Andy Levy wrote:
> On 6/14/07, Troy Bull <troy.bull@uni.edu> wrote:
>> Andy Levy wrote:
>> > On 6/14/07, Troy Bull <troy.bull@uni.edu> wrote:
>> >> Greetings:
>> >>
>> >> I posted yesterday about this but I have much more information today.
>> >> First, I have a subversion repository I am responsible for, I
>> recently
>> >> switched over from CVS. My repository has 419 "projects" on the top
>> >> level directory. In each of these project directories I have 3
>> folders
>> >> trunk, branches, and tags.
>> >>
>> >> I have an initial import of the head of my cvsnt repository which
>> >> consisted of each project (419 of then) for a total of about 16k
>> files,
>> >> these are a mix of html, perl, images, and pdf's. My repository has
>> >> been live about a week, I think we are about r 140 as I type
>> this. My
>> >> revision 1 adds almost all of the 16k files in one import.
>> >>
>> >> I use http as my access method, (mod_dav_svn), my problem is svn
>> log -v
>> >> http://machine.domain.com/svn/Access seems to take forever.
>> >>
>> >> Here are some numbers in the format command(time in seconds to
>> complete)
>> >>
>> >> svn log -v http://source.its-is.uni.edu/svn/Access (104) - fs-type
>> = bdb
>> >> svn log -v http://source.its-is.uni.edu/svn/AccessFSFS (79) - fs-type
>> >> = fsfs
>> >> svn log -v file:///source/AccessFSFS(1) - fs-type=fsfs
>> >>
>> >> Is this what I should be expecting? Is there anything I can do to
>> >> improve my access times for this operation? Many of my "clients" use
>> >> TortoiseSVN on windows and it times out before the command completes.
>> >> Any help would be greatly appreciated.
>> >
>> > My repository has 1490 revisions, not sure how many files but it's
>> > fewer than yours (log operation speed should depend more on the # of
>> > revisions, not the size of the repository I think) and it took me 15
>> > seconds to get svn log -v on the root of the repository via HTTP. FSFS
>> > backend, SVN 1.2.3 on Apache 2.0.59, Win2K.
>> >
>> > Have you ruled out network problems?
>> These numbers were done at the command line of source.its-is.uni.edu so
>> local on the machine where the repository resides..
>
> Please remember to use Reply To All.
>
> In this scenario, is it looping back, or is it still making a trip out
> to the network? Do you have any kind of packet filtering going on?
>
> HTTP access is the slowest of the 3 access methods, but that's a
> *huge* difference which doesn't make a lot of sense to me. Someone
> else suggested directory access controls & if there's a lot or they're
> particularly complex, that may be taking a performance toll as well.
There aren't any access controls really, if you have access to the
repository, you can commit, so valid user is required for access no
other controls.
I tried svn+ssh and it was really fast too. I may just switch my
developers to that, but conceptually I liked the http access methods..
Troy
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Jun 14 17:03:20 2007