You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Trent Fisher <tr...@oracle.com> on 2010/02/19 19:23:59 UTC

svnadmin dump runs out of memory...

I was going to do a dump/load sequence on a large repository to 
rearrange things.  I got an "out of memory" error!  At first I thought 
it was sloppiness in my perl script doing the filtering, but, no, it was 
"svnadmin dump" which ran out of memory!  What's worse is it seems the 
"out of memory" error was printed on stdout rather than stderr.

I've been able to reproduce this. I've done this three times and each 
time it gets to the revision 54942 (out of 181168), at that time it is 
taking up 3.1 gigs of memory (VmSize).

Admittedly, I am doing this on a test machine which doesn't have a ton 
of memory (4 gigs physical + 10 gigs swap). It would take more to run 
the main server out of memory (it has 32+16 gigs). But the concerning 
thing is that it is seems that it is possible to do so, given a large 
enough repository.

Is this a known problem, or is it just me?

FYI, this is with Subversion 1.6.5 on an Oracle (Red Hat) Enterprise 
Linux 4 update 7 machine.

++thanks,
trent...

Re: svnadmin dump runs out of memory...

Posted by Justin Connell <ju...@propylon.com>.
Trent Fisher wrote:
> I was going to do a dump/load sequence on a large repository to 
> rearrange things.  I got an "out of memory" error!  At first I thought 
> it was sloppiness in my perl script doing the filtering, but, no, it 
> was "svnadmin dump" which ran out of memory!  What's worse is it seems 
> the "out of memory" error was printed on stdout rather than stderr.
>
> I've been able to reproduce this. I've done this three times and each 
> time it gets to the revision 54942 (out of 181168), at that time it is 
> taking up 3.1 gigs of memory (VmSize).
>
> Admittedly, I am doing this on a test machine which doesn't have a ton 
> of memory (4 gigs physical + 10 gigs swap). It would take more to run 
> the main server out of memory (it has 32+16 gigs). But the concerning 
> thing is that it is seems that it is possible to do so, given a large 
> enough repository.
>
> Is this a known problem, or is it just me?
>
> FYI, this is with Subversion 1.6.5 on an Oracle (Red Hat) Enterprise 
> Linux 4 update 7 machine.
>
> ++thanks,
> trent...
>
>
>
Hi Trent
I have had similar experience with huge repos (150 GB). To handle this 
have you considered using incremental dumps?
ie svnadmin dump /path/to/repo --revision x:y --incremental > dump1

Justin