How to monitor Internet (ISP link) uptime / downtime with Dynamic IP

[cw.wang]

Hi, Im kinda new with Zabbix. Our organization plan to monitor total up to 100 sites from our HQ Data Center. But how can we achieved this without putting any additional zabbix agent / PC at every sites? Because every sites is using Dynamic IP address. Each sites have a Fortigate FW.

Thanks

[kloczek]

In monitored host interface you don't need to use IPs. You can use names. Such name doesn't need to be even DNS name.
In case of using passive agent setup prx/srv needs to know DNS name or IP of the host where is working zabbix agent.
In case active agent setup agent process asks prx/srv for the monitored metric configuration and it doing queries with the name specified in zabbix_agentd.conf::Hostname.
If this configuration field is not specified zabbix agent uses here what "hostname -f" command returns.
In other words, in case of the active agent, you can use your own namespace of the monitored hostnames independent from DNS.

Another solution, of course, is updating DNS entries after each IP change.

[cw.wang]

Thanks Kloczek. Actually we just need to monitor the Internet link via SNMP and ICMP. There will be no PC. I wonder where to set the hostname in the branch Firewall?

[kloczek]

What you've got is not fully clear so I'll be assuming some details.

My understanding is that those ~100 locations are guarded by the firewall which is not passing traffic from outside to the inside of those networks.
If it is still possible to establish connectivity from inside those networks to the zabbix server all that you need is just active proxy (zabbix_proxy.conf::ProxyMode=1) which will be querying srv for monitoring cfg data and in the same connections sending back batches of monitoring data sampled from agents connected to the proxy and/or SNMP agents.
You need to guarantee enough long buffer of the data stored on the proxy in case of connections issues (zabbix_proxy::UnreachablePeriod)
If all monitored hosts with SNMP agents are behind NAT in each isolated networks all those monitored locations can use even with the same names/IPs because srv distinguishes monitored host by path [<zabbi-prx-name>:]{<IP>|<hostname>}. Exact <zabbi-prx-name> needs to be set up in each zabbix_proxy.conf::Hostname, or in internal networks DNS servers such proxies must have the unique names not used in any other zabbix prx-GW.

Nevertheless, SNMP monitoring by SNMP protocol definition is always is passive monitoring, so in case of SNMP monitoring, you cannot use this trick with the active agent which I've described previously.
Even SNMP traps as kind of active communication on receiving by zabbix srv/prx are identified by source IP/hostname (not by hostname on SNMP trap sender side settings).

Other options: add VPN connections between all those networks. Because connectivities between srv all proxies may be time to time interrupted (even by VPN lost connection) in each isolated network should be at least one proxy (depends how many metrics NVPS will be flowing over single proxy.
Nevertheless active proxy with encryption of the traffic as it is now possible using Zabbix 3.4 is kind of 1:1 equivalent of the VPN so if you do not have already VPNs or you do not have plans to connect those locations using VPNs using the active proxy with encrypted traffic will provide the same level of privacy.

If in above picture/sketch something does not much to your situation you need to clarify those details

[cw.wang]

Thanks kloczek. Will try on this.