feature request: ping out of the box - who is not using ping for network monitoring?

[xabbiz]

https://support.zabbix.com/browse/ZBXNEXT-739

At least the agent should simply be able to open a network connection - this is still something else than [what ping does|https://superuser.com/a/279132], but not having such a feature at all is like driving a car that has no lights.

I can not express my astonishment and disappointment about having to manually fiddle for such a basic thing strong enough - however I am (hopefully) still able to reflect myself, so I would like to ask: is pinging across hosts not part of any standard monitoring setup in the world of zabbix users? What are you doing instead? How are you monitoring your network infrastructure else without ping?

Zabbix devs are missing another low hanging fruit here: such a small thing, but such a big plus if it would be available by default - one less item on the never ending todo list for admins, more time for important things, not having to fiddle to achieve the basics.

I give you one more lovely feature suggestion here that will distinguish zabbix from other tools in a sensational way: just think one step further and enable pinging between hosts by default for all hosts and let zabbix draw nice network connectivity graphs by default! Of course that should be configurable for every host group and with variations like "ping only between host groups" and "define global ping target", "ping discovered hosts by default" etc. - well, you get the idea, the details of course must be worked out. That would be the opposite of the current situation: instead of forcing admins into digging into micro-details of configuration for very basic functionality you just could give admins a complete overview of their network connectivity status just out of the box!

This kind of feature-oriented thinking would be very appreciated!

[steeladept]

I am still somewhat new to Zabbix, but we never use ping. Why would we? You use it to troubleshoot and then what? It tells you if there is a network connection or not. That is it. Nothing else. Nada. No state, no details on why it isn't connecting. Nothing. You know what else works when it is connected? The agent. SNMP. SSH. In other words, pretty much everything.

Ping is critical in troubleshooting to see if there is a connection, there is no doubt or argument there. Once the fact is established, however, its value is over. There is no need to monitor it because you already know the network connectivity exists. If it goes down, ANY other tool that connects even once can tell you more, and will prove more valuable once you know you have a connection. Moreover, ping is chatty and creates a lot of network overhead in larger environments. If I am going to use a connection that is that noisy, I want it to provide information that is useful so the signal to noise is worth while. Therefore, once I have a connection setup, I use the agent, SNMP, or whatever connection I am monitoring to tell me when the connection goes down. IF it doesn't recover and I have to investigate, I may use ping then, but I would do that, not Zabbix.

Anyway, that is my 2 cents. Others may have other points of view for it.

[kloczek]

Not to mention that in many networks ICMP is limited by separated set of rules like max bandwidth or packets prioritisation.

[Mike2K]

https://support.zabbix.com/browse/ZBXNEXT-739
I give you one more lovely feature suggestion here that will distinguish zabbix from other tools in a sensational way: just think one step further and enable pinging between hosts by default for all hosts and let zabbix draw nice network connectivity graphs by default! Of course that should be configurable for every host group and with variations like "ping only between host groups" and "define global ping target", "ping discovered hosts by default" etc. - well, you get the idea, the details of course must be worked out. That would be the opposite of the current situation: instead of forcing admins into digging into micro-details of configuration for very basic functionality you just could give admins a complete overview of their network connectivity status just out of the box!

I rather like that idea but you should really do your homework.
As mentioned, ping is the absolute lowest on the QoS ladder so any and all network traffic is given priority over ICMP. The results would fluctuate like crazy and tell you....nothing....

For the creation of network graphs you should use CDP (Cisco Discovery Protocol) this will tell you how your switches are connected to each other...