Zabbix waste a lot of traffic

[Roke]

Hello,

I have several servers monitorized by zabbix (30 or more), i check the traffic with cacti tool and i can see that there is the same graphic for the traffic inbound for all the server.

I check with a sniff tool and i look that the traffic waste 200KB of constant traffic for the checks.

This traffic 24 hours is a lot of traffic, anyone with the same problem?? Is anyway to reduce this traffic? The autodiscovery is off.

Best Regards.

[nelsonab]

If you are referring to the actual protocol, I would argue that the Zabbix protocol is one of the most efficient due to it's simplicity. The easiest way to reduce your network load is to reduce the frequency with which you monitor your items. The default Linux template is overly aggressive with regards to timings. In most cases you can get by with monitoring items with a frequency of 30 seconds or more for CPU and Network, and 60 seconds for everything else.

[JBo]

Hi Nelsonab,

I agree with you on "default Linux template is overly aggressive" and reducing polling frequency to 30 or 60 seconds is a good start.

However, Zabbix protocol in passive mode opens a new TCP connection for each item, so several packets are exchanged for opening and closing the connection.

If several items are collected, switching to active mode may help since all items will be packed in one TCP connection.

Regards,
JBo