Using proxies to failover active agents

[mmachado]

We are using Zabbix to monitoring geographically distributed agents with (primarily) active checks. Eventually, when my main datacenter's internet link gets down, we lose the data of these agents.

But we have two others datacenters where we can put some proxies to handle these requests on behalf of our Zabbix server, which will be useful to synchronize when the main link get up again.

This approach will be useful not only in a broken link scenario, but in cases of routing problems, where my main server is unreachable to some agents but one of the proxies continues to receive its data, bouncing it to the server.

Some sites have only one or two servers so, keep a proxy aside them is out of question.

Is there any way to accomplish this active checks failover?

Regards,
Marcos

[mmachado]

I was trying to balance my active requests (2 servers in master/master setup) using 2 A registers in my DNS. But the agents seem to get stuck with the first response and never try to connect to the second IP.

I'll take a look at the buffersend and buffersize parameters. This maybe can be used to minimize the lost of data problem.