This is revisiting a support question I've emailed you back in January. I followed your recommendation and for a time performance was better. However, after adding more items and hosts, it is slowing up again. This is my current configuration:

Number of hosts (monitored/not monitored/templates/deleted) 107(83/1/23/0)
Number of items (monitored/disabled/not supported)[trapper] 11726(9170/840/1716)[0]
Number of triggers (enabled/disabled)[true/unknown/false] 2468(2372/96)[5/4/2363]

zabbix_server.conf:
StartPollers=3
StartTrappers=90

my.cnf
max_connections=600
thread_concurrency = 500
innodb_buffer_pool_size = 2G
innodb_thread_concurrency = 64

Initially I was using mostly Zabbix Active Agents. However, now it is about 80% Zabbix Agents, while the rest are Zabbix Active Agents.

I would think I would need to increase the StartPollers because of the use of Zabbix Agents (non-active), but when I did that (StartPollers=150), mysql CPU utilization shot up high ~40% (load average 20+).

Keeping StartPollers=3 will keep the load average around 1. However, I fear that not data is not being collected in a timely manner. I see breaks in graphs.

Can you please advise what should be the settings for best practice.