Isolate two simultaneous faults without guessing

applied · 55 min · Objective 5.1

Task

Diagnose a host with two independent faults at once, using divide-and-conquer to halve the search space and the discipline of changing one thing at a time. Two faults is where undisciplined troubleshooting falls apart, because fixing one does not clear the symptom and the temptation is to undo the fix and try something else.

Steps

  1. Identify and separate. List the distinct symptoms: name resolution fails, and some but not all local addresses are reachable. Treat them as two problems, not one — approach multiple problems individually, because they may or may not share a cause.
  2. Divide and conquer. Start in the middle at layer 3 rather than the bottom. Ping the gateway by IP: it answers, so layers 1 to 3 to the gateway are healthy and the "some hosts unreachable" fault is not a dead interface. One test removed half the candidates.
  3. Test the mask theory. Ping a local host that the current mask excludes, then compare the configured mask against the IPAM record. A mask that is too long makes the host treat some local addresses as remote. Fix the mask — only the mask — and confirm local reachability returns.
  4. Re-scope what remains. Name resolution still fails, but the address it should resolve to is now reachable by IP. That split — IP works, name does not — is the signature of a DNS fault, isolated cleanly because you changed one thing.
  5. Test and fix DNS. Query the configured resolver directly, see it not answer, point at a working resolver, and confirm resolution returns. Change one thing, verify, then stop.

Verify

ping -c 2 -W 2 192.168.10.10; echo "local host exit $?"
getent hosts router.lab || nslookup router.lab
dig +short router.lab @192.168.10.1; echo "dns query exit $?"

Each fault has its own proof. The local-host ping succeeding shows the mask is right; a successful name lookup shows DNS is fixed. Because you verified after each single change, you can say which fix cleared which symptom — which is the whole point of not changing two things at once.

Notes

The methodology names the trap directly: change one thing at a time, and change it back if it did not help. With two faults present, fixing the mask does not make the DNS symptom disappear, and an undisciplined technician reads that as "the mask fix did not work", undoes it, and now has an unrecoverable mess. The fix worked; it just was not the whole problem.

Divide and conquer earns its place here. Bottom-up would have you check cables and link lights first, which are fine; top-down would start at the application. Starting at layer 3 with a single ping to the gateway told you in one test that the lower layers were healthy, which is why it is usually fastest when you have no strong prior about where the fault sits.