Monitoring multiple servers?

Colleagues, good afternoon.
In the last time there was a need to monitor virtual machines on servers, their statuses and alerts in case of any incidents.

That, we have about 10-15 VMS, each of which made his task; has a different amount of cpu, ram, hdd, etc.
We have monitoring. There is a production server that "themonitored" reference. Actually like and want to see.
Since this prod is one - netdata it is convenient to monitor in real-time; watch what happens in a wheelbarrow, when it comes to alert from the same netdatы (for example, increased background 5xx on the web server), I immediately go to netdata in the logs of nginx.
There is also a bunch of netdata+influxdb. The second pushes all of the monitoring server in grafana, which nicely and clearly draws it to herself. Thus I can see what happened in the car at some ancient time.

Comfortable? Yes. In fact, all that is necessary is. Alerts, realtime+nerealnym monitoring.
That "not" out of this - firstly, the dependence on multiple systems: for realtime need netdata+install influxdb to push data (by the way, you can save the BD influx on a local wheelbarrow, and the wheelbarrow monitoring? as local gigs already ate 10 with all the values).

This is useful when the condition monitoring 1, 2. A maximum of 3 cars.
But not 15.

I'd like to find a solution like zabbix-server, zabbix-agent, but Zabbix I don't particularly enjoy (maybe not used?)

Anyway, for the most part, the VM don't need realtime (although it would be convenient to have) (anyway everywhere you can put netdata, if necessary), but need realtime alerts+save charts monitoring.
Another condition is that we have a separate car for monitoring. That is very simply want to put the agents (as in the case of zabbix) on the server to carry out the minimum configuration and seostamine to configure the server.

The question is whether there are integrated alternatives zabbixу? Or only he will have to deal with it?
Thank you!
March 19th 20 at 08:47
3 answers
March 19th 20 at 08:49
How to install and configure Elasticsearch, Logstash, Kibana (ELK Stack) on Ubuntu/Debian/Centos
https://serveradmin.ru/ustanovka-i-nastroyka-elast...
ELK - a visualization and search the logs. However, servers have different metrics, which are not written in logs, but they still need to watch. - Lue84 commented on March 19th 20 at 08:52
@Lue84, it is metricbeat
https://www.elastic.co/guide/en/beats/metricbeat/c... - Einar.Monahan commented on March 19th 20 at 08:55
@Einar.Monahan, @Ayana35 something with ELK I wouldn't start. besides, he seems to be in java (I could be wrong?) and eats a lot of space.
for example on the VM to access a VPN (1cpu, 512ram) it barely can zapustil, I think.

plus Yes, indeed, it seems to me, a small crutch. or not? - Priscilla.Thompson commented on March 19th 20 at 08:58
@Priscilla.Thompson,
java is only used in the elasticsearch server, the agents are self-contained and written in Go - Einar.Monahan commented on March 19th 20 at 09:01
@Einar.Monahan, but either way, as I understand it, it's still the collector logs a La splunk. - Priscilla.Thompson commented on March 19th 20 at 09:04
@Priscilla.Thompson, Yes, but splunk is a collector not only of logs, and elastic
example
https://www.elastic.co/guide/en/beats/metricbeat/c... - Einar.Monahan commented on March 19th 20 at 09:07
@Einar.Monahan, right, but, as I understand it, there are some things like monitoring nginx do not go out of the box? plus then you can send alerts, monitor cpu wait normal, etc., how does netdata?
or is it all set up and has been custom? - Priscilla.Thompson commented on March 19th 20 at 09:10
@Priscilla.Thompson, the Solution to memory leak
https://stackoverflow.com/questions/27359885/elast... - Ayana35 commented on March 19th 20 at 09:13
@Priscilla.Thompson, mmm, it depends on what you mean by the box.
You have a server in elasticsearch, there are agents that send data, agents need to include modules locally through the config. Those config comes not from the server itself, means or handles on or through any ansible. Have elastica "the official user interface" - kibana, it is possible to create "dashboards" or handles, or load standard. Graphs can be configured as you like.
In General, you'd better browse the article some type https://ru.bmstu.wiki/Elastic_Beats
Actually, I'm not campaigning for elastic, and I can't say he is better than zabbix is in your case. - Einar.Monahan commented on March 19th 20 at 09:16
@Ayana35, it is a strange solution for the problem "how to run elasticsearch on a server with 512MB RAM" - Einar.Monahan commented on March 19th 20 at 09:19
@Einar.Monahanlooks, of course, beautiful. especially with lots of charts and TA.
but well, I'm almost ready to take Zabbix.
please tell us the cases. ie, how to monitor servers that to them to slip and how to build the necessary graphics (graphics for webmasters including RPS, number of 5xx etc)

what we have: production web server; apache. want to monitor system+Apache in particular, the network (probably logs) and not leaving it to disk (in the same system?)
multiple web servers with different VM number of cpu, ram. monitor system+web server.
networks server (and monitor Zabbix?)
2 storage server; here it is important to monitor io svobodnoe/used space, etc.
the mysql server is slave, monitor is it normal arrives replica, some seconds behind master
the vpn server. I want to monitor number of connected users

well, the little things some number of servers monitoring just the system in General (cpu, ram etc).

any idea how it all "beautiful" tomonitoring? for each part of the service to create their own templates, etc. (for example, networks servers, and other) - Priscilla.Thompson commented on March 19th 20 at 09:22
@Priscilla.Thompson, in short, to purchase splunk (Oh, not to buy in Russia)
To draw everything yourself, or look for ready-made templates and edit them for yourself. For example, there is a chart template hits copied-edited-do filter by event_code, throw on the dashboard.
As I recall sabika exactly the same thing, so what's here you need to choose?
https://www.zabbix.com/integrations/nginx
Actually, if you have identified a list of the metrics you already collect through netdata. Data need somewhere to keep. Judging by https://docs.netdata.cloud/backends/
netdata can send data in different databases. Carry InfluxDB on a separate machine with grafana that you already use.
Yes, it's not out of the box, and you need to deploy a garland, but zabbix, apparently you will configure from scratch and suffer. - Einar.Monahan commented on March 19th 20 at 09:25
@Einar.Monahan, as I understand it, only splunk collects logs?
I have it, it pouchitsa access.log production web server. I not really came

that is your tip - ligament netdata+influxdb+grafana slightly better than Zabbix on all hosts and zabbix-server monitoring-server? - Priscilla.Thompson commented on March 19th 20 at 09:28
@Priscilla.Thompson, for example here kilo-tech.blogspot.com/2017/12/making-simple-splun...

In a first approximation, Yes, maybe you don't have to configure everything from scratch on a new unfamiliar instrument. Plus there is a nuance, a universal system type Zabbix not likely to have templates for all of your metrics that you already use. They certainly can adjust, but how much time you will spend is unknown, and in principle, it will be with any system.
Zabbix will have to use if you start to monitor very very different things, like thousands of ports on the switches using snmp, the temperature in a different server, the voltage at opsah, and a bunch of bespoke scripts receiving the different weather on Mars. And even then, the trend in recent years to put it all in a horizontally-scalable database type elastica, and not to force poor relational mysql/postgre
If you are trying to solve the problem that each new host will have to configure the data collector, it is necessary to look in the direction ansible/puppet to work with configs. - Einar.Monahan commented on March 19th 20 at 09:31
@Einar.Monahan, got it, thanks! - Priscilla.Thompson commented on March 19th 20 at 09:34
March 19th 20 at 08:51
it also comes linked server agent?
or what? it can be a bit "under the hood" to tell - Priscilla.Thompson commented on March 19th 20 at 08:54
@Priscilla.Thompson, consider a direct analogue of Zabbix - oma40 commented on March 19th 20 at 08:57
@oma40, that is, with their graphs and so and so?
and, as I recall, there is still icigna2. it seems like it's advise, no? - Priscilla.Thompson commented on March 19th 20 at 09:00
@Priscilla.Thompsoncan be icinga2? In fact it's the same Nagios with their poets. I prefer Zabbix - oma40 commented on March 19th 20 at 09:03
@Priscilla.Thompson, what is there to tell? There is a nagios server, which continuously polls the agents and remove them from the data. There are agents who give the data. The server raises an alert if the agent gave the status of the alert and removes it if the agent ceased to give it. Graphs any basic (free) version doesn't draw, though there is a pnp4nagios.
Its advantage is the ability to fasten it to any agents. Monitor the voltage of power supply via IPMI? Easy. Monitor how much powder is left in the cartridge? Easily. To monitor whether you are running a service on Windows? Yes, without the Bazaar. - Rozella.Cro commented on March 19th 20 at 09:06
March 19th 20 at 08:53
For small quantities you can use PRTG
I watched, but it is not the little specifics that you need.
**as I understand** - Priscilla.Thompson commented on March 19th 20 at 08:56

Find more questions by tags UbuntuMonitoring