Sorry, I haven't had a chance to try this patch out.

While not ideal, the static number of threads isn't as bad as I initially 
thought. When I first tried it, ulimit -u was too low (the default, 1024) to 
support the number of threads specified (5000) and the machine went bonkers 
and triggered a watchdog reboot. This happened a few times before I figured out 
what was wrong. Once I upped the ulimit for user processes, things worked much 
better.

At both low and heavy load, we see a pretty constant ~330 megs of resident 
memory with 5000 static threads and negligble CPU usage. To me, this isn't a 
big deal.

Just a suggestion, but it might be a good idea to up the ulimit before 
dropping root, since the required max can easily be determined.

- Neil

On Monday 19 March 2012 11:21:07 Joe Gooch wrote:
> True, stock 2.6 doesn't have this, but it's not all that hard to implement.
> 
> Try this patch, and set ThreadModel dynamic in your config.
> http://goochfriend.org/pound/pound_2.6_threadmodels.patch
> 
> Joe
> 
> > -----Original Message-----
> > From: Leo [mailto:[email protected]]
> > Sent: Monday, March 19, 2012 3:53 AM
> > To: [email protected]
> > Subject: Re: [Pound Mailing List] Pound threading model in 2.6
> > 
> > Hello Neil,
> > 
> > AFAIK there is no possibility to get the old threading model in Pound
> > 2.6. You have to determine the maximum number of threads you will need
> > at peak time (in your case 20,000 threads) and set it in your pound
> > config.Pound will start all these threads at startup then :-(
> > 
> > I have already complained about this behaviour
> > (http://www.apsis.ch/pound/pound_list/archive/2012/2012-
> > 01/1326878271000#1326878271000)
> > and some of my suggestions have been added to the feature list for
> > Pound
> > 2.7
> > (http://www.apsis.ch/pound/pound_list/archive/2012/2012-
> > 02/1328374222000#1328374222000)
> > 
> > Hope this helps ...
> > 
> > Regards,
> > Leo
> > 
> > On 03/16/2012 07:23 PM, Neil Skrypuch wrote:
> > > I was wondering if there is any way to get the Pound 2.5 behaviour
> > 
> > for
> > 
> > > threads in Pound 2.6+.
> > > 
> > > We have a somewhat atypical setup involving Pound where the old
> > > behaviour was strongly preferable. Basically, we handle large volumes
> > > of long lived connections (think Websockets). With Pound 2.5, we had
> > > no trouble supporting over 10,000 concurrent connections, whereas
> > > using the default Pound 2.6 setup things fell over around 110.
> > > 
> > > Now, I can bump up the Threads value to 10,000 or 20,000 (or some
> > > other equally large number), and this works, but this means that
> > 
> > Pound
> > 
> > > always has
> > > 10,000 or 20,000 threads running, which isn't very nice to the
> > 
> > system.
> > 
> > > I should point out that we don't normally have 20,000 concurrent
> > > connections, this is an unusual event, but we do need to support it.
> > > 
> > > So, is there any way to get back the old threading behaviour, or do
> > 
> > we
> > 
> > > just have to tough it out and always run a huge number of Pound
> > 
> > threads?
> > 
> > > - Neil
> > > 
> > > --
> > > To unsubscribe send an email with subject unsubscribe to
> > 
> > [email protected].
> > 
> > > Please contact [email protected] for questions.
> > 
> > --
> > To unsubscribe send an email with subject unsubscribe to
> > [email protected].
> > Please contact [email protected] for questions.
> 
> --
> To unsubscribe send an email with subject unsubscribe to [email protected].
> Please contact [email protected] for questions.

--
To unsubscribe send an email with subject unsubscribe to [email protected].
Please contact [email protected] for questions.

Reply via email to