On 4/24/07, Malcolm Rowe <malcolm-svn-dev@farside.org.uk> wrote:
> On Tue, Apr 24, 2007 at 09:13:48AM +0100, Taylor, Richard wrote:
> > We are using FSFS.
> > When I say 'corrupted the directory' mean the following. When the server
> > runs out of disc space during a commit and it errors at the point of sending
> > file X in directory Y. Any subsequent checkout, update or commit of the
> > repository will fail when it reaches an operation on directory Y.
> >
>
> That is utterly, utterly, strange. FSFS doesn't finalise the commit
> until after the whole transaction has been written, and up to that
> point, the transaction is invisible to other users. So I'd expect to
> see dead transactions lying about, but no impact on the actual
> repository.
What is *very* important in this context is the APR version RHEL4 is
using. Any version before 0.9.7 is bad business in terms of low-disk
situations.
> > I'd like to add this into the bug database so that it can be given some
> > consideration.
>
> Definitely, please do. If you can generate a reproduction testcase,
> that would be even better (preferably one starting by loopback-mounting
> a disk of a specified size, so anyone can run it).
bye,
Erik.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Apr 24 11:39:19 2007