Read a sample: 1. Read the Evidence Before Changing the Network
“The network is down” is a useful report, but it is not yet a diagnosis. A user may mean that a website will not open, a file transfer has stopped, or a wireless connection has disappeared. Each symptom deserves attention, but none identifies the failed component by itself.
The technician’s first job is to turn the report into a question that can be tested. Which device is affected? What destination is it trying to reach? Does the failure occur by name, by IP address, or both? Can another device on the same network perform the same task? The answers begin to define the failure domain: the portion of the path where the problem could reasonably exist.
This chapter establishes a repeatable way to investigate those questions. The introductory case involves a user who reports that a business application is unavailable. The path to the answer will move through a client, a gateway, and a name-resolution test—but the important skill is not memorizing the commands. It is learning what each result proves, what it leaves unresolved, and which test should come next.
A symptom is an observation. A diagnosis is an explanation supported by evidence.
When someone says, “The server is down,” record what they actually experienced before accepting the proposed cause. Perhaps a browser displayed an error when the user entered a hostname. That is a real symptom. It does not establish whether the server, a router, a DNS resolver, or the client application is responsible.
Ask focused questions:
- What were you trying to reach, and what happened instead?
- When did the problem begin? Was there a recent change or interruption?
- Is the problem limited to one application or destination?
- Are other users, devices, or locations affected?
- Has this exact task worked from this device before?
Avoid turning the interview into an interrogation. The user’s answers establish context and urgency; they do not need to contain technical terminology. If possible, observe the failure or record the exact message. “Cannot find the server” and “connection refused” may lead to very different tests.
Also establish the scope of impact. One affected client suggests a different starting point from an entire floor, although either situation can still have more than one cause. If a nearby colleague can perform the same task, that colleague’s device may become a known-good comparison. The comparison is valuable only when you understand its limits: two devices on adjacent desks might use different VLANs, DNS servers, wireless networks, or security policies.
A troubleshooting workflow is not a rigid sequence of commands. It is a way to keep observations, assumptions, and decisions separate.
Start by defining the symptom and its scope. Then map the relevant path at a useful level of detail: client, access network, gateway, routed network, name service, and destination. You rarely need the entire enterprise topology to begin, but you do need to know which components are shared by affected and unaffected users.
Next, collect evidence that can distinguish plausible causes. A client address and mask help establish its local network. A default gateway identifies where traffic for other networks is expected to go. A route table shows a device’s forwarding choices. A DNS query tests a different dependency from an IP reachability test.
Form a small number of hypotheses and choose a test that separates them. For example:
- If a client cannot reach its gateway, investigate local addressing, the access network, and the gateway interface before focusing on a distant server.
- If the gateway is reachable but a remote IP address is not, investigate the routed path, return path, and applicable controls.
- If a remote service works by IP address but not by hostname, investigate name resolution and the name being used.
These are starting points, not universal rules. A gateway can answer a ping while failing to forward some traffic. A destination can ignore ICMP echo requests while its application remains healthy. A successful test proves only that the tested transaction succeeded under the tested conditions.
Finally, make a controlled correction, verify the original task, and communicate the result. A change is not verified merely because a command was accepted or a status indicator turned green.
A user in the purchasing department reports that the inventory application will not open. The user says, “The inventory server is down,” and asks for it to be restarted.
The technician does not dismiss the report, but separates the observation from the conclusion. The user explains that entering inventory.example.test in a browser produces a message that the site cannot be found. The application worked earlier that morning.
A second employee in the same department can open the application. Both employees are connected to the wired office network. This comparison makes a complete server outage less likely, but it does not rule out every server-side issue: the users could still be reaching different addresses or receiving different application responses.
The technician records the relevant details:
- Affected client: PC-PUR-17
- Reported destination: inventory.example.test
- Observed symptom: the browser cannot find the site
- Known-good comparison: another purchasing client opens the application by the same name
- Recent context: the affected user says the application worked earlier
The environment’s documentation identifies the inventory service at 198.51.100.20. That address is useful for a controlled comparison, but the technician does not yet assume that the documentation is current or that IP reachability alone will prove…