[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

working copy is too big

From: Kevin Brannen <kevin_brannen_at_stercomm.com>
Date: 2004-09-02 22:53:25 CEST

I'm reasonably certain I know the answer to this, but I want to ask just
in case there's an answer I'm not seeing because I'm so new to this.

Q: I've imported my code and pulled out a working copy, watching my disk
space get sucked dry. I know that SVN keeps a copy of the pristine file
in the metadata .svn dirs, but is there any way to avoid that and still
be able to make edits and commits at a later time? A performance
penalty to do this is quite acceptable.

A few facts to help the reader understand why I'm asking.

Each developer and server machine needs a "vendor branch" working copy
(and our application working copy too but I'm ignoring that for now).
Presently, we're in CVS. This is an issue because of our size:

$ find /usr/local/vendor -print | grep -v CVS | wc -l
$ du -s /usr/local/vendor
5841896 /usr/local/vendor

So we've got over 51K files & dirs that take up over 5.8GB. (CVS
metadata is 35MB of that size, mostly negligable).

But my working copy out of SVN is much bigger:

$ du -s ./vendor
11341133 vendor

Here I'm using over 11GB. This is a major problem for us as many of our
machines just don't have that much disk space available, and adding it
is not an option in some places (try pricing DASD for a zSeries
mainframe [running Linux]. :-)

Part of me would like to just do a "svn export" sort of thing as there
are only a dozen of us out of about 250 developers who change those
files, and of the 51K files, we probably only change 1 or 2 a day
(ignoring when we drop whole new vendor releases into a subdir). So
that 5.3G penalty is hard to take.

Another part of me wouldn't mind putting the working copy on 1 machine
with lots of disk and NFS mounting it out to everyone, but we're
geographically spread out, so NFS over a WAN does not excite any of us
(especially the security folks. :-)

Our app code is only about 1GB, so that's not too bad, but the vendor
branch (it'll be multiple under a single dir in reality) is another
story. As much as I like Subversion's better features, is CVS really
the better tool for *this situation*? Or is there something I'm not
seeing and understanding?


To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Sep 2 22:53:45 2004

This is an archived mail posted to the Subversion Users mailing list.