On Fri, Nov 18, 2011 at 8:20 PM, Mykola S. Grechukh <gns@altlinux.org> wrote:

Is it really impossible for Poller to have own named pipe ?

I'm doing research of how to migrate big distributed nagios/check_mk
system to Shinken distributed cluster.  Check_mk runs itself as single
active check per host, which in turn provide service states to nagios
command pipe as passive checks.

So it seems that i have to run Receiver near every Poller instance. It
could be a problem from point of consistency and fault tolerance -
i.e. if receiver dies, scheduler continues to assign check-mk tasks to
this poller :(

Quite easy to workaround it, although...
Yes, the reveiver is a way of course, but can be indeed a problem if the receiver die when the poller stay up.

I added in the sources to the CommandFile module a way to be load in the poller daemon, so you can just add this module to your distant poller with this.

But I'm wondering how this passive way will scale. Shinken is done for huge active checks, and the check_mk way is an hack to avoid Nagios low active way by using a "less low" passive one. When you test it, look at the arbiter activity. I think for a middle size installation there will be no problems at all, but for a large one it can be a problem and we will have to find solutions to enhance this.

A perfect way can be to wrote a check_mk poller module of course, so there will be no more "passive" way for all of this, but it can take some times to get it working (but check_mk is in Python, so it's definitively possible :) ).



Mykola Grechukh

CTO, Satelliz-Ukraine

All the data continuously generated in your IT infrastructure
contains a definitive record of customers, application performance,
security threats, fraudulent activity, and more. Splunk takes this
data and makes sense of it. IT sense. And common sense.
Shinken-devel mailing list