Hi,
I've been looking at Zenoss recently and like what I see. It could
be a good fit to replace our existing Nagios system.

However, the lack of layer2 network dependencies is pretty much a
show-stopper for us. I understand that it is being worked on and I
understand that it is a tricky problem to solve as there are so
many variables and different situations you have to detect and model.

I have two suggestions :-

[*] Allow setting of dependencies manually. Yes, it's not as sexy
as automated discovery, but it is easy to implement, it doesn't
have to be mandatory (i.e. manual 'parent' settings can override or
augment auto-discovery data for a device). For those who Zenoss
works well for now, they don't have to set any overrides. Even when
you get automated Layer 2 discovery working, is it likely to work
in all situations? Discovering which devices are behind a switch is
probably very different to determining which physical host a
virtual machine resides on for instance.

[*] Expose the dependency information to the administrator. While
graphical maps would be a 'nice to have', they are not necessary.
It is reassuring for an administrator simply to be told textually
that host 'x' has been detected as being behind router 'y'.

Am I missing something fundamental to the way Zenoss works as to
why these can't be done? Or is there a way of doing so outside of
the GUI?

------------------------
Dave Sexton







_______________________________________________
zenoss-dev mailing list
zenoss-dev@zenoss.org
http://lists.zenoss.org/mailman/listinfo/zenoss-dev

 

 

Thanks for the response,

Nagios takes the exact opposite approach from Zenoss in this
regard. All such dependencies are manual. They are built up as a
tree with the Nagios server as the root. Each monitored host
becomes a node on the tree and you explicitly state what each
host's parent nodes are.

For example, we have a server that is plugged into a monitored
switch, so I set the switch as the parent of the server (and other
hosts), then a router as the perent of the switch (and other hosts/
switches) and finally the nagios server is the parent of the router.

If we lose total connectivity to the switch, we know not to flood
the server admins with alerts about their server (though we may
just drop them a single one as a notification).

------------------------
Dave Sexton







_______________________________________________
zenoss-dev mailing list
zenoss-dev@zenoss.org
http://lists.zenoss.org/mailman/listinfo/zenoss-dev

 

 

edahl wrote:

 

 

Clearly with a
big network you can't manually build a map of the network! :)

I beg to differ, there are many really large Nagios installations
out there that either do not use dependencies or must have done it
manually. The way it's implemented is just as an additional field
in the host definition. If your network is adequately documented,
the information is easy to find at the time you add the node into
the monitoring system.

Admittedly, Nagios does not cope well with dependency 'loops', as
I believe that Zenoss does. I think a good work-around would be to
do automated discovery as best as possible but allow the
administrator the ability to manually override this on a host-by-
host basis.

------------------------
Dave Sexton







_______________________________________________
zenoss-dev mailing list
zenoss-dev@zenoss.org
http://lists.zenoss.org/mailman/listinfo/zenoss-dev