Re: server problems

On Thu, Jan 24, 2002 at 11:29:22AM -0600, Karl Fogel wrote:
>...
> Ack, -1. We still need to backup the full repository on every commit.
> We have had to restore from that at least once before, when something
> really weird happened to the repository and we couldn't fix it. (Yes,
> of course we have to figure out what went wrong, but in the meantime
> we need a working repository).

Are we sure that it had a problem? During some of my work on the box last
night, I found that it simply took some directories a long time to run
db_recover. The larger the logs (or the more there were), the longer it
took.

IOW, did you let the thing run long enough?

I'd like to look into this; where did you put the copy of the repository?
I'm finding it hard to believe that it is truly broken / unfixable.

> The solution is to remove old repository backups after they're no
> longer relevant. I imagine we only should keep the most recent 20 or
> so around. We can use that script (headify.sh, just checks out the
> head tree from a repos backup) to convert the others and remove the
> repositories. That way we don't lose any information, we just make it
> a bit harder to retrieve.

Yes.

> We've already headified everything up to rev 793. I'll remove those
> repositories, and then headify some more and remove some more.

At your leisure. I've done a recovery and archive/log-toss on each of the
backups. That free'd a ton of space, so headify'ing isn't as important.

Note that I moved all the heads into /usr/backup/heads, and the older
backups into /usr/backup/old-repositories.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org