Zabbix agent instances are not really isolated from each other

[aysabzevar]

We have some servers that each one has many items should be monitored (lots of them are UserParameters). Some of this items need to access to internet, other network access check, etc.
Although I increased number of pre-forked instances of Zabbix agent on all of these server (StartAgents=40) but Zabbix server still shows gap on all graphs of all items on these servers because the lag of internet or other network access for those servers.

The main local network (Zabbix server and Zabbix agents connections) is stable and Zabbix server works perfectly and has not any items on the queue.

If this is and instances are not really isolated from each other, can I put one or more UserParameters in a single instance that has no effect on the other instances on the same machine?

[kloczek]

We have some servers that each one has many items should be monitored (lots of them are UserParameters). Some of this items need to access to internet, other network access check, etc.
Although I increased number of pre-forked instances of Zabbix agent on all of these server (StartAgents=40) but Zabbix server still shows gap on all graphs of all items on these servers because the lag of internet or other network access for those servers.

The main local network (Zabbix server and Zabbix agents connections) is stable and Zabbix server works perfectly and has not any items on the queue.

If this is and instances are not really isolated from each other, can I put one or more UserParameters in a single instance that has no effect on the other instances on the same machine?

Just stop using passive zabbix agent setup (StartAgents=0), setup in agent proxy/server name (ServerActive=<host>) and replace all "zabbix agent" items by "zabbix (active) agent" one.
You have probably some congestion in reading some metric data and this stops whole process of reading other metrics.

[aysabzevar]

Just stop using passive zabbix agent setup (StartAgents=0), setup in agent proxy/server name (ServerActive=<host>) and replace all "zabbix agent" items by "zabbix (active) agent" one.
You have probably some congestion in reading some metric data and this stops whole process of reading other metrics.

Actually we can not use Zabbix active check, because we have several commands that should be run on the different servers via Zabbix agent.

[aysabzevar]

Solved the problem

I solved the problem with defining "timeout" option on all of my external commands executed by UserParameters.
For example if the command was curl command, I used curl command with option -m=2.