What metrics should be taken from switches l3, l2?

Question

Monitoring

Hello!
I'm just starting to deal with zabbix and monitoring in general. Therefore, there are several questions.
1. What are the main metrics to shoot?
2. How should I monitor access switches, monitor only trunk ports? Or all?

Answer 1

1. What are the main metrics to shoot?

The state of the switch itself: utilization of the CPU, memory, readings of temperature sensors, fan speed, input / output voltage of the power supply.
Network subsystem status: port utilization (bps, pps) with traffic separation (unicast, multicast, broadcast), number and type of errors/dropped frames, interface status (duplex, speed).
Generally speaking, I think it's better to have a lot of data and discard some if necessary, than to have not enough data and guess what's happening on the network. Therefore, I recommend collecting all the metrics that are available.

2. How should I monitor access switches, monitor only trunk ports? Or all?

See the previous paragraph, I think it's better to monitor all interfaces. Of course, it is better to display selected charts on the main page (dashboard).

Answer 2

The more you shoot and monitor, the better. However, do not overdo it - proceed from the capabilities of iron. Download the MIB from the manufacturers of your network equipment and create a template based on OIDs or use ready-made templates for zabbix (if available) and collect all the metrics. Over time, you will understand what you need and what you don't.
Well-configured triggers will help you analyze the collected information. With their help, it is quite possible to detect abnormal loading on network interfaces, for example.
A couple of months ago, Alexey Vladyshev (the creator of Zabbix) hosted free webinars. Look for them online. I strongly advise you to familiarize yourself with them - the most for beginners.

Answer 3

Everything needs to be monitored, metrics need to be written and analyzed. Any deviation from the statistics should be closely examined.