Need help on Subversion dump/load feature....Thanks

[Reddy1413 <re...@gmail.com>]

Greetings,

As part of the experiment with svn dump/load ,I am using svnadmin dump and
load utilities on the same box.

The dump and load provide some very nice features for making the migration
of large-ish repositories a manageable task. My thanks to those responsible.
However, am having trouble with an aspect of the load on the target
migration system.

Background: Am a relative newbie but have read the svn book particularly
Repository Maintenance/Migrating Data Elsewhere (more than once, to be
honest  ; done a subversion software upgrade and a bit of other hands-on.

To keep the dump file size "under control" am using both the -r switch to
dump a range of revisions, and the --deltas switch for "compressed, binary
differences" as the book says. For example,

svnadmin dump /repos_path -r0:50 --deltas > repos_r0-50.dmp
svnadmin dump /repos_path -r51:HEAD --deltas > repos_r51-HEAD.dmp
et cetera...

I moved the *dmp files where I want to load.

Now, the moment of truth:

The load appears to work fine for the rev 0-50 dump file. "Appears" because
only svnlook history /repos-path has been used on the target system as a
quick check to see that the intended revisions are listed.

However, for the next dump file and contiguous set of revisions:

svnadmin create target_repos_path
svnadmin load /target_repos_path < repos_r0-50.dmp  #worked fine no problem.
svnadmin load /target_repos_path < repos_r51-HEAD.dmp  #bouncer, Thrown
error as below

the above command returns nearly immediately and the following is output to
stderr:


<<< Started new transaction, based on original revision 51
svnadmin: File already exists: filesystem 'loadtest/db', transaction
'50-1e', path 'proj16'
     * adding path : proj16 ...


The /target_repos_path is named the same as the source /repos_path (fyi)

The svn book says --deltas is useful and a dump using -r followed by
svnadmin load in order from bottom up is supposed to work fine.

So there must be something else about the dump switches or about the load
that I do not understand or missed in the svn book.

Thanks very much in advance for any illumination on the subject.

Re: Need help on Subversion dump/load feature....Thanks

[Nico Kadel-Garcia <nk...@gmail.com>]

On Fri, Oct 8, 2010 at 12:19 PM, Reddy1413 <re...@gmail.com> wrote:
> Greetings,
> As part of the experiment with svn dump/load ,I am using svnadmin dump and
> load utilities on the same box.
> The dump and load provide some very nice features for making the migration
> of large-ish repositories a manageable task. My thanks to those responsible.
> However, am having trouble with an aspect of the load on the target
> migration system.
> Background: Am a relative newbie but have read the svn book particularly
> Repository Maintenance/Migrating Data Elsewhere (more than once, to be
> honest  ; done a subversion software upgrade and a bit of other hands-on.
> To keep the dump file size "under control" am using both the -r switch to
> dump a range of revisions, and the --deltas switch for "compressed, binary
> differences" as the book says. For example,


Stop here. Why are you doing this? Do you need to compress your dump
files? Or can you run the output through the 'split' command, and cat
the contents of the split files into the svnadmin load command to
avoid splitting your dumps?

Also, which Subversion version are you using? The more recent versions
do some very interesting things to clean up old formatting errors,
which you may find useful but can make the overall task more complex.