Overall it works like a

> > charm, but sometimes (eg. 1/30) discovery fails. 
>
> does mco ping exhibit the same behavior if you run it often?  Does it
> tend to happen more after a period of the collective being idle or just
> really randomly?
>
Am facing the same problem where the discovery fails (1/4). I have a setup 
with 
a network of 2 activemq brokers and the client and nodes fails over from 
one to other.
The mco ping also exhibits the same problem, mostly after beeing idle and 
when one of the broker 
boxes is heavily loaded. I have nodes with 4000, 5000 ms ping and with 60, 
70 ms ping 
but when i run mco ping with default i get no nodes discovered. Atleast it 
should discover the
low ping boxes right? 
*I have configured fail over as in the documentaion but yet i find these 
discovery problems especially when*
*one of the brokers is loaded. Can mcollective to tuned to use the less 
loaded broker ?*
 

> > 
> > I would like to ask for some feedback on the following ideas, that
> > could fix this problem.
> > 
> > a) Increase discovery timeout
> > mco offers a option to tweak the discovery timeout. What is your
> > experience with increasing this value? When running
> > "mco ping", I see ping times of 130ms, so 2 seconds (the default
> > should be enough), or?
> > Is there a way to configure is global?
>
> it's not global - its a client setting but with those ping times it 
> should be sufficient.  Discovery does exactly what mco ping does so its
> a good way to diagnose
>
by this documentation it shows its a server config  
http://docs.puppetlabs.com/mcollective/reference/basic/configuration.html

when i add the command line option --dt 5 it works fine. I dono why the low 
ping nodes
are not discovered in the default discovery timeout. I find this option 

export MCOLLECTIVE_EXTRA_OPTS="--dt 5 --timeout 3 --config 
/home/you/mcollective.cfg"

*Is there a way to configure default discovery timeout for the client in 
the client.cfg?*

Can anyone clarify me on this whole discovery timeout thing?
 
 

> Might be worth enabling verbose gc logging on your activemq its possible
> that during these times it just did a big full garbage collection which
> would block it and that might indicate some tuning is needed
>
> > 
> > b) Mco should exit != 0 when no nodes are found
> > I would like to see a "--batch" or "--non-interactive" mode, where
> > mco has a exit code different from 0, when no nodes
> > are found.
>
> ok, you can file tickets for this
>
I haven't setup any mcollective monitoring, Is checking if mco ping works 
fine 
enough ?
 

> > c) Add "expected count" to mco command
> > I thing there are some situation, where one knows the number of
> > MCollective nodes. So what about adding a options
> > "--expect-number" option to mco, where I can either give a count or
> > range of expected nodes.
>
> mcollective 1.3.x which will soon become 2.0 have a new mode of 
> communications
> where you can provide it a host list etc and it will bypass discovery, this
> is ment to be used for things like deployers where you know what machines 
> you
> wish to affect and it will probably help
>
Am using the direct addressing stuff and i works just fine

 - activemq doing long full garbage collections
>  - network is interrupted after long periods of idle time
>  - activemq is idle for a long time and was swapped etc
>  - you have very busy machines that do not respond at all - unlikely in 
> your case
>
> there are probably other reasons too but these are the rough likely causes.
> Amazon has a pretty aggressive idle connection timeout though so you might
> enable registration just to keep the stomp connections from being idle too 
> long
>

Mine is a data centre setup, this is my first experience with message 
brokers so can 
i get some ideas in fine tuning activemq or ll using rabbitmq solve any of 
the problems? 
 
 

-- 
You received this message because you are subscribed to the Google Groups 
"Puppet Users" group.
To view this discussion on the web visit 
https://groups.google.com/d/msg/puppet-users/-/VRlFkZKguGcJ.
To post to this group, send email to puppet-users@googlegroups.com.
To unsubscribe from this group, send email to 
puppet-users+unsubscr...@googlegroups.com.
For more options, visit this group at 
http://groups.google.com/group/puppet-users?hl=en.

Reply via email to