Hi I have a question about network discovery an answer to which I haven't been able to find anywhere yet.
I would like to use network discovery as described here (https://www.zabbix.com/documentation...work_discovery), but there's one crucial problem (?) with it: speed. In my tests where I deployed a zabbix server in a dockerized environment, it took 4-6 seconds per IP. If the we're talking a /16 network, it will take 5 days (((2^16) * 6) / 86400) to scan it with the single threaded discoverer process. That's not great because in my case I have dozens of those /16 networks.
So, my question is, do I understand correctly that it takes that long? Is there a way to mitigate this?
The other alternative I see is https://www.zabbix.com/documentation...o_registration but the problem with this approach is that the rules for autoregistration are evaluated only at the time of the agent's registration. If, for example, I later accidentally detach a host from a template, Zabbix will not fix that for me next time the agent reports itself. The registration is over at that point and the rules will not be enforced. With network discovery, according to my tests, all the rules are replayed when a host is found during the next scan. But again, it's impractically slow.
Am I missing something simple or is there a better option besides those two options?
Thanks!!
I would like to use network discovery as described here (https://www.zabbix.com/documentation...work_discovery), but there's one crucial problem (?) with it: speed. In my tests where I deployed a zabbix server in a dockerized environment, it took 4-6 seconds per IP. If the we're talking a /16 network, it will take 5 days (((2^16) * 6) / 86400) to scan it with the single threaded discoverer process. That's not great because in my case I have dozens of those /16 networks.
So, my question is, do I understand correctly that it takes that long? Is there a way to mitigate this?
The other alternative I see is https://www.zabbix.com/documentation...o_registration but the problem with this approach is that the rules for autoregistration are evaluated only at the time of the agent's registration. If, for example, I later accidentally detach a host from a template, Zabbix will not fix that for me next time the agent reports itself. The registration is over at that point and the rules will not be enforced. With network discovery, according to my tests, all the rules are replayed when a host is found during the next scan. But again, it's impractically slow.
Am I missing something simple or is there a better option besides those two options?
Thanks!!
Comment