We currently have 7 zabbix systems (1 server vm and 1 database per "system") each handling about 12k hosts and 3k kvps each. in total there are roughly 76k hosts. its been a struggle to grow beyond this without scaling horizontally. We aggregate data across databases with custom scripts to unify all host data on a single database. We have roughly 1000 passive proxies per 10 devices. For our current setup, we seem to be bottlenecked by the 250 "StartProxyPoller" option in the config. According to the service status charts, the Proxy Pollers are roughly around 50% capacity. In order to handle surges of data, we don't want to exceed this. So I'll ask directly, is there a way to bypass the 250 Proxy Poller limit? Is compilation of the software an option to do this manually? Are there technical reasons why the limit is 250?
-
-
TBH, you should ask support for that answer...
But do I get it correctly... 1k proxies and just 10 hosts per proxy? The other way would be be better... Of course there can be reasons, what we do not know....
Here I have approx. same as your "one system"... 12k hosts, 3k nvps... running 16 proxies, just running 8 proxy pollers... process average seems to be ~20% -
We require many proxies due to firewalls, etc. Its not something we can easily change. 250 ProxyPoller average is about 37%, but can spike in the 100% range when system reboots happen, or some large event.Comment
Comment