Graph and screen are displayed differently

[Pada]

The discovery rules should typically be once per hour or so, and not once every second! You can perhaps make it once every 5 minutes, but even that is a bit overboard, unless your hosts are started up / shut down faster than that, which I doubt.

You should really change that update interval from 1s to say 30 seconds.

The lowest update interval that we have on our monitoring system is 5 seconds for a select few icmppingsec items.

Having the trend history set to 365 is fine. The longer you keep it, the more hard drive space it will take. Trends are stored/calculated on an hourly basis.

[sccuser]

I had restore the state of Zabbix server to one month ago + already changed all the interval time of items into 60sec, then it should be OK then.

except some parameters still keep 100% busy as image.

from the image, should I increase the value of StartPollers/StartPingers/StartPollersUnreachable?

I got a problem when changing the value of zabbix_server.conf, my method as below:
1. ssh to the Zabbix using WinSCP
2. do change zabbix_server.conf > saving the file then
3. sudo /etc/init.d/zabbix_server restart (but zabbix_server will be defunct very fast afterward)

also no luck if I do reboot the Zabbix server after the #3 step, as an error appears "Zabbix server is not running: the information displayed may not be current"

[Pada]

You can perhaps increase the respective pollers to say 20.

If your machine does not have swap space, then my guess is that the Zabbix server is going into '<defunct>' because its running out of memory. You will really have to assign more resources (CPU + memory) to that VM of yours.

Also make sure that you're not doing the discovery at a too short interval.

[sccuser]

Any adjustment on the zabbix_server.conf is always make the error "Zabbix server is not running: the information displayed may not be current" then, as it might insufficient of resource (Core, RAM, HDD...).

that's why I just did:
- change interval time = 60s
- remove all of discovery rules

and it's working aright then even some processes still getting busy at 100.

Perhaps I go ahead with a dedicated server for Zabbix, not VM as current.

Thank you both so much for the strong help, Pada & tchjts1!

[tchjts1]

VM's will be fine if you put your Zabbix App server on one, and your DB on another... giving your DB enough resources to do it's thing. I run my setup like that, currently with ~600 hosts.

As Pada mentioned, it comes down to getting the right combination of config settings, not only in Zabbix, but in MySql as well, and to some extent at the OS layer.

[sccuser]

Thanks tchjts1! it's very clear for me on your points.

I should go ahead with a dedicated MySQL server first (a separated server), to see how it works.

Much appreciated.

[tchjts1]

To give you an idea of our setup, the Zabbix App server has 4vCPU's and 8GB of memory. Our DB server has 4vCPU's and 16GB of memory.

These all live on Red Had Enterprise.

[sccuser]

Great info, thanks a billion!