What can delay 1 second in collecting the value of the item?

[cesarsj]

I need the synchronized report of values of three items, which read numbers recorded in their respective files.

The items have the keys:
vfs.file.contents [/var/run/zabbix/col.txt,ANSI], ranging only to col.txt, col2.txt, col3.txt.

By clicking on latest values and selecting the three items, I see that there is usually a value that causes the report to break because of 1 second.

How could I solve this break?

Our Zabbix Server have: 16G RAM, 4 CPU Core and is a VM (VMWare)

Zabbix have: 900 hosts and 6000 items active.

Zabbix_Server.conf have configs:

StartPollers = 5;
StartPollersUnreachable = 2;
CacheSize = 16M;
HistoryCacheSize = 16M;
ValueCacheSize = 8M;
Timeout = 4;

Would it be the case that I adjust some Zabbix Server configuration? Where do I see which zabbix_server.conf optimal settings, showing calculations for me to base on?

[cesarsj]

My boss wants to understand why there is this one second difference. He believes Zabbix Server is dying. And until I convince him the real cause of this one-second difference, he hasn't forgotten that question. : /

Do you know if there is anything in the documentation about it?

I have provided some server and system settings, if there is anything I can do to improve performance, please ask them to help me.

[ingus.vilnis]

Master item with dependent ones is a good idea but not always you can implement nicely moreover if you need to check contents of three different files.

How about using passive agent items with scheduling intervals?

https://www.zabbix.com/documentation/4.0/manual/config/items/item/custom_intervals

Set the "normal" interval to 0 and use exact scheduling instead.

Still, from real life experience there is a chance that the results can be off by a second and it has nothing to do with poor performance of a dying Zabbix.

[cesarsj]

Set the "normal" interval to 0 and use exact scheduling instead.

Still, from real life experience there is a chance that the results can be off by a second and it has nothing to do with poor performance of a dying Zabbix.

It's already like this. Thank's for reply.

[ingus.vilnis]

What is going on there
- Zabbix server initiates the check for the item
- Network transports the request to the agent
- agent has to gather the required data (and the monitored server there could be busy with other tasks)
- Network transports the value back
and all of that should happen within a single second for your boss to be happy....