Wadsworth, Eric (Contractor) wrote:
> We've been using subversion for about a year, and our repository is
> getting
> bulky. We are looking for a way to amputate the really old stuff, so our
> backups are smaller and more managable. We very rarely need revisions more
> than 6 months old. We'd keep an archive of the complete repository, of
> course, just in case.
>
> Is there some sort of approved way of doing this? As time goes on, this
> problem will only get worse.
There's no real approved or elegant way to do this that I know of. I have a
few ideas how it might be possible to script something along these lines,
but I think there would be scripted munging of dumpfiles involved - so, not
ideal - however...
Is the repository server short on disk space? If not, I have a much nicer
suggestion for you:
Keep the full history online in the repository.
Then use "svnadmin dump" with revision range specifiers to create sections
of dumpfile for backup.
I'm going to choose some arbitrary numbers here for illustrative purposes:
Suppose up to r5000 is considered 'old' and can be 'frozen' - by which I
mean, we will back it up once, secure those backups, and not back it up ever
again. Note the tradeoff is that you will not be backing up any changes to
log messages or other revprops in the old range - that shouldn't be a
problem, if you've set the threshold sufficiently far in the past, but to be
sure, you may want to edit the pre-revprop-change hook to disallow revprop
changes on 'frozen' revisions.
We do a 'freezing' backup: "svnadmin dump -r 1:5000 repos".
Now, your routine backup becomes "svnadmin dump --incremental -r 5001:HEAD
repos".
Note that you *will* *need* the r1-r5000 dump too should you ever need to
restore this backup.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Wed Sep 8 07:17:22 2004