Zabbix_agentd ko, and no trigger "red"

[Tinto]

Hi all, i'm quite new to Zabbix and i'm still studying it.

I've noticed my Zabbix server's cpu load was high so i took a look at the graph.
The graph rises suddenly over 7 and then "dies" in sense that the server does not recieve any data.

I've tried to restart the zabbix_agentd with no result. Then i tried to stop it.

With my suprise, the trigger "Zabbix_agentd is not running on {HOSTNAME}" remained green, i've been waiting for more than 10 minutes but still green.

So the first question is... why the data arent'receieved any more?
The second question is "sightly" more important: why the "death" of an agent does not trigger any alarm on the server??? How can I trust a monitoring system if "no data recieved" is equal to "no alarm"???
It seems the lastest data recieved is treated as the lastest data?!

I was used to use Hobbit (surely a very limited tool if seen vs Zabbix): when data cannot be recieved from the agent it's clearly put in evidence with the dedicated "purple" condition...

[walterheck]

So the first question is... why the data arent'receieved any more?

The logs can tell you more about that, both on the agent as well as on the server.

The second question is "sightly" more important: why the "death" of an agent does not trigger any alarm on the server??? How can I trust a monitoring system if "no data recieved" is equal to "no alarm"???
It seems the lastest data recieved is treated as the lastest data?!

This is a bit of a flaw in the default linux template: it doesn't include a trigger for nodata. I usually trigger off of something very simple for that, like CPU load or so: {Template_Linux:system.cpu.load[,avg1].nodata(600)}=1