[pmwiki-users] Help! Performance problem after server move

Patrick R. Michaud pmichaud at pobox.com
Wed Apr 5 11:02:25 CDT 2006


On Wed, Apr 05, 2006 at 12:00:35PM -0400, Robin Hills wrote:
> Since I started this thread, it would be nice if I could report a 
> specific resolution to the problem but I can't. We're now on a 
> completely different server running on Windows 2003. I have some issues 
> to look at but we've had 1400 visitors in ten hours and the lights are 
> still on so I'm in much better shape than I was. Except that I really 
> need some sleep and a day off.

I'm very glad to hear the site is running again.  I did a lot of
work yesterday looking at ways to potentially speed up PmWiki itself
but didn't come across anything specific.

> I reported back to the previous host that it might be an issue related 
> to Apache. I doubt that I'll hear anything more about it.

Out of curiosity, how many pages do you have on your site?

Pm



> Patrick R. Michaud wrote:
> 
> >On Tue, Apr 04, 2006 at 11:19:33PM +0200, Joachim Durchholz wrote:
> > 
> >
> >>Patrick R. Michaud schrieb:
> >>   
> >>
> >>>I'm fairly 
> >>>convinced it's not directly PmWiki's fault, because it happens very
> >>>inconsistently.  For more details -- see my message at
> >>>http://www.pmichaud.com/pipermail/pmwiki-users/2006-March/025470.html.
> >>>     
> >>>
> >>Now that "happens very inconsistently" triggered something in my mind.
> >>
> >>I know of one thing that happens at random intervals in PHP: session GC. 
> >>With the standard settings, it is run with a 1% probability - very 
> >>difficult to reproduce any problems with that.
> >>[...]
> >>   
> >>
> >
> >I'm pretty certain the problem isn't session GC.  Although session GC
> >does occur randomly (with a 1% probability), when it does occur
> >it always happens at session initialization time -- i.e., when
> >session_start() is executed.  In PmWiki (on pmwiki.org), this
> >would have to be during calls to RetrieveAuthPage.
> >
> >The testing I've been doing is pretty fine grained--well below
> >the function-call level.  So, if the delays were consistently
> >occurring as part of session initialization (i.e., during
> >page authorization), I could narrow it down pretty quickly
> >to that.  But that's not what I'm seeing -- the delays occur
> >at different times in pmwiki processing.
> >
> >Also, I see the slow processing times even when no session
> >handling is taking place.  So, I'm pretty sure it must
> >be something else.
> >
> >Pm
> >
> >_______________________________________________
> >pmwiki-users mailing list
> >pmwiki-users at pmichaud.com
> >http://host.pmichaud.com/mailman/listinfo/pmwiki-users
> >
> > 
> >
> 




More information about the pmwiki-users mailing list