I've added a patch to Python 2.1 on sourceforge that helped me considerably 
when diagnosing multi-threaded Python lock ups.  See 
http://sourceforge.net/tracker/index.php?func=detail&aid=421709&group_id=5470&atid=305470
 
for the submitted patch.

I created this to give me at least as much functionality as Java has to dump 
the stack trace of all Java threads and Java monitors.

For an example of how I used it see 
http://lists.zope.org/pipermail/zodb-dev/2001-May/000703.html in the ZODB dev 
list.

Hope this helps,

John


On Friday 18 May 2001 09:13, Christian Scholz wrote:
> Hi!
>
> Just FYI: The server still hangs once a day and another one is doing
> the same..
> It even does this when simply accessing one page or image over and over
> again.
> More precisely I've started the server and did
>
> ab -n 100000 -c 10 http://foo.bar/we/pics/someimg.gif
>
> (ab being the apache benchmarking utility)
>
> and after a while I get the broken pipe error.
>
> I've also changed my python version to 2.0 and tested it with this
> but still the same effect. Also changing the server (and thus
> hardware and system) did not help it.
> I've also removed the mysqlda from the zope installation but also this
> did not help..
> I will try now to create the smallest version of the site that still
> hangs..
>
> Dunno if anybody has some idea, I actually have none..
>
> I'll keep you informed..
>
> -- christian
>
> PS: strangely it works on the development server but also here I am doing
> some tests with ab right now.. and this one is still 1.5.2
>
> On Thu, May 03, 2001 at 02:16:30PM -0400, Chris McDonough wrote:
> > > Well, my problem might be that in my case most sql statement are done
> >
> > inside
> >
> > > some attribute provider of ZPatterns.. This is using ZSQL methods
> >
> > internally
> >
> > > but actually those then won't show up I guess..
> >
> > Even if you weren't using ZPatterns, the error probably wouldn't jump out
> > and say "here I am!"  So I don't think there's much difference between
> > using ZPatterns and not using ZPatterns.  The process of detecting when
> > something hangs is just like any other troubleshooting process, it's a
> > matter of exclusion.  If you notice that the request named
> > "foobargorf/fleafang" *always* hangs, you investigate what it does, and
> > try to reproduce it.  If it's incidental, so be it, and move on to the
> > next theory.
> >
> > > I've also seen that some more recent version of the mysql stuff is
> > > around and I am using this now..
> > >
> > > The problem's also that I just have a few methods to invoke from the
> >
> > outside
> >
> > > which do lots of things by calling other object. Thus I might not
> > > really
> >
> > see
> >
> > > what's really causing the problem.. I hope though that my upgrade will
> >
> > show some
> >
> > > benefit..
> >
> > The -M log perhaps won't show you the actual operation that's causing the
> > hang, but it will show you the entry point into a routine which causes
> > the hang.  It's your job from there to track down the cause.  This is
> > just like debugging a program.  You get an error somewhere, and you need
> > to track it back to its root, which may be six levels up the call stack
> > buried in some godforsaken regex.  ;-)

-- 
. . . . . . . . . . . . . . . . . . . . . . . .

John D. Heintz | Senior Engineer

1016 La Posada Dr. | Suite 240 | Austin TX 78752
T 512.633.1198 | [EMAIL PROTECTED]

w w w . d a t a c h a n n e l . c o m

_______________________________________________
Zope-Dev maillist  -  [EMAIL PROTECTED]
http://lists.zope.org/mailman/listinfo/zope-dev
**  No cross posts or HTML encoding!  **
(Related lists -
 http://lists.zope.org/mailman/listinfo/zope-announce
 http://lists.zope.org/mailman/listinfo/zope )

Reply via email to