cmpilato@collab.net writes:
> "Jay Freeman \(saurik\)" <saurik@saurik.com> writes:
>
> > Is there a reason why the size on the server after an import is twice as
> > large as the size of the files that were uploaded into the repository?
> > This seems really screwy to me.
>
> I've not duplicated this example (yet), but if you use 'du -k' do you
> get different results?
Okay, now I have duplicated it. I imported a directory of song
lyrics, sized as such:
~/public_html % du -sk lyrics
352 lyrics
Then I looked at the size of the repository:
~/public_html % du -k /usr/www/repositories/test
4 /usr/www/repositories/test/dav
4 /usr/www/repositories/test/conf
8 /usr/www/repositories/test/locks
24 /usr/www/repositories/test/hooks
1496 /usr/www/repositories/test/db
1544 /usr/www/repositories/test
Wow. That looks big. But let's check out that db/ folder. I
suspect...
~/public_html % du -k /usr/www/repositories/test/db/*
8 /usr/www/repositories/test/db/__db.001
268 /usr/www/repositories/test/db/__db.002
36 /usr/www/repositories/test/db/__db.003
228 /usr/www/repositories/test/db/__db.004
12 /usr/www/repositories/test/db/__db.005
560 /usr/www/repositories/test/db/log.0000000001
8 /usr/www/repositories/test/db/nodes
8 /usr/www/repositories/test/db/representations
8 /usr/www/repositories/test/db/revisions
348 /usr/www/repositories/test/db/strings
8 /usr/www/repositories/test/db/transactions
Ah, yes. The strings table (full of the actual file contents + dir
entries list) looks to be about the right size. But those __db* files
are shared memory regions, storing state for would-be concurrent
accessors. And the log file? That's BDB journaling data. So,
really, the extra data is BDB overhead.
So there you have it.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:51 2006