You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Kylo Ginsberg <ky...@gmail.com> on 2007/04/18 19:39:29 UTC
Subversion robust to server-side disk full errors?
I'm running subversion 1.4.3 under Fedora Core 4 with both svn:// and
http:// access methods supported. Is subversion tested against the
error case where the server runs out of disk space mid-commit? (I
seem to recall a report of a repository corruption problem resulting
from this scenario.)
If not, is there a recommended practice for a pre-commit script that
will reject a commit if it would cause disk full? (This second
question assumes subversion is robust to the case where it runs out of
disk space while creating the transaction file itself.)
Thanks,
Kylo
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: Subversion robust to server-side disk full errors?
Posted by Ryan Schmidt <su...@ryandesign.com>.
On Apr 18, 2007, at 14:39, Kylo Ginsberg wrote:
> I'm running subversion 1.4.3 under Fedora Core 4 with both svn:// and
> http:// access methods supported. Is subversion tested against the
> error case where the server runs out of disk space mid-commit? (I
> seem to recall a report of a repository corruption problem resulting
> from this scenario.)
>
> If not, is there a recommended practice for a pre-commit script that
> will reject a commit if it would cause disk full? (This second
> question assumes subversion is robust to the case where it runs out of
> disk space while creating the transaction file itself.)
I don't know if Subversion is robust in either of these cases.
But you could write a start-commit hook to reject all commits when
disk space is less than some number. The start-commit hook is
executed before the client has transferred the transaction data to
the server.
--
To reply to the mailing list, please use your mailer's Reply To All
function
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: Subversion robust to server-side disk full errors?
Posted by Erik Huelsmann <eh...@gmail.com>.
On 4/19/07, Kylo Ginsberg <ky...@gmail.com> wrote:
> On 4/18/07, Erik Huelsmann <eh...@gmail.com> wrote:
> > On 4/18/07, Kylo Ginsberg <ky...@gmail.com> wrote:
> > > I'm running subversion 1.4.3 under Fedora Core 4 with both svn:// and
> > > http:// access methods supported. Is subversion tested against the
> > > error case where the server runs out of disk space mid-commit?
>
> > Subversion checks all the return values of all I/O calls, including
> > read/write calls. Yes, it should be safe to run it in a low-diskspace
> > situation.
>
> Thanks; that's reassuring to hear. Is a low-diskspace situation part
> of the regression test suite? (I looked briefly through the
> subversion/tests source but didn't get oriented well enough to answer
> this question.)
No, it's not... (So you don't have to look any further)
> > There's one exception: you have to use a recent enough APR
> > version: more recent than 0.9.7, I think. ... I'm not
> > sure which version FC4 has, but apr-config --version should be able to
> > tell you.
> >
>
> Well, apr-config --version shows 0.9.6, but if I cd to the directory
> in which I built subversion 1.4.3, apr/apr-config --version shows
> 0.9.12. How can I confirm that svn/svnserve/mod_dav_svn will find the
> more current apr lib(s)? (I built from a directory with both the
> subversion tarball and the subversion dependencies tarball.)
mod_dav_svn will use the version against which your Apache has been
built, regardless of whether you built it against 0.9.12, so, if you
use the system-delivered Apache, I'd say you're using 0.9.6...
I just verified: yes, you do need at least 0.9.7 which contains the
fix for buffered files to fix the low-disk condition.
You should be able to replace the 0.9.6 image without problems with
the 0.9.12 image though: APR is designed to be 100% backward
compatible.
HTH,
Erik.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: Subversion robust to server-side disk full errors?
Posted by Kylo Ginsberg <ky...@gmail.com>.
On 4/18/07, Erik Huelsmann <eh...@gmail.com> wrote:
> On 4/18/07, Kylo Ginsberg <ky...@gmail.com> wrote:
> > I'm running subversion 1.4.3 under Fedora Core 4 with both svn:// and
> > http:// access methods supported. Is subversion tested against the
> > error case where the server runs out of disk space mid-commit?
> Subversion checks all the return values of all I/O calls, including
> read/write calls. Yes, it should be safe to run it in a low-diskspace
> situation.
Thanks; that's reassuring to hear. Is a low-diskspace situation part
of the regression test suite? (I looked briefly through the
subversion/tests source but didn't get oriented well enough to answer
this question.)
> There's one exception: you have to use a recent enough APR
> version: more recent than 0.9.7, I think. ... I'm not
> sure which version FC4 has, but apr-config --version should be able to
> tell you.
>
Well, apr-config --version shows 0.9.6, but if I cd to the directory
in which I built subversion 1.4.3, apr/apr-config --version shows
0.9.12. How can I confirm that svn/svnserve/mod_dav_svn will find the
more current apr lib(s)? (I built from a directory with both the
subversion tarball and the subversion dependencies tarball.)
Thanks,
Kylo
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: Subversion robust to server-side disk full errors?
Posted by Erik Huelsmann <eh...@gmail.com>.
On 4/18/07, Kylo Ginsberg <ky...@gmail.com> wrote:
> I'm running subversion 1.4.3 under Fedora Core 4 with both svn:// and
> http:// access methods supported. Is subversion tested against the
> error case where the server runs out of disk space mid-commit? (I
> seem to recall a report of a repository corruption problem resulting
> from this scenario.)
>
> If not, is there a recommended practice for a pre-commit script that
> will reject a commit if it would cause disk full? (This second
> question assumes subversion is robust to the case where it runs out of
> disk space while creating the transaction file itself.)
Subversion checks all the return values of all I/O calls, including
read/write calls. Yes, it should be safe to run it in a low-diskspace
situation. There's one exception: you have to use a recent enough APR
version: more recent than 0.9.7, I think.
Older versions had a problem with buffered files in a low-diskspace
situation and could end up in an endless loop because of it. I'm not
sure which version FC4 has, but apr-config --version should be able to
tell you.
HTH,
Erik.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org