You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by ge...@printsoft.com on 2006/06/19 11:35:27 UTC
More on possible client authentication
Hi all (again)
This is addition to an earlier email regarding what appears to be long
authentication times for SVN clients.
I built another debian sarge system at home, and have absolutely not issues with
it. Is there a possibility that by running Subversion on a User Mode Linux
system could cause connection issues with clients?
Is there a detailed list of supporting services and runtime libraries that, if
missing, could caase delays in authentication users between the subversion
client and server?
Thanks one again - I hope to hear/read from somebody soon
Regards,
Gerhard
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: More on possible client authentication
Posted by ge...@printsoft.com.
Quoting Ryan Schmidt <su...@ryandesign.com>:
> On Jun 20, 2006, at 01:40, Gerhard Hannemann wrote:
>
> >> One possibility would be that it's taking a long time to generate
> >> random bytes. If you machine doesn't have a suitable source of
> >> entrypy, /dev/random could be quite slow. You might try recompiling
> >> APR and telling it to use /dev/urandom (via the --with-devrandom flag
> >> to configure).
> >
> > I thought APR is only relevant when using apache and access
> > repositories
> > using http.
> > Currently I only access the repository using svn and file.
> > It APR or random number generator device still relevant in this case?
>
> Yes, the advice is still relevant. The Apache Portable Runtime (APR)
> is not related to the Apache web server, other than that they're both
> made by the same group, and that the Apache web server uses APR.
> Subversion also uses APR, always, regardless of what repository
> access method is being employed. As I understand it, APR gives you a
> common interface to things which typically vary between operating
> systems, and since Subversion is supposed to run on many operating
> systems, they decided to use APR rather than reinvent several wheels.
>
>
>
>
Thanks for the information.
I will rebuild APR with the sugested settings and let the list know if this
improves things.
Regards,
Gerhard
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
RE: More on possible client authentication
Posted by Gerhard Hannemann <ge...@printsoft.com>.
Hi all,
Success!
The subversion server has behaved well so far.
It appears some extra little steps are required when running Subversion on
User Mode Linux.
Regards,
Gerhard
-----Original Message-----
From: Gerhard Hannemann [mailto:gerhard.hannemann@printsoft.com]
Sent: Wednesday, 21 June 2006 5:19 PM
To: 'Ryan Schmidt'
Cc: users@subversion.tigris.org
Subject: RE: More on possible client authentication
Hi,
I reconfigured the apr section or the subversion source using:
./configure --with-devrandom=/dev/urandom
I found that I also had to turn on the urandom daemon.
So far so good - no hang-ups using the previous tests that originally caused
the symptoms in the svn client.
I will run this for the next 24 hours or so and report back.
Thanks for all the help.
Regards,
Gerhard
-----Original Message-----
From: Ryan Schmidt [mailto:subversion-2006q2@ryandesign.com]
Sent: Tuesday, 20 June 2006 5:54 PM
To: Gerhard Hannemann
Cc: users@subversion.tigris.org
Subject: Re: More on possible client authentication
On Jun 20, 2006, at 01:40, Gerhard Hannemann wrote:
>> One possibility would be that it's taking a long time to generate
>> random bytes. If you machine doesn't have a suitable source of
>> entrypy, /dev/random could be quite slow. You might try recompiling
>> APR and telling it to use /dev/urandom (via the --with-devrandom flag
>> to configure).
>
> I thought APR is only relevant when using apache and access
> repositories
> using http.
> Currently I only access the repository using svn and file.
> It APR or random number generator device still relevant in this case?
Yes, the advice is still relevant. The Apache Portable Runtime (APR)
is not related to the Apache web server, other than that they're both
made by the same group, and that the Apache web server uses APR.
Subversion also uses APR, always, regardless of what repository
access method is being employed. As I understand it, APR gives you a
common interface to things which typically vary between operating
systems, and since Subversion is supposed to run on many operating
systems, they decided to use APR rather than reinvent several wheels.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
RE: More on possible client authentication
Posted by Gerhard Hannemann <ge...@printsoft.com>.
Hi,
I reconfigured the apr section or the subversion source using:
./configure --with-devrandom=/dev/urandom
I found that I also had to turn on the urandom daemon.
So far so good - no hang-ups using the previous tests that originally caused
the symptoms in the svn client.
I will run this for the next 24 hours or so and report back.
Thanks for all the help.
Regards,
Gerhard
-----Original Message-----
From: Ryan Schmidt [mailto:subversion-2006q2@ryandesign.com]
Sent: Tuesday, 20 June 2006 5:54 PM
To: Gerhard Hannemann
Cc: users@subversion.tigris.org
Subject: Re: More on possible client authentication
On Jun 20, 2006, at 01:40, Gerhard Hannemann wrote:
>> One possibility would be that it's taking a long time to generate
>> random bytes. If you machine doesn't have a suitable source of
>> entrypy, /dev/random could be quite slow. You might try recompiling
>> APR and telling it to use /dev/urandom (via the --with-devrandom flag
>> to configure).
>
> I thought APR is only relevant when using apache and access
> repositories
> using http.
> Currently I only access the repository using svn and file.
> It APR or random number generator device still relevant in this case?
Yes, the advice is still relevant. The Apache Portable Runtime (APR)
is not related to the Apache web server, other than that they're both
made by the same group, and that the Apache web server uses APR.
Subversion also uses APR, always, regardless of what repository
access method is being employed. As I understand it, APR gives you a
common interface to things which typically vary between operating
systems, and since Subversion is supposed to run on many operating
systems, they decided to use APR rather than reinvent several wheels.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: More on possible client authentication
Posted by Ryan Schmidt <su...@ryandesign.com>.
On Jun 20, 2006, at 01:40, Gerhard Hannemann wrote:
>> One possibility would be that it's taking a long time to generate
>> random bytes. If you machine doesn't have a suitable source of
>> entrypy, /dev/random could be quite slow. You might try recompiling
>> APR and telling it to use /dev/urandom (via the --with-devrandom flag
>> to configure).
>
> I thought APR is only relevant when using apache and access
> repositories
> using http.
> Currently I only access the repository using svn and file.
> It APR or random number generator device still relevant in this case?
Yes, the advice is still relevant. The Apache Portable Runtime (APR)
is not related to the Apache web server, other than that they're both
made by the same group, and that the Apache web server uses APR.
Subversion also uses APR, always, regardless of what repository
access method is being employed. As I understand it, APR gives you a
common interface to things which typically vary between operating
systems, and since Subversion is supposed to run on many operating
systems, they decided to use APR rather than reinvent several wheels.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
RE: More on possible client authentication
Posted by Gerhard Hannemann <ge...@printsoft.com>.
Thanks for your reply.
I thought APR is only relevant when using apache and access repositories
using http.
Currently I only access the repository using svn and file.
It APR or random number generator device still relevant in this case?
I vaguely remember seeing some messages about the random number generator
when the UML starts up.
Thanks once again for your help
Regards,
Gerhard Hannemann
-----Original Message-----
From: rooneg@gmail.com [mailto:rooneg@gmail.com] On Behalf Of Garrett Rooney
Sent: Tuesday, 20 June 2006 12:28 AM
To: gerhard.hannemann@printsoft.com
Cc: users@subversion.tigris.org
Subject: Re: More on possible client authentication
On 6/19/06, gerhard.hannemann@printsoft.com
<ge...@printsoft.com> wrote:
>
>
> Hi all (again)
>
> This is addition to an earlier email regarding what appears to be long
> authentication times for SVN clients.
>
> I built another debian sarge system at home, and have absolutely not
issues with
> it. Is there a possibility that by running Subversion on a User Mode
Linux
> system could cause connection issues with clients?
>
> Is there a detailed list of supporting services and runtime libraries
that, if
> missing, could caase delays in authentication users between the subversion
> client and server?
>
> Thanks one again - I hope to hear/read from somebody soon
One possibility would be that it's taking a long time to generate
random bytes. If you machine doesn't have a suitable source of
entrypy, /dev/random could be quite slow. You might try recompiling
APR and telling it to use /dev/urandom (via the --with-devrandom flag
to configure).
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: More on possible client authentication
Posted by Garrett Rooney <ro...@electricjellyfish.net>.
On 6/19/06, gerhard.hannemann@printsoft.com
<ge...@printsoft.com> wrote:
>
>
> Hi all (again)
>
> This is addition to an earlier email regarding what appears to be long
> authentication times for SVN clients.
>
> I built another debian sarge system at home, and have absolutely not issues with
> it. Is there a possibility that by running Subversion on a User Mode Linux
> system could cause connection issues with clients?
>
> Is there a detailed list of supporting services and runtime libraries that, if
> missing, could caase delays in authentication users between the subversion
> client and server?
>
> Thanks one again - I hope to hear/read from somebody soon
One possibility would be that it's taking a long time to generate
random bytes. If you machine doesn't have a suitable source of
entrypy, /dev/random could be quite slow. You might try recompiling
APR and telling it to use /dev/urandom (via the --with-devrandom flag
to configure).
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org