Hi,
I've been searching the internets about this, but the majority of results deal with installing a Zabbix cluster.
So: in dealing with monitoring cluster, the classic solution is to define Zabbix hosts for each cluster node, on which to monitor node-specific items (like resource usage) and to define a cluster host, to which the VIP address is assigned, on which to monitor cluster-specific items (like number of connections).
How do you deal with monitoring clusters which do no have a VIP? A basic example would be Couchbase, on which you can get the cluster-specific data from every node?
You could monitor the cluster on each node, but that would mean that you overload the servers with duplicate items and, if you add/remove nodes, you risk losing historical data.
You could use only passive items, and do some external "magic", but the how would you deal with necessarily active checks (like jmx items on Kafka clusters)?
Thanks for any input here.
I've been searching the internets about this, but the majority of results deal with installing a Zabbix cluster.
So: in dealing with monitoring cluster, the classic solution is to define Zabbix hosts for each cluster node, on which to monitor node-specific items (like resource usage) and to define a cluster host, to which the VIP address is assigned, on which to monitor cluster-specific items (like number of connections).
How do you deal with monitoring clusters which do no have a VIP? A basic example would be Couchbase, on which you can get the cluster-specific data from every node?
You could monitor the cluster on each node, but that would mean that you overload the servers with duplicate items and, if you add/remove nodes, you risk losing historical data.
You could use only passive items, and do some external "magic", but the how would you deal with necessarily active checks (like jmx items on Kafka clusters)?
Thanks for any input here.