I'm trying to set up a Zabbix installation for a customer. They're primarily interested in various performance metrics for their cluster (CPU, memory, load, disk, etc.) with high granularity.
We'd also like to collect that data for ourselves.
We have our own central Zabbix server installation which does our internal stuff.
We've got a "bridge" host that lets us access the customer's network (which is very heavily isolated).
We want to have a Zabbix server in their network that we can use to store the performance data for their cluster, and have it be available to them on their network.
And we want that same data to be available from our central server for our own use.
Is there a simple way to set this up?
At the moment, the only way I can see to do it is:
It's workable, but I'm concerned about the CPU overhead of running twice the Zabbix agent queries per cluster host as well as the network overhead on the bridge host from constant Zabbix agent queries.
We can't (obviously) expose our central Zabbix system to the customer. We also don't want to allow network traffic from our central Zabbix system to route directly into their cluster via the bridge host.
What I'd like to be able to do is put a Zabbix proxy in their cluster (to do the grunt work for both Zabbix servers). That proxy would collect data for both the customer's Zabbix server (routed directly) and our central server (via another Zabbix proxy on the bridge host).
Except, as far as I can tell, Zabbix doesn't let you relay traffic through two proxies like that.
Is there something I'm missing? Another solution that someone could recommend?
The primary goal here is to minimize the network and CPU overhead on the bridge host, because it's our only way of accessing their network (and other customer networks).
We'd also like to collect that data for ourselves.
We have our own central Zabbix server installation which does our internal stuff.
We've got a "bridge" host that lets us access the customer's network (which is very heavily isolated).
We want to have a Zabbix server in their network that we can use to store the performance data for their cluster, and have it be available to them on their network.
And we want that same data to be available from our central server for our own use.
Is there a simple way to set this up?
At the moment, the only way I can see to do it is:
- Install a Zabbix server on an infrastructure node in their cluster that hits the cluster with Zabbix agent queries.
- Install a Zabbix proxy on our bridge host that also hits the cluster with Zabbix agent queries.
It's workable, but I'm concerned about the CPU overhead of running twice the Zabbix agent queries per cluster host as well as the network overhead on the bridge host from constant Zabbix agent queries.
We can't (obviously) expose our central Zabbix system to the customer. We also don't want to allow network traffic from our central Zabbix system to route directly into their cluster via the bridge host.
What I'd like to be able to do is put a Zabbix proxy in their cluster (to do the grunt work for both Zabbix servers). That proxy would collect data for both the customer's Zabbix server (routed directly) and our central server (via another Zabbix proxy on the bridge host).
Except, as far as I can tell, Zabbix doesn't let you relay traffic through two proxies like that.
Is there something I'm missing? Another solution that someone could recommend?
The primary goal here is to minimize the network and CPU overhead on the bridge host, because it's our only way of accessing their network (and other customer networks).