ZABBIX Agent 2 vs Telegraf Ideas

[baggord]

Hi everyone,

I've some questions if you had any experience on Agent 2 and Telegraf. I would like to hear from you.
I'm using zabbix agent (1) on most of the hosts in my env. and testing telegraf in 3-5 servers. In my opinion telegraf does a better job and it's easier thanks to plugins. Now, I read about agent 2 written with go and plugin support. I wonder if agent 2 is better than agent 1 and if any of you using in production?

For example, zabbix agent 1 does not aware of each disk cant return disk IO if you dont do a LLD for disks. Telegraf can return metrics for all disks easily.

[Atsushi]

For the benefits of using Zabbix Agent 2, see the announcement material in the other day's summit at the following URL:

https://assets.zabbix.com/files/zabbix_summit_2019/Aleksey_Petrov-Magic_of_the_new_Zabbix_Agent.pdf

The official release of Zabbix agent 2 seems to be from Zabbix 5.0.

LLD is used to find disks and network interfaces because Zabbix needs to have a place to store data as items before it can retrieve the value of the item. In Zabbix, if you know the pre-connected disks and network interfaces, you can start monitoring immediately by defining them in the template.
The advantage of using LLD is that you can monitor servers with different hardware configurations with the same template. If you want to end the discovery process quickly, shorten the discovery interval or use Check Now.

[gofree]

In my opinion one of the big drawbacks of zabbix agent 2 is that you will have to recompile the agent everytime youd like to add new plugin - at least in the begining of the zabbix agent 2 production release - see https://support.zabbix.com/browse/ZBXNEXT-5453

Hopefully there will be some usefull plugins already out of the box - guess the version 5 will be the one that might bring zabbix in to the new time of cloud apps, k8s, etc...

Telegraf seemed interesting - event there were some talks about telegraf zabbix integration

https://github.com/influxdata/telegraf/issues/5652

https://github.com/influxdata/telegraf/pull/3966

At the end of the day there will be the question of how much can/will be integrated and why not use the original product....just my two cents.