Hey There!
I've been having a pickle of a dilly understanding if my current set up is optimal. I'm using zabbix 7.4 and monitoring windows servers with active checks primarily.
I have the following servers set up [zabbix postgres db] <- [zabbix server] <- [4 zabbix proxies]
There are 2 sets of 2 zabbix proxies, each proxy group (set up in the front end) is for a specific hosting region.
Initially, my agent config file looked like this:
SERVER=<ip of zabbix server and proxies separated by commas>
SERVERACTIVE=<ip of zabbix server and proxies separated by semicolons>
After noticing a ton of noise in the logs, I separated into 2 agent deployments for regions, and now my agent config looks like this
SERVER=<IP of zabbix Server, IP of regionXproxy1, IP of regionXproxy2>
SERVERACTIVE=<IP of regionXproxy1>
It's my understanding that the server field should have the main zabbix server and both proxies, because it needs to "reach" the main zabbix server for load balancing, and the proxy servers in case i add some passive checks (i probably wont).
It's also my less solid understanding that I only need to place one proxy in the server active field, and that the zabbix server itself will "load balance" the host based off its auto registered proxy group into a specific proxy.
My main question is if this set up is acceptable for a somewhat large (1800 host) environment (may need to add proxies eventually but load is fine for now.) I notice in the agent logs lots of passive checks being attempted by the zabbix server itself, but I'm thinking those might be some automatic discovery actions I put in place in the front end. I would say that largely, this set up is working. But so was the initial set up and it was only after I started scouring logs that I noticed tons and tons of failed connections. I still see some but they are much more toned down. In such a large environment if i can minimize traffic I would like to do so. Would appreciate any advice, thanks in advance!
I've been having a pickle of a dilly understanding if my current set up is optimal. I'm using zabbix 7.4 and monitoring windows servers with active checks primarily.
I have the following servers set up [zabbix postgres db] <- [zabbix server] <- [4 zabbix proxies]
There are 2 sets of 2 zabbix proxies, each proxy group (set up in the front end) is for a specific hosting region.
Initially, my agent config file looked like this:
SERVER=<ip of zabbix server and proxies separated by commas>
SERVERACTIVE=<ip of zabbix server and proxies separated by semicolons>
After noticing a ton of noise in the logs, I separated into 2 agent deployments for regions, and now my agent config looks like this
SERVER=<IP of zabbix Server, IP of regionXproxy1, IP of regionXproxy2>
SERVERACTIVE=<IP of regionXproxy1>
It's my understanding that the server field should have the main zabbix server and both proxies, because it needs to "reach" the main zabbix server for load balancing, and the proxy servers in case i add some passive checks (i probably wont).
It's also my less solid understanding that I only need to place one proxy in the server active field, and that the zabbix server itself will "load balance" the host based off its auto registered proxy group into a specific proxy.
My main question is if this set up is acceptable for a somewhat large (1800 host) environment (may need to add proxies eventually but load is fine for now.) I notice in the agent logs lots of passive checks being attempted by the zabbix server itself, but I'm thinking those might be some automatic discovery actions I put in place in the front end. I would say that largely, this set up is working. But so was the initial set up and it was only after I started scouring logs that I noticed tons and tons of failed connections. I still see some but they are much more toned down. In such a large environment if i can minimize traffic I would like to do so. Would appreciate any advice, thanks in advance!
Comment