Please don't top-post on this list. Put your reply below or inline.
More below ...
> -----Original Message-----
> From: Tennebø Frode [mailto:frode.tennebo_at_saabgroup.com]
> Sent: Friday, November 11, 2011 6:39 AM
> To: Waseem Shahzad; users_at_subversion.apache.org
> Subject: RE: Database Corruption in SVN
>
>> Is there any way that SVN database may be corrupted. Any
>> story , Issue , Experience..
>
> The only time I have encountered this was when a collegue set up a BDB repository and accessed it using file:/// on an NFS mounted disk from several clients.
>
> -Frode
>
On Fri, Nov 11, 2011 at 12:46 PM, Waseem Shahzad
<waseem.shahzad_at_curemd.com> wrote:
> This means that corruption only if BDB , not in case of FSFS?
>
It happens, also with FSFS, though it's quite rare. Please search the
mailing-list archives [1] if you want to know more.
Some of these corruptions have been attributed to low-level problems
(problems with the filesystem, incorrect NFS configuration, harddisk
crash, ...). Others have lead to clear bug reports and subsequent
fixes for FSFS. And still others have not been successfully diagnosed,
were impossible to reproduce, ...
It's quite rare, but to be on the safe side, I'd recommend creating an
incremental dump file from every revision from the post-commit hook,
as described in another thread [2]. That technique is also a safeguard
for detecting corrupt revision files as early as possible, because
those will give an error when you try to dump them (you'll probably
want to make it a little bit more sophisticated then, capturing the
exit code of the dump-invocation, and sending an email to the svn
admin in case it fails or something like that (left as an exercise for
the reader)).
[1] http://subversion.apache.org/mailing-lists.html
[2] http://svn.haxx.se/users/archive-2011-07/0410.shtml
--
Johan
Received on 2011-11-11 21:20:07 CET