You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by georg <ge...@w3x.org> on 2005/11/15 20:17:04 UTC

2nd+ co too slow

Hello!

I set up a Gentoo Linux Box with subversion 1.2.3, running svnserve on a
custom port (9999). I added a new repository (standard db), which
contains 3 empty dirs: trunk, branches, tags

On a second machine (same net, no fw), I try to checkout:
	svn co svn://192.168.1.5:9999/swwe/trunk

xinetd spawns 2 svnserve processes, and the checkout works as expected
(with authorisation), no svnserve process keeps running.
(btw: The problem occurs in daemon mode _and_ with xinetd.)

Now the trouble starts:
I delete the checked out trunk, and make sure that there is no remaining
.svn directory, etc.
Finally I issue the same checkout command again, and svn hangs. However,
the trunk directory is already checked out. On the server side there are
again the 2 svnserve processes.

If I delete ~/.subversion/auth/* the authentication takes place as
normal, but afterwards svn also hangs...

After several minutes (15-30) passed, the checkout gets finished.
Everything is as expected.

When I reboot the server, the first checkout works, then the same
problem starts again...

Thanks in advance!

georg

-- 
We could change the world if god gave us the source.


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: 2nd+ co too slow

Posted by Ryan Schmidt <su...@ryandesign.com>.
On Nov 16, 2005, at 12:34, georg@w3x.org wrote:

>> I set up a Gentoo Linux Box with subversion 1.2.3, running  
>> svnserve on a
>> custom port (9999). I added a new repository (standard db), which
>> contains 3 empty dirs: trunk, branches, tags
>>
>> On a second machine (same net, no fw), I try to checkout:
>> 	svn co svn://192.168.1.5:9999/swwe/trunk
>>
>> xinetd spawns 2 svnserve processes, and the checkout works as  
>> expected
>> (with authorisation), no svnserve process keeps running.
>> (btw: The problem occurs in daemon mode _and_ with xinetd.)
>>
>> Now the trouble starts:
>> I delete the checked out trunk, and make sure that there is no  
>> remaining
>> .svn directory, etc.
>> Finally I issue the same checkout command again, and svn hangs.  
>> However,
>> the trunk directory is already checked out. On the server side  
>> there are
>> again the 2 svnserve processes.
>>
>> If I delete ~/.subversion/auth/* the authentication takes place as
>> normal, but afterwards svn also hangs...
>>
>> After several minutes (15-30) passed, the checkout gets finished.
>> Everything is as expected.
>>
>> When I reboot the server, the first checkout works, then the same
>> problem starts again...
>
> some additions:
> - anon access works without problems: read+write (!) is it an auth
> problem?!
> - when i delete the repos and create a new one, i do not need to  
> reboot
> - problem exists with fsfs and bdb
> - local co does work normal (-> auth problem)


Could you be running out of entropy? See for example this discussion:

http://svn.haxx.se/users/archive-2005-02/0828.shtml

Check if your APR is using /dev/random (which may be slow) or /dev/ 
urandom (which should always be fast), and recompile it with the  
latter if necessary.

This may not be it though, since your problem is so predictable.  
Usually entropy problems occur more randomly.



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: 2nd+ co too slow

Posted by "georg@w3x.org" <ge...@w3x.org>.
some additions:
- anon access works without problems: read+write (!) is it an auth
problem?!
- when i delete the repos and create a new one, i do not need to reboot
- problem exists with fsfs and bdb
- local co does work normal (-> auth problem)
- i'm getting really depressed

any suggestions?

thx, georg

Am Die, den 15.11.2005 schrieb georg um 21:17:
> Hello!
> 
> I set up a Gentoo Linux Box with subversion 1.2.3, running svnserve on a
> custom port (9999). I added a new repository (standard db), which
> contains 3 empty dirs: trunk, branches, tags
> 
> On a second machine (same net, no fw), I try to checkout:
> 	svn co svn://192.168.1.5:9999/swwe/trunk
> 
> xinetd spawns 2 svnserve processes, and the checkout works as expected
> (with authorisation), no svnserve process keeps running.
> (btw: The problem occurs in daemon mode _and_ with xinetd.)
> 
> Now the trouble starts:
> I delete the checked out trunk, and make sure that there is no remaining
> .svn directory, etc.
> Finally I issue the same checkout command again, and svn hangs. However,
> the trunk directory is already checked out. On the server side there are
> again the 2 svnserve processes.
> 
> If I delete ~/.subversion/auth/* the authentication takes place as
> normal, but afterwards svn also hangs...
> 
> After several minutes (15-30) passed, the checkout gets finished.
> Everything is as expected.
> 
> When I reboot the server, the first checkout works, then the same
> problem starts again...
> 
> Thanks in advance!
> 
> georg


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org