Hi,

Well....the workers always server requests. Think if is this way.  I have a
list that random servers pages test=[a,b,c] e.g. random.choice(test).


So...randomly, a page will be shown, either a,b,c for eash request.

Then I update the list to just show two test=[a,b].  Only a and b should
now be served.  C should stopped being served.  Yes..per the chart, there
is what I call leakage.  Somehow that page is still being served even
though I removed C from the list.  If I restart all servers.....problem
solved but dont want to do that if I dont have to.










On Sun, Dec 30, 2012 at 8:19 PM, Roberto De Ioris <[email protected]> wrote:

>
> > Hi,
> >
> > See the attached.  Note the abrupt stop..thats where I updated the hash.
> > I
> > have 4 servers running one gevent loop per core.  So..its clear that the
> > code does what its supposed to do...but I guess there is a runaway
> worker.
> > I dont know how to explain it.
> >
>
> If a whole worker lose the message you will not see such an abrupt stop
> (each core in the worker would still be able to manage requests), so i
> have to suppose a bunch (really few from the graph) of greenlets are still
> up answering requests.
>
> How each greenlet knows it should stop answering requests ? Or better,
> when a greenlet read that dictionary to stop itself ?
>
> --
> 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