On Tue, Nov 13, 2012 at 5:36 AM, Roberto De Ioris <[email protected]> wrote:

>
> > On Tue, Nov 13, 2012 at 4:29 AM, Roberto De Ioris <[email protected]>
> > wrote:
> >
> >>
> >> > Last update, after downgrading to newrelic-1.1.0.192 this (so far)
> >> made
> >> > the
> >> > bug disappear - and I have forwarded all this to their support guys to
> >> > hopefully get it fixed.
> >> >
> >> > Roberto - if possible, could we make uWSGI show the full traceback for
> >> > problems like this in the future, rather than just a single line of
> >> debug?
> >> > (it took about an hour of looking through every __iter__() method to
> >> find
> >> > this problem lol).
> >> >
> >> > Thanks!
> >>
> >>
> >> Not tested, but if i remember correctly you can set it with
> >>
> >> import sys
> >> sys.tracebacklimit = N
> >>
> >> where N is the depth.
> >>
> >
> > Thanks!
> >
> >
> >>
> >> So basically you get the Popen on thread bug, only in newrelic ? Or you
> >> get it everytime ?
> >>
> >
> > Happens every time, no matter if new relic is enabled or not.
>
> It should be that one (as you already found):
>
> http://bugs.python.org/issue14308
>
> Does the django_wsgi module has some code generating threads ?
>

Not as far as I'm aware, but subprocess/popen was being called when the
problem happened obv.


>
> If the bug is triggered by the usage of fork() (as in the issue14308) you
> should be able to fix it adding --lazy-apps. In that way the code in
> django_wsgi will be reset in each worker (this could break uwsgi signals,
> but we have workaround for that too)
>

Thanks for the detailed response Roberto :)

Typically we *never* run threads directly from uWSGI (we always use a
proper queuing system) and in the case of SimpleAES, we weren't aware it
was using Popen.


> --
> Roberto De Ioris
> http://unbit.it
> _______________________________________________
> uWSGI mailing list
> [email protected]
> http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
>
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi

Reply via email to