Re: svnadmin load runs out of memory

Dave Oxley wrote:

> I know there is an incompatability between 0.33.1 and 0.34.0. Thats one
> of the reasons I was doing the dump and load cycle. The other reason is
> that I'd forgotten to do the deltify and I needed to reduce the size of
> the repositiory (It had got up to 4.5Gb although there are only 146
> revisions!). Another question on this: Does a svnadmin load do the
> deltification?

I believe svnadmin load *does* do deltification automatically.

> There are several revisions that fail. rev 65 fails but then I can run a
> recover and it says that rev 65 exists(?) so I can then continue and
> load the other revisions. But before the load is complete it fails on
> about (128) and a recover doesn't fix it.

I'm not completely sure how svnadmin load works, but I could speculate
that it successfully loads r65, then fails to deltify. And it also
fails on r128 (possibly because of the r65 failure).

Seems to me like deltification is the problem. If there was a way to do
svnadmin load without deltification, you might have more luck. Though
of course you're trying to reduce the size.

I imagine a 65 MB SQL file would have lots of changes (millions?) all
over the place. Maybe the deltification code doesn't scale to it.

> The only file that was changed in the revisions that fail is the large
> SQL script.

Are they incremental dumps? (Not sure how much difference it makes
though; I suppose all the unchanged nodes in a non-incremental dump
would be easily ignored.)

If this is a bug it's a very strange and serious one.

> I think you're right about the memory. I'm not entirely sure how Windows
> works (Who is?), but I think a program can only ever allocate a max of
> 4Gb. The additional 2Gb may or may not be usable by other programs?

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Dec 8 02:25:56 2003