Young Hyun wrote:
> Why do the Berkeley DB log files grow for conceptually read-only
> operations like svnlook, 'svn list', and 'svn cat'? I am especially
> surprised about svnlook.
>
> Specifically, try this:
>
> perl -e 'my $x = 100; while ($x--) { print $x, "\n"; \
> `svnlook youngest /path/to/repos`; }'
>
> and look at the size of your log files before and after.
>
> I'm managing a slowly changing repository that will get a moderate amount
> (~400-800/day) of read-only requests, and it seems wasteful for the logs
> to grow when nothing is actually changing. It becomes wasteful if I
> choose to slavishly back up the logs (which I probably won't).
>
> I'm running Subversion 1.0.0 with Berkeley DB 4.1.25. (Yes, these are not
> the latest, but I have to use them for now.)
Upgrade. Seriously, upgrade!
There have been 2 security bugs fixed on the 1.0 line, each of which warranted an immediate release.
And BDB 4.1 is suspected to corrupt databases from time to time.
Anyway, to answer your question, most subversion operations open a Berkeley transaction, to ensure they view a consistent set of
data throughout the operation.
There's some thought about eliminating these transactions where it is safe to do so, but it's not a particularly high priority
issue.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Fri Jun 11 12:27:22 2004