Testing connectivity with ping, curl, traceroute, and MTR
Use several tools because each answers a different question. ICMP failure alone does not prove that a TCP application is down.
ping -c 4 SERVER_ADDRESS
curl -v --connect-timeout 10 https://host.example.com/
traceroute SERVER_ADDRESS
mtr -rwzc 50 SERVER_ADDRESS
Run tests in both directions when possible and record UTC timestamps. Loss shown only on an intermediate MTR hop is not conclusive when later hops answer normally; routers may deprioritise diagnostic traffic.
For a refused TCP connection, inspect the listening socket with ss -lntp. For a timeout, check routes, guest firewall rules, service bind addresses, and any upstream filtering.