Re: "400 bad request" when trying svn co to NFS-mounted disk?

(sorry, did not CC this to list)

Tan Ruyan wrote:
> Please check the settings of your anti-virus software.

Hi Tan,

Thank you for taking an interest.

We're not using any Anti-Virus software in this 3-machine setup;

It's not installed on the client, the SVN server or the NFS server.

I've performed some further experiments, and have tried the svn co on a
filesystem exported from a modern NFS server (also an CentOS 4 machine -
the other NFS server is ancient) and there it works, so this is
essentially an NFS thing that the svn-client runs into.

One of my developers has looked at some straces, and figures that the
issue is with writing to a file (on the client), and then immediately
checking the size of the file from the file system. On our (old)
NFS-server, the filesize is reported as '0', and the svn-client aborts
with this '400 bad request' message.

We'll be upgrading our NFS server in any case, but an option like
'--ignore-bad-fstat' for the svn client might make life easier for
others :-)

-S

> On 3/26/07, *Simen Thoresen* <simentt@dolphinics.no
> <mailto:simentt@dolphinics.no>> wrote:
>
> Hi Subversion Users,
>
> We're in the process of establishing a new SVN-server to hold our old
> CVS repository, but I've run into an issue with the client checking out
> to an NFS-mounted disk;
>
> Our experience pretty much mirrors this;
>
> http://www.centos.org/modules/newbb/viewtopic.php?topic_id=4875
>
> To recap;
>
> cd to ~, mounted on an old NFS-server
>
> optimist-1:~$ svn co http://scubversion/repos/test/DIS/trunk
> <http://scubversion/repos/test/DIS/trunk>
> svn: REPORT request failed on '/repos/test/!svn/vcc/default'
> svn: REPORT of '/repos/test/!svn/vcc/default': 400 Bad Request
> (http://scubversion)
> optimist-1:~$
>
> cd to /tmp (locally mounted)
> optimist-1:/tmp$ svn co http://scubversion/repos/test/DIS/trunk
> A trunk/build
> A trunk/build/WindowsView
> A trunk/build/WindowsView/mkmakefiles.sh
> A trunk/build/WindowsView/mkmakefiles.pl
> A trunk/build/WindowsView/Makefile
> A trunk/adm
> A trunk/adm/test
> A trunk/adm/test/sisci_benchmarks....
> (etc).
>
> This is whith the CentOS-provided svn-1.1.4 , but also applies to an
> updated package (svn-1.3.2).
>
> The problem also occurs with the https protocol (apparently a favourite
> fix to avoid caching), none of the other '400-bad request' usual
> suspects like caching-proxies or antivirus-agents are involved (I've
> googled this for some time now)
>
> In short - svn co works to local disk, but fails to otherwise accessible
> NFS-mounted disk.
>
> Is this a known issue, and is there a known fix for it?
>
> Yours,
> -S
> --
> Simen Thoresen, Dolphin ICS
> Systems Administration and Wulfkit Support
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
> <mailto:users-unsubscribe@subversion.tigris.org>
> For additional commands, e-mail: users-help@subversion.tigris.org
> <mailto:users-help@subversion.tigris.org>
>
>
>
>
> --
> Tan Ruyan
> http://www.iUseSVN.com (Chinese SVN Forum)
> http://www.PCplayer.cn (My blog)

-- 
Simen Thoresen, Dolphin ICS
Systems Administration and Wulfkit Support
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org