Re: [munin-users] munin node randomly fails to connect to many nodes simultaneously

[Nicolai L. <ja...@la...>]

On 08/22/2011 07:13 PM, Lonni J Friedman wrote:
> On Mon, Aug 22, 2011 at 9:28 AM, Lonni J Friedman<net...@gm...>  wrote:
>> On Mon, Aug 22, 2011 at 7:56 AM, Steve Schnepp<ste...@gm...>  wrote:
>>> On Sun, Aug 21, 2011 at 00:47, Lonni J Friedman<net...@gm...>  wrote:
>>>> Are you referring to the max_processes parameter in munin.conf, or
>>>> something else?
>>>
>>> Yes.
>>
>> ok, i've set it to 8.  although I'm rather confused how this works.
>> The documentation doesn't state an explicit default value, only "as
>> many processes as necessary".  How does munin-update determine how
>> many are necessary?
>
> When set to 8, its increased from 190s to 255s.
> When set to 9, its taking 228s
>
> My bigger concern here is how reliably we can draw any conclusions
> from setting the value (especially when I don't understand what the
> default is, or how its being selected), since this failure only
> happens roughly once/week.

There is no inteligence in selecting the default which is "infinite", 
i.e. fork one process for each [node] configured.  Each process uses 
relatively little resources so most often this is all right.

  max_processes <value>
     This directive specifies the maximum number of processes to be used
     for gathering information from nodes.  If left blank, munin will use
     as many processes as necessary.  Affects: munin-update.

Nicolai