Follow a failed connection from DNS to TCP, NAT, return routing, TLS, and application processing when a firewall log shows allow.
Define success at the application level
An allow log answers a policy question for recorded traffic. A user needs a complete service transaction. The connection can still fail at a later device, on the server, during authentication, or on its return path. Start with the operation that fails, the exact time, and the endpoint performing it.
Create a connection worksheet
| Field | Example to replace |
|---|---|
| Client and server | 10.10.10.25 and 10.40.20.50 |
| Service | TCP 443 and the application hostname |
| Original and translated tuple | Source and destination before and after NAT |
| Device context | Firewall, VDOM or virtual system, and active member |
| User-visible failure | Timeout, reset, TLS alert, or application error |
Test the layers in order
Resolve-DnsName app.example.com
Test-NetConnection app.example.com -Port 443 -InformationLevel Detailed
First, validate the destination
Check the DNS answer against the intended service address. Consider both A and AAAA records and record which address the failing application uses. A browser and a command-line test can select different paths. A working IP test alongside a failing hostname test directs attention to name resolution or hostname-dependent processing.
Next, follow TCP establishment
For TCP, look for the client SYN, the server SYN-ACK, and the final ACK. Compare captures near both sides of the firewall. A repeated SYN narrows the question to request delivery and the missing reply. A reset provides a different lead. Record which device sends it before changing a firewall rule.
Compare NAT and routing with the design
Identify the actual destination after translation and the source seen by the server. Confirm that replies return through the firewall or state-sharing cluster responsible for the session. Unexpected translations, overlapping subnets, or a different default gateway can all change this path. Use the same test tuple when correlating route lookups and session records.
Then inspect TLS and application behavior
If TCP completes, inspect TLS alerts, certificate trust, hostname matching, and any decryption or security inspection logs. If TLS completes, review HTTP responses, redirects, identity-provider access, and server logs. An application can rely on several additional hosts after its first connection. Trace the failing transaction and its dependencies.
Make a scoped, testable correction
Write down the evidence, the proposed cause, and the predicted result of the change. Save the previous setting. Change the route, translation, service configuration, or rule that the evidence identifies, then repeat the original operation. Broad temporary access rules can obscure which condition was actually responsible.
Verify and document
Confirm the application succeeds, compare the final traffic log and packet sequence, and test one representative unaffected service. Record what the allow log established and what later evidence revealed. That record prevents a future incident from restarting with the assumption that policy permission guarantees application success.