> clean up.
> 
> Thanks for the heads up!
> sage

just tested the ceph-client-standalone/master-backport - the messages* 
are gone. Thanks!

- Thomas

* http://chaschperli.ch/testing/testDetails.php?test=88&build=47