Re: [Xen-users] Monitoring Xen via Nagios

[Simon Hobson <linux@xxxxxxxxxxxxxxxx>]

Jan Vejvalka <jan.vejvalka@xxxxxxxxxxxxxxx> wrote:

> IMHO, SNMP is more standardized than NRPE, and
> Nagios/NRPE was easier for me to run than Nagios/SNMP, because of the
> (mostly mental, I think) overhead of the SNMP infrastructure. OTOH,
> SNMP is more standardized than NRPE and many things are ready to use
> (e.g. detailed openvswitch monitoring, as Simon wrote) just that by
> quick g**gling I can't find anything useful (for me) about SNMP and
> Xen... more than https://github.com/cvarta/xen-monitoring .

I'd agree with that.
To monitor something by SNMP you "only" need the thing being monitored to have 
SNMP installed and working - and it can be used regardless of what that "thing" 
is. But, it's quite a lot to get your head around.

NRPE is more specialised in that it only supports things you can install it on. 
However, configuring it is fairly easy as you'll have learned most of it while 
setting up Nagios itself.


Just to add another complication. I include it for completeness, and strongly 
recommend not trying it until you have SNMP polling under control.
With SNMP you have the option of setting traps. Here you can configure checks 
on the device being monitored - so for a UPS it might be "if the mains fails, 
raise an SNMP trap". These can be sent to the system running Nagios, and via 
another layer of configuration that largely duplicated the polling setup, you 
can have the trap event alter the status without waiting for Nagios to poll.
So taking that UPS example.
Basic polling will tell you some period after the event that the mains power 
has failed. The period may be "almost nothing" or it may be the maximum polling 
interval - so Nagios might not react for several minutes. There's a tradeoff 
here - you can poll very frequently but that increases load which may be a 
factor with a lot to monitor. Or you can poll infrequently to keep the load 
down, but then not know about an event for some time after it happens. This is 
one of those things you need to tune to suit your requirements, approach to 
risk, etc.
If you configure a trap on the UPS, and configure Nagios to process it, then 
you can keep your normal load down, while reacting quickly to important events.
But as I say, it's another load of config, which like most things SNMP, is "not 
trivial".


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
https://lists.xen.org/xen-users