[op5-users] question about merlin poller/noc setup

Andreas Ericsson ae at op5.se
Mon Jan 18 10:20:05 CET 2010


On 01/12/2010 08:36 PM, Christian Anton wrote:
> Hello!
> 
> I am trying to use merlin to set up a poller/noc setup and i have
> several questions:
> 
> Is my understanding right every of the nagios boxes will have its own
> database? Or is it sufficient for the pollers having merlin installed
> _without_ an own database? Having multiple databases, will all databases
> contain the same data? Or are only databases of peered nodes equal?
> 

You only need a database on the servers where you want to run Ninja. The
database will contain info on all hosts. If you're using two peers in a
loadbalanced/redundant fashion, they will both contain the same info,
with a microscopic delay between updates (the time it takes to send, read
and insert a record on the other peer).

> In a poller/noc setup, must both machines have the exact same
> configuration files? In my case the poller would serve a vpn-connected
> location. If poller goes down there will be no possibility (routing) to
> check these services from the central machine.
> 

In a poller/noc setup, the noc must have the hosts and services configured
that the poller has, but the poller doesn't need to have the hosts and
services it shouldn't check. If checks will fail when run from the noc,
you can disable the checks there, but in that case you really should
check the connectivity to the poller instead. A plugin parsing the netstat
output and possibly also checks update-time of one or more of the poller-
monitored hosts should suffice to make sure you get notified if the
poller goes to lunch.

> In my tests with the "next" head merlin installed on CentOS5/nagios
> 3.2.0 on two hosts i had to install only the hosts which had to be
> monitored by the peer to the peer's configuration files, otherwise the
> peer and the noc did loadbalancing on the hosts which are not included
> in the hostgroup which is "delegeted" to the poller.
> 
> Example:
> NOC has configs for host1, host2, host3 while host2 and host3 are in the
> hostgroup "remotehosts". This hostgroup is set up in merlin.conf to be
> delegated to the poller.
> 
> POLLER has configs only for host2 and host3, and also has them in a
> hostgroup "remotehosts". If this node's configuation contained host1
> also, load-balancing would be done for host1.
> 

Yes, that's the plan, sort of. A poller that shouldn't check a host
shouldn't have that host configured, or checks for it enabled.

> This setup worked for me but changing configuration or sometimes simply
> restarting the nagios daemon on one side resulted in a crash of the
> nagios daemon on the other side. I would like to know if that is because
> of my configs or if the poller/noc architecture is still a bit unstable
> in merlin.
> 

The poller/noc configuration is still a bit unstable. I wish I could say
we're working on that right now, but other things (such as completing the
missing pieces of Ninja) have taken priority here at the office. Patches
are ofcourse welcome. Merlin is scheduled to be complete and 100% stable
in september this year. The current functionality will most likely be
stable around march or april this year, unless we get very detailed bug-
reports (including backtrace printouts) from users that make the bughunt
really quick for us, in which case we'll divert some work-efforts from
Ninja into Merlin to make some quick fixes.

-- 
Andreas Ericsson                   andreas.ericsson at op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

Considering the successes of the wars on alcohol, poverty, drugs and
terror, I think we should give some serious thought to declaring war
on peace.


More information about the op5-users mailing list