Some suggestions / questions

[dom]

Hello,
i'm new to this forum, but not new to the supervision market, as I work for years now in this area with commercial products (OpenView, Tivoli, ...).
I'm currently starting to evaluate Zabbix (and Nagios), for our own activity (monitoring services, and IT services), and it already has very nice features, like integrating both the monitoring and reporting functions, and a nice agent !
So, first of all : congratulations for this really good work !

My goal would be to use it for "big" environments (our clients, or our own services) : several hundred of "machines" (servers, routers, ...), and a "client" organisation inside it (Groups of machines, User profiles, ...).
Then the important points for us are : security, distributed architecture (scalability), ability to manage globally.

Some suggestions / questions :
- It would be nice to have a centralized management of all agents (already seen in somes posts).
- Is it planned to have an autonomous agent. The principle would be to push the agent's configuration, and then work only by exception (no polling => traffic optimisation).
- Distributed monitoring : is it planned / possible to have a distributed management : local autonomous manager reporting to a centralized one, with a central configuration, and if possible no polling between the local and the central manager ?
- Redundancy : ability to have two Zabbix on a cluster or 2 machines ?
- Calculated indicators : one important point, is the ability to calcule consolidated indicator from simple ones. For example bandwidth from ifInOctets, ifOutOctets, and ifSpeed, or the availability of a machines group from the individual availability. Is this possible/planned ?
- Screens management : very interesting (ability to design screens). A way to organise the screens (Tree, Folders, ...), in a more global vision would be a great thing in a big environment where you could have hundreds of them.
- Group management : some improvements could be done, but I have to work more on Zabbix before saying anything interesting and precise enough.

[Alexei]

Thanks for your suggestions! I like them all.

A small note regarding clustering. I would suggest 2 servers 1 shared disk array approach with one virtual IP address for ZABBIX server (kind of HP Service Guard's install). If one box fails, the second one will take care of database data on the shared array.

Virtual IP address will guarantee that failover does not require any configuration changes for SNMP trappers, active agents, etc.

[bbittman]

yes! Distributed monitoring

- Distributed monitoring : is it planned / possible to have a distributed management : local autonomous manager reporting to a centralized one, with a central configuration, and if possible no polling between the local and the central manager ?

I'm a little lost in that but I think the jist of it is what we're looking for, let me rephrase

we deploy quite a few servers to our client's locations, these need to be monitored, of course this is exactly what zabbix is for. However, some clients also have critical services of their own going on that we don't nesscessarily want to monitor in our central database, what would be ideal is if we could put the zabbix_server (and DB) on our deployed machine, configure it to monitor their local services and other machines - allow the client's IT people to view the Zabbix data for their network, off of their server (meaning the one we're monitoring) - and put a zabbix_agent on the server to report summarized information about their network back to our central zabbix server. hmm - i hope i'm making sense, please inquire if you feel otherwise.

it seems to me that to implement something like this we have a couple options
1) let the server act as an agent so that another server could query it about the things it's monitoring
2) allow the zabbix_agent to query a zabbix_server (on the same machine.. or otherwise) about the things it's monitoring.

so, either way the server needs to be extended somewhat to allow something (an agent, or another server) to query it about what it's monitoring.

Thanks for the excellent product!