Understanding Zabbix proxy

[lmontalvanr]

Hello everyone.
Before I deployed zabbix on a single server, but I have a requirement from a client that has more than 5000 devices in different geographical areas, I just request your help to understand the distributed zabbix system.

I do not understand how this works and what components I must separate. The only thing I know is that I must have a proxy for each geographical area, now my doubts come:

- I must have a separate server for web-front, zabbix server, database and that the proxies point to both the server and the database. (arquitecturqa 01.jpg)


- Should I have a separate server for the web front and zabbix server, and on each zabbix proxy server install the database? (arquitecturqa 02 .jpg)


Finally I want to know what database to use.
Does mysql or postgresql work with zabbix proxy or sqlite3 works only with zabbix proxy?

[Jason]

Your second architecture is better. You can do some HA with zabbix server and corosync/pacemaker. It's far better to have your database on a separate server and spend time getting that optimised.
Proxys can be sqlite3, mysql or postgres. Sqlite3 being the simplest to set up although not best on performance.

[lmontalvanr]

Thanks Jason!!!.

It's true sqlites is simplest. I choose the second architecture.

Now I have one server with zabbix-web (only zabbix-web and zabbix-web-mysql package installed), other zabbix independent (zabbix-server-mysql installed) and a database server (mysql).

The severs zabbix-server and database works, but the zabbix-web I don't know how to integrate with zabbix-server.

It possible zabbix-web and zabbix-server separate?

[ewestdal]

We have a very large environment collecting about 8000 NVPS. All of our proxies have the sqlite DBs setup and we don't have any issues.

We also have our zabbix web and zabbix server completely seperate.

[Deriuz]

The second drawing is very promising.

[LenR]

My next "research" project it to see if I can replace my smaller VM mysql proxies with a docker sqlite proxy. I only have 1 or 2 candidates, but due to our billing model, I could save (imaginary) money.

[c1support]

Hi there

I know that the recommended architecture is as above #2 (HA with corosync/pacemaker).
However, I am still wondering if #1 would be possible.

The motivation behind my question is that we have two zones spread across two datacenters (DC-1 and DC-2), mainly used for redundancy.
The information from DC-1 is only available through Zabbix proxy to DC-2 Zabbix server - and vice versa.
The two datacenters are operated by different teams, and both would like to have the full information (i.e. covering both DC's) available.
If disaster strikes (say in DC-1), the DC-2 Zabbix Server should be seamlessly available (without any reconfiguration) (of course only reaching her own datacenter's info, i.e. DC-2 Zabbix proxy) since the guys should be allowed to work on more critical issues...

=> Is it even possible to use one Zabbix Proxy for two Zabbix Servers?
=> Would it be better to cascade Zabbix Server instances and let them communicate via Zabbix API? (I think I saw something like this, but can't find the link to this proposed architecture right now)
=> Or is there a better alternative?

Any opinion is helpful :-)

[vendrusculo]

I'm using the #2 and getting 7900+ vps, proxies are using sqlite as database and I'm using a postgress (1.1TB) behind the zabbix server, I like to think that the database for proxies are volatile

As far I understand the #1 is possible but you will not gain anything puting in the same database, all the data will be sent over the net from proxies to the server and save it again to the database and maybe you will create an unacessary overhead in the database

About your "motivation"... if Zabbix didn't changed anything on 5.0 is not possible to have 2 servers "active-active"... so you can have a mirrored database and a standby server on DC2 in case of failure you need to start the secondary server and point the proxies to this server located on DC2

=> Is it even possible to use one Zabbix Proxy for two Zabbix Servers?
I don't think so because the configuration from the secondary server would overwrite the configuration from the primary

=> Would it be better to cascade Zabbix Server instances and let them communicate via Zabbix API? (I think I saw something like this, but can't find the link to this proposed architecture right now)
I've never seen this before I believe it is not a "native" feature

=> Or is there a better alternative?
would be awesome if someone has