More that 500 server. Active or Pasive more?

[verop]

Hi,

We have more than 500 servers monitoring throgth Zabbix.
The servers are in passive mode.

The problem is that when we have a network problem in the zabbix server, all of the environments that we are monitoring seems like unreachable.
To resolve this problem we have go down stop zabbix agent in all servers and after a few minutes go up the service in each server.
This problem is because the zabbix server don't see a change on the host status item and it think that all of servers follow unreachable and don´t collect data.

The agent on the servers are working and if we probe with zabbix_get we can collect data.

The question is if we can solve this problem with passing our servers in active mode or the number the servers (more than 500..) is to much to do this change.

Thanks in advance!!!

[untergeek]

Active checks are really quite nice. You might give them a try. They also reduce load on your database server.

[verop]

Thanks for the prompt response!

I forgot to say that in each server is configured an internal network where there are some equipments connected. So we have configured Zabbix Proxy to collect equipments's data.

We thougth that was better if the servers work in passive mode to lower the polling and balance the load in the zabbix server.
Currently the zabbix server is asking the agents's data and waiting the proxies's data.

With the zabbix proxy running on each server is better change to active mode?
Why in active mode reduce load in the database server?

Thank in advance!

[untergeek]

The names are kind of misleading.

Passive means that for each item, the Zabbix server queries the database for information, then makes a request to the remote agent, waits for data, then writes it to the db after it has been received.

Active agent means that at start up (and at the proscribed interval) the agent requests a list of all items marked with type "Zabbix Agent (active)" in the UI. It holds this list, complete with intervals and other pertinent data, and sends the data back to the server without subsequent checks from the server.

In other words, passive checks are all one at a time, active checks are like a bulk "to-do" list sent to the agent to process at the proper time. The server then just waits for the data to come back.

Unless your checks are really far apart, it can save a considerable amount of db cycles to use active checks. The bigger your zabbix environment gets, the more benefit you will see from using active checks.

[verop]

Thanks again!

I just have one more doubt: we are worried to overload the zabbix server due to too many concurrent connections.
Given a number of connections, the main difference between passive and active mode is only the number of DB queries, right? No polling limitations?

Thanks!

[untergeek]

That is correct.