Which monitoring tool to choose?

Question

Monitoring

Long used to monitor nagios and munin servers.
Nagios as a system for alerts is quite satisfied: it is easy to write plugins, easy to configure, there are many ready-made plugins.
But munin is just a nightmare. Terrible graphics, it is impossible to scale, make selections, it is inconvenient to compare information.
Plus, I get two different monitoring systems with different data captured by different plugins.
Ideally, I would like a system that could build graphs that can be scaled and that it could send alerts, had a plugin base. There are not many servers yet, about 10, but growth is planned.
You need to monitor system information, raid controllers, disks (especially raid and disks) and the elasticsearch cluster.
Immediately as an option - zabbix, but its monstrosity repels me.
The second option is graphite and grafana.
The third option is Shinken.
What do you advise?
I use puppet as a server configuration system.

Answer 1

zabbix is ​​not a monster, you just need to figure it out.
You can look at a simple thing monit and mmonit

Answer 2

Increasingly, these systems are used together. For example, Nagios for monitoring the availability of servers and networks, Zabbix - for taking metrics from end machines.
In your case, Zabbix is ​​perfect. A definite plus is autodiscover, ease of setting up servers of the same type and a large number of metrics out of the box. But it can easily generate a large load if it is not carefully tuned (an excessive number of frequently updated metrics). For example: a database with close to standard metrics and settings for 4 linux servers has grown to 35Gb in 2 years.

Answer 3

In fact, now I am faced with the fact that people have some very strange understanding of what monitoring is. Many people mean by it beautiful graphics in graphana and forget about alerting.
1. Zabbix is ​​certainly good, but its convenience is somewhat exaggerated. Autodiscovery is also not the most convenient option, you still have to generate a bunch of different templates, triggers, etc., the JSON format is strictly formalized. Another disadvantage: you can’t automatically place all items from one discovery on one chart, adding a chart to latest data, you can’t save it as a permanent one and put it on the screen. there is no automatic generation of screenshots from discovery either. The graphs are again very depressing because they are still rendered on the server, and it is not very convenient to look at the values ​​of metrics on graphs consisting of several data sources. But still there is https://github.com/alexanderzobnin/grafana-zabbix for zabbix.
2. graphite + grafana, there is no alerting here, you will have to install a third-party one (such as graphite-beacon). But there is also something else to collect data (collectd/diamond). Also, 0.9.10 is dead, there are a lot of requests in the turnip pool, no one is involved in the project, it was decided that 0.9.13 would be the last version and then 1.0, but somehow it all barely moves.
3. Shinken - subjectively, I think he is dying.
4. If you already have experience with Nagios, then look at Icinga 2. After they rewrote the core and updated the web interface, it became possible to somehow use it.

Answer 4

In your version, only zabbix will do, but damn it is from the beginning of the 2000s, of course I want something new, but so far there is nothing good.