Home > Articles > Troubleshooting Processes for Complex Enterprise Networks

Troubleshooting Processes for Complex Enterprise Networks

Chapter Description

This chapter shows you how to troubleshoot principles and approaches, implement troubleshooting processes and integrate troubleshooting into the network maintenance process.

Implementing Troubleshooting Procedures

The troubleshooting process can be guided by structured methods, but it is not static, and its steps are not always the same and may not be executed in the exact same order every time. Each network is different, each problem is different, and the skill set and experience of the engineer involved in a troubleshooting process is different. However, to guarantee a certain level of consistency in the way that problems are diagnosed and solved in an organization, it is still important to evaluate the common subprocesses that are part of troubleshooting and define procedures that outline how they should be handled. The generic troubleshooting process consists of the following tasks:

  1. Defining the problem
  2. Gathering information
  3. Analyzing the information
  4. Eliminating possible problem causes
  5. Formulating a hypothesis about the likely cause of the problem
  6. Testing that hypothesis
  7. Solving the problem

It is important to analyze the typical actions and decisions that are taken during each of these processes and how these could be planned and implemented as troubleshooting procedures.

The Troubleshooting Process

A network troubleshooting process can be reduced to a number of elementary subprocesses, as outlined in the preceding list. These subprocesses are not strictly sequential in nature, and many times you will go back and forth through many of these subprocesses repeatedly until you eventually reach the solving-the-problem phase. A troubleshooting method provides a guiding principle that helps you move through these processes in a structured way. There is no exact recipe for troubleshooting. Every problem is different, and it is impossible to create a script that will solve all possible problem scenarios. Troubleshooting is a skill that requires relevant knowledge and experience. After using different methods several times, you will become more effective at selecting the right method for a particular problem, gathering the most relevant information, and analyzing problems quickly and efficiently. As you gain more experience, you will find that you can skip some steps and adopt more of a shoot-from-the-hip approach, resolving problems more quickly. Regardless, to execute a successful troubleshooting exercise, you must be able to answer the following questions:

  • What is the action plan for each of the elementary subprocesses or phases?
  • What is it that you actually do during each of those subprocesses?
  • What decisions do you need to make?
  • What kind of support or resources do you need?
  • What kind of communication needs to take place?
  • How do you assign proper responsibilities?

Although the answers to these questions will differ for each individual organization, by planning, documenting, and implementing troubleshooting procedures, the consistency and effectiveness of the troubleshooting processes in your organization will improve.

Defining the Problem

All troubleshooting tasks begin with defining the problem. However, what triggers a troubleshooting exercise is a failure experienced by someone who reports it to the support group. Figure 2-4 illustrates reporting of the problem (done by the user) as the trigger action, followed by verification and defining the problem (done by support group). Unless an organization has a strict policy on how problems are reported, the reported problem can unfortunately be vague or even misleading. Problem reports can look like the following: "When I try to go to this location on the intranet, I get a page that says I don't have permission," "The mail server isn't working," or "I can't file my expense report." As you might have noticed, the second statement is merely a conclusion a user has drawn perhaps merely because he cannot send or receive e-mail. To prevent wasting a lot of time during the troubleshooting process based on false assumptions and claims, the first step of troubleshooting is always verifying and defining the problem. The problem has to be first verified, and then defined by you (the support engineer, not the user), and it has to be defined clearly.

Figure 2-4

Figure 2-4 A Reported Problem Must First Be Verified and Then Defined by Support Staff

A good problem description consists of accurate descriptions of symptoms and not of interpretations or conclusions. Consequences for the user are strictly not part of the problem description itself, but can be helpful to assess the urgency of the issue. When a problem is reported as "The mail server isn't working," you must perhaps contact the user and find out exactly what he has experienced. You will probably define the problem as "When user X starts his e-mail client, he gets an error message saying that the client can not connect to the server. The user can still access his network drives and browse the Internet."

After you have clearly defined the problem, you have one more step to take before starting the actual troubleshooting process. You must determine whether this problem is your responsibility or if it needs to be escalated to another department or person. For example, assume the reported problem is this: "When user Y tries to access the corporate directory on the company intranet, she gets a message that says permission is denied. She can access all other intranet pages." You are a network engineer, and you do not have access to the servers. A separate department in your company manages the intranet servers. Therefore, you must know what to do when this type of problem is reported to you as a network problem. You must know whether to start troubleshooting or to escalate it to the server department. It is important that you know which type of problems is your responsibility to act on, what minimal actions you need to take before you escalate a problem, and how you escalate a problem. As Figure 2-4 illustrates, after defining the problem, you assign the problem: The problem is either escalated to another group or department, or it is network support's responsibility to solve it. In the latter case, the next step is gathering and analyzing information.

Gathering and Analyzing Information

Before gathering information, you should select your initial troubleshooting method and develop an information-gathering plan. As part of this plan, you need to identify what the targets are for the information-gathering process. In other words, you must decide which devices, clients, or servers you want to collect information from, and what tools you intend to use to gather that information (assemble a toolkit). Next, you have to acquire access to the identified targets. In many cases, you might have access to these systems as a normal part of your job role, but in some cases, you might need to get information from systems that you cannot normally access. In this case, you might have to escalate the issue to a different department or person, either to obtain access or to get someone else to gather the information for you. If the escalation process would slow the procedure down and the problem is urgent, you might want to reconsider the troubleshooting method that you selected and first try a method that uses different targets and would not require you to escalate. As you can see in Figure 2-5, whether you can access and examine the devices you identified will either lead to problems escalation to another group or department or to the gathering and analyzing information step.

Figure 2-5

Figure 2-5 Lack of Access to Devices Might Lead to Problem Escalation to Another Group

The example that follows demonstrates how information gathering can be influenced by factors out of your control, and consequently, force you to alter your troubleshooting approach. Imagine that it is 1.00 p.m. now and your company's sales manager has reported that he cannot send or receive e-mail from the branch office where he is working. The matter is quite urgent because he has to send out a response to an important request for proposal (RFP) later this afternoon. Your first reaction might be to start a top-down troubleshooting method by calling him up and running through a series of tests. However, the sales manager is not available because he is in a meeting until 4:30 p.m. One of your colleagues from that same branch office confirms that the sales manager is in a meeting, but left his laptop on his desk. The RFP response needs to be received by the customer before 5:00 p.m. Even though a top-down troubleshooting approach might seem like the best choice, because you will not be able to access the sales manager's laptop, you will have to wait until 4:30 before you can start troubleshooting. Having to perform an entire troubleshooting exercise successfully in about 30 minutes is risky, and it will put you under a lot of pressure. In this case, it is best if you used a combination of the "bottom-up" and "follow-the-path" approaches. You can verify whether there are any Layer 1–3 problems between the manager's laptop and the company's mail server. Even if you do not find an issue, you can eliminate many potential problem causes, and when you start a top-down approach at 4:30, you will be able to work more efficiently.

Eliminating Possible Problem Causes

After gathering information from various devices, you must interpret and analyze the information. In a way, this process is similar to detective work. You must use the facts and evidence to progressively eliminate possible causes and eventually identify the root of the problem. To interpret the raw information that you have gathered, for example, the output of show and debug commands, or packet captures and device logs, you might need to research commands, protocols, and technologies. You might also need to consult network documentation to be able to interpret the information in the context of the actual network's implementation. During the analysis of the gathered information, you are typically trying to determine two things: What is happening on the network and what should be happening. If you discover differences between these two, you can collect clues for what is wrong or at least a direction to take for further information gathering. Figure 2-6 shows that the gathered information, network documentation, baseline information, plus your research results and past experience are all used as input while you interpret and analyze the gathered information to eliminate possibilities and identify the source of the problem.

Figure 2-6

Figure 2-6 Useful Factors That Can Feed and Support the Interpret and Analyze Task

Your perception of what is actually happening is usually formed based on interpretation of the raw data, supported by research and documentation; however, your understanding of the underlying protocols and technologies also plays a role in your success level. If you are troubleshooting protocols and technologies that you are not very familiar with, you will have to invest some time in researching how they operate. Furthermore, a good baseline of the behavior of your network can prove quite useful at the analysis stage. If you know how your network performs and how things work under normal conditions, you can spot anomalies in the behavior of the network and derive clues from those deviations. The benefit of vast relevant past experience cannot be undermined. An experienced network engineer will spend significantly less time on researching processes, interpreting raw data, and distilling the relevant information from the raw data than an inexperienced engineer.

Formulating/Testing a Hypothesis

Figure 2-7 shows that based on your continuous information analysis and the assumptions you make, you eliminate possible problem causes from the pool of proposed causes until you have a final proposal that takes you to the next step of the troubleshooting process: formulating and proposing a hypothesis.

Figure 2-7

Figure 2-7 Eliminating Possibilities and Proposing a Hypothesis Based on

After you have interpreted and analyzed the information that you have gathered, you start drawing conclusions from the results. On one hand, some of the discovered clues point toward certain issues that can be causing the problem, adding to your list of potential problem causes. For example, a very high CPU load on your multilayer switches can be a sign of a bridging loop. On the other hand, you might rule out some of the potential problem causes based on the gathered and analyzed facts. For example, a successful ping from a client to its default gateway rules out Layer 2 problems between them. Although the elimination process seems to be a rational, scientific procedure, you have to be aware that assumptions play a role in this process, too, and you have to be willing to go back and reexamine and verify your assumptions. If you do not, you might sometimes mistakenly eliminate the actual root cause of a problem as a nonprobable cause, and that means you will never be able to solve the problem.

An Example on Elimination and Assumptions

You are examining a connectivity problem between a client and a server. As part of a follow-the-path troubleshooting approach, you decide to verify the Layer 2 connectivity between the client and the access switch to which it connects. You log on to the access switch and using the show interface command, you verify that the port connecting the client is up, input and output packets are recorded on the port, and that no errors are displayed in the packet statistics. Next, you verify that the client's MAC address was correctly learned on the port according to the switch's MAC address table using the show mac-address-table command. Therefore, you conclude that Layer 2 is operational between the client and the switch, and you continue your troubleshooting approach examining links further up the path.

You must always keep in mind which of the assumptions you have made might need to be reexamined later. The first assumption made in this example is that the MAC address table entry and port statistics were current. Because this information might not be quite fresh, you might need to first clear the counters and the MAC address table and then verify that the counters are still increasing and that the MAC address is learned again. The second assumption is hidden in the conclusion: Layer 2 is operational, which implies that the client and the switch are sending and receiving frames to each other successfully in both directions. The only thing that you can really prove is that Layer 2 is operational from the client to the switch, because the switch has received frames from the client.

The fact that the interface is up and that frames were recorded as being sent by the switch does not give you definitive proof that the client has correctly received those frames. So even though it is reasonable to assume that, if a link is operational on Layer 2 in one direction it will also be operational in the other direction, this is still an assumption that you might need to come back to later.

Spotting faulty assumptions is one of the tricky aspects of troubleshooting, because usually you are not consciously making those assumptions. Making assumptions is part of the normal thought process. One helpful way to uncover hidden assumptions is to explain your reasoning to one of your colleagues or peers. Because people think differently, a peer might be able to spot the hidden assumptions that you are making and help you uncover them.

Solving the Problem

After the process of proposing and eliminating some of the potential problem causes, you end up with a short list of remaining possible causes. Based on experience, you might even be able to assign a certain measure of probability to each of the remaining potential causes. If this list still has many different possible problem causes and none of them clearly stands out as the most likely cause, you might have to go back and gather more information first and eliminate more problem causes before you can propose a good hypothesis. After you have reduced the list of potential causes to just a few (ideally just one), select one of them as your problem hypothesis. Before you start to test your proposal, however, you have to reassess whether the proposed problem cause is within your area of responsibilities. In other words, if the issue that you just proposed as your hypothesis causes the problem, you have to determine whether it is your responsibility to solve it or you have to escalate it to some other person or department. Figure 2-8 shows the steps that you take to reach a hypothesis followed by escalating it to another group, or by testing your hypothesis.

Figure 2-8

Figure 2-8 Formulating a Hypothesis Is Followed by Escalation or Testing the Hypothesis

If you decide to escalate the problem, ask yourself if this ends your involvement in the process. Note that escalating the problem is not the same as solving the problem. You have to think about how long it will take the other party to solve the problem and how urgent is the problem to them. Users affected by the problem might not be able to afford to wait long for the other group to fix the problem. If you cannot solve the problem, but it is too urgent to wait for the problem to be solved through an escalation, you might need to come up with a workaround. A temporary fix alleviates the symptoms experienced by the user, even if it does not address the root cause of the problem.

After a hypothesis is proposed identifying the cause of a problem, the next step is to come up with a possible solution (or workaround) to that problem, and plan an implementation scheme. Usually, implementing a possible solution involves making changes to the network. Therefore, if your organization has defined procedures for regular network maintenance, you must follow your organization's regular change procedures. The next step is to assess the impact of the change on the network and balance that against the urgency of the problem. If the urgency outweighs the impact and you decide to go ahead with the change, it is important to make sure that you have a way to revert to the original situation after you make the change. Even though you have determined that your hypothesis is the most likely cause of the problem and your solution is intended to fix it, you can never be entirely sure that your proposed solution will actually solve the problem. If the problem is not solved, you need to have a way to undo your changes and revert to the original situation. Upon creation of a rollback plan, you can implement your proposed solution according to your organization's change procedures. Verify that the problem is solved and that the change you made did what you expected it to do. In other words, make sure the root cause of the problem and its symptoms are eliminated, and that your solution has not introduced any new problems. If all results are positive and desirable, you move on to the final stage of troubleshooting, which is integrating the solution and documenting your work. Figure 2-9 shows the flow of tasks while you implement and test your proposed hypothesis and either solve the problem or end up rolling back your changes.

Figure 2-9

Figure 2-9 Testing a Proposed Hypothesis

You must have a plan for the situation if it turns out that the problem was not fixed, the symptoms have not disappeared, or new problems have been introduced by the change that you have made. In this case, you should execute your rollback plan, revert to the original situation, and resume the troubleshooting process. It is important to determine if the root cause hypothesis was invalid or whether it was simply the proposed solution that did not work.

After you have confirmed your hypothesis and verified that the symptoms have disappeared, you have essentially solved the problem. All you need to do then is to make sure that the changes you made are integrated into the regular implementation of the network and that any maintenance procedures associated with those changes are executed. You will have to create backups of any changed configurations or upgraded software. You will have to document all changes to make sure that the network documentation still accurately describes the current state of the network. In addition, you must perform any other actions that are prescribed by your organization's change control procedures. Figure 2-10 shows that upon receiving successful results from testing your hypothesis, you incorporate your solution and perform the final tasks such as backup, documentation, and communication, before you report the problem as solved.

Figure 2-10

Figure 2-10 The Final Step: Incorporate the Solution and Report the Problem as Solved

The last thing you do is to communicate that the problem has been solved. At a minimum, you will have to communicate back to the original user that reported the problem, but if you have involved others as part of an escalation process, you should communicate with them, too. For any of the processes and procedures described here, each organization will have to make its own choices in how much of these procedures should be described, formalized, and followed. However, anyone involved in troubleshooting will benefit from reviewing these processes and comparing them to their own troubleshooting habits.

3. Integrating Troubleshooting into the Network Maintenance Process | Next Section Previous Section