Aniruddha Apte <aniapte@gmail.com> writes:
> My company is considering moving to Svn from Cvs. As part of this
> process we are evaluating svn. After creating the repository, I
> imported our whole tree that has around 5300 files in 1600
> directories. Then i checked out one directory with 4 files for
> testing. However "svn log" is too slow inside this dir. It takes close
> to 15 seconds to produce the output that has 2 revisions. During this
> time the CPU usage shoots up to 100% with 97% used by Apache.
>
> I noticed that the first revision log info has a list of all paths
> affected, which in my case is 5300+1600 paths because the whole tree
> was added in this revision. This is quiet a lot of data. I guess this
> is the reason it is so slow. So this means that the 1st revision log
> will be included in the logs of all files.
> Am i right?
>
> Can someone please suggest a workaround to make svn log to take less time?
>
> Our setup is as follows:
> Svn server running on P4 2.3Ghz, Win2k box with 512 mb ram.
> svn, version 1.1.1 (r11581) compiled Oct 23 2004, 14:47:47
> Apache 2.0.52
The duration of log command should be roughly proportional to the
amount of data returned. Since every file begins at r1, and r1 is
very big, that's probably part of the problem. Still, 15 seconds is
surprising.
Can you do an experiment: put the same repository on a Unix box (that
is, create a new repository on a Unix box, and use 'svnadmin dump',
'svnadmin load' to transport the data) and see if log takes as long
from that server. Leave all other variables the same, just change the
server's operating system. That would help us diagnose this.
Thanks,
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Wed Jan 5 18:43:00 2005