Distributed Monitoring against Centralised

[cm2000]

Hello,

Background:
We are running two sites, office (non-critical equip) and 1 data centre (critical equip). Running Zabbix 1.6.6 from the office, which is monitoring both sites.
Alerts are sent via email to sms gateway via the office internet feed.

Problem:
I've been asked\suggested to get a second zabbix server set up at the data centre, in the event comms fail, or alerts cannot be raised thro the office www link. So, this new zabbix server would send alerts the same mechanism, but via the DC's local internet feeds.

So, do i do a distributed setup, or do i keep both zabbix servers seperate?
I'm keen by the distributed setup, although nervous (after reading forum posts) about the converting table process and loosing my current data. Also what scares me is if we had to revert back for any reason.

If I had two seperate zabbix's, i'd have two seperate logons for monitoring, and i'd have to copy templates\hosts across, and i guess I'd set it so each zabbix monitors its own local office along with a few checks to the other office (e.g. if there is comms)

I looked at the zabbix proxy, but it cannot do notifications which is a shame.

what do you guys suggest is the best method?

Cheers,

Tom

[cm2000]

Ooops

Apologies, I'm in the wrong forum...

feel free to move me to "Zabbix Help"...

[richlv]

it probably depends on how much you want to avoid two separate setups, which would be the most simple way.
distributed setup would be easier to administrate later, but i'd strongly suggest having backups and testing it before converting.

[chrisf]

I'd advise against the distributed setup.
It breaks too easily and troubleshooting it is practically impossible.
I had it set up for a mission critical site and the communications broke down and essentially left us open to critical failure.

I would do the following:
data center zabbix -- Monitors all data center hosts and office zabbix server
office zabbix -- Monitors all office hosts and datacenter zabbix server

Both systems send out their own alerts.

Again I cannot with good conscious advise using the distribute node setup in a production environment. It's just not ready.

-Chris