Re: gkrellmd problem

Hi,

thanks for your answer. The problem has gone for some time, but it's back 
again. 

Now to you questions:

What leaking means is shown in a screen shot I attached. All of the upper 
computer are running at 100% on all cores. But it seems that the client 
doesn't recieve information continuously.

A reconnection of the client doesn't resolve the problem. A new connection to 
the server is possible but take a while.

A restart of the daemon solves the problem. But I realy don't now what must 
happen to reproduce this behaviour. It's seems, that this happen when to many 
people connect and reconnect ovre the time.

Bye,
Andreas

Am Samstag, 31. Dezember 2011, 16:26:14 schrieb Stefan Gehn:
> Hi,
> 
> On 12/09/2011 10:09 AM, Andreas Hauffe wrote:
> > Dear Gkrellm developer,
> > 
> > we have a problem with gkrellmd. We use gkrellm to check if a computer
> > idles and can be used for calculations. On each computer gkrellm is
> > startet with: /usr/bin/gkrellmd --pidfile /var/run/gkrellmd.pid --user
> > nobody --group nogroup --max-clients 50 --update-hz 4 --port 19150
> > --allow-host 192.0.0.* --io-timeout 0 --reconnect-timeout 0 -d
> Disabling all timeouts is probably not a good idea, clients will never
> reconnect in case of network problems.
> 
> > Then up to six or seven people open there gkrellm client and connect to
> > that server (gkrellmd). After some time the gkrellmd server seems to
> > leak. For up to 30 second the client will get no information. Atfer
> > data is send to the client for about 5 seconds and so on.
> What are you referring to with "leak"? Does the gkrellmd process consume
> more memory over time or does it cause any high cpu-usage in this scenario?
> 
> > What could be the problem? A reconnection of the client doesn't resolve
> > the problem. Only a restart of the daemon.
> Can new clients actually connect to the server or will it stop
> accepting new connections?
> 
> 
> In theory, clients blocking eachother should not happen, all
> client-connections are set to non-blocking mode. I have started working
> on porting gkrellmd to GIO-based networking but the code hasn't been
> published yet. If you're willing to build gkrellmd yourself I can push
> the current work in January so you can try out wether it improves your
> situation.
> 
> Bye,
> Stefan Gehn
> 
> ______________________________________________
> Gkrellm mailing list
> [email protected]
> Read archives-join-leave-set RSS feed:
> http://archive.lists.netservicesgroup.com
> This service is provided by The Network Services Group:
> http://www.netservicesgroup.com
-- 

----------------------------------------------------------------------------------------------------
Technische Universität Dresden
Institut für Luft- und Raumfahrttechnik / Institute of Aerospace Engineering
Lehrstuhl für Luftfahrzeugtechnik / Chair of Aircraft Engineering

D-01062 Dresden
Germany

phone : (++49)351 463 38496
fax : (++49)351 463 37263
mail : [email protected]
Website : http://tu-dresden.de/mw/ilr/lft
----------------------------------------------------------------------------------------------------