Passive auto-discovery?

[mpotter-xiss]

We will be using zabbix to monitor 5000+ servers and are currently working out the scaling issues. What I would like to know is if it is possible to have Zabbix set up in such a manner that I can install the agent on a host, use the agent to contact the Zabbix server, and do auto-discovery in that manner? There is no way we can put the amount of traffic on the wire necessary to use active auto-discovery regardless of the configuration.

The task of adding the hosts by hand is fairly daunting as you can imagine. If anyone has any advice please let me know.

[troffasky]

There is no way we can put the amount of traffic on the wire necessary to use active auto-discovery regardless of the configuration.

The task of adding the hosts by hand is fairly daunting as you can imagine. If anyone has any advice please let me know.

Could you put the Zabbix server behind a router/firewall and rate-limit it, perhaps? Alternatively, if you already have the key bits of data about your 5000 servers in tabular form, you could write a script to import them into the database directly, bypassing the Zabbix front end. Personally if I were about to monitor 5000 nodes with Zabbix, I'd be thinking about paying for commercial support :-)

[Alexei]

What I would like to know is if it is possible to have Zabbix set up in such a manner that I can install the agent on a host, use the agent to contact the Zabbix server, and do auto-discovery in that manner?

This functionality will be supported soon in one of 1.6.x.

[mpotter-xiss]

Could you put the Zabbix server behind a router/firewall and rate-limit it, perhaps? Alternatively, if you already have the key bits of data about your 5000 servers in tabular form, you could write a script to import them into the database directly, bypassing the Zabbix front end. Personally if I were about to monitor 5000 nodes with Zabbix, I'd be thinking about paying for commercial support :-)

The overview is that the vast majority of the monitored pieces will be cluster nodes. We cannot have auto-discovery turned on at all if it is active. However we can use it to discover one cluster at a time (two subnets per cluster for our monitoring purposes) and then turn it off. This is what we are currently doing and it works alright.

This functionality will be supported soon in one of 1.6.x.

That is awesome! Zabbix's features already have it a head above the monitoring solution I was using previously in other environments (Nagios). This feature would be a major piece of the puzzle for a lot of people I think.