Troubleshooting Cisco IP Telephony

It's 5:30 a.m. on a Monday and your pager goes off. You recognize the phone number— it's your CEO's administrative assistant. As the administrator of the company's 8000-phone IP Telephony network, you assume there's a big problem. You rush into work and find the CEO's administrative assistant, who states that several calls for the CEO have been disconnected in the middle of the call, including a call from a very important customer. Where do you start?

Troubleshooting a Cisco IP Telephony network can be a daunting task. Rather than describing step-by-step how to solve specific problems (subsequent chapters provide that information), this chapter focuses on teaching a good troubleshooting methodology: learning how to find clues and track down your "suspect" by breaking the problem into smaller pieces and tackling each piece individually.

A typical IP Telephony network consists of—at the very least—one or more of the following components:

• Cisco CallManager servers
• IP phones
• Voice gateways

These components are in addition to the data network infrastructure that supports voice over IP (VoIP) traffic. More-complex installations can have dozens of servers for different services and redundancy, each server running a variety of applications, as well as hundreds or thousands of IP phones and a large number of voice gateways.

Before exploring the myriad of tools, traces, and techniques available to you that aid in troubleshooting, you must develop a systematic method by which you can focus on the problem and narrow it down until you determine the root cause.

In addition to the information in this book, you should become familiar with the various standard protocols that are used in an IP Telephony network, such as the following:

• H.323

• Media Gateway Control Protocol (MGCP)

• Telephony Application Programming Interface/Java Telephony Application Programming Interface (TAPI/JTAPI)

You should also become familiar with the protocols used when interfacing with the traditional time-division multiplexing (TDM)-based Public Switched Telephone Network (PSTN), such as the following:

• Q.931 (an ISDN protocol)

• T1- or E1-Channel Associated Signaling (T1-CAS or E1-CAS)

• Foreign Exchange Office (FXO)

• Foreign Exchange Station (FXS)

Additionally, because an IP Telephony network runs over a data network, it is important to understand the protocols that transport VoIP data, such as the following:

• Internet Protocol (IP)

• Transmission Control Protocol (TCP)

• User Datagram Protocol (UDP)

• Real-Time Transport Protocol (RTP)

Later chapters cover some of these concepts. However, each of the mentioned protocols could take up an entire book on its own, so you should refer to the specifications and RFCs or to other materials that go into detail about these protocols. Appendix A, "Cisco IP Telephony Protocol and Codec Information and References," provides references to where you can find additional information for each protocol discussed in this book.

On the other hand, because the Skinny Client Control Protocol (SCCP or Skinny protocol, the Cisco-developed protocol that Cisco IP Phones use) is not the product of an industry-wide standards body, this book goes into additional detail about how this protocol works. Understanding the Skinny protocol is essential to understanding how the phone operates and how to troubleshoot problems with it. The Skinny protocol is covered in greater detail in Chapter 5, "IP Phones."

Developing a Troubleshooting Methodology or Approach

To track down a problem and resolve it quickly, you must assume the role of detective. First, you need to look for as many clues as you can find. Some clues lead you to additional clues, and others lead you to a dead end. As soon as you've got all the clues, you need to try to make sense of them and come up with a solution. This book shows you where to look for these clues and track down the problem while trying to avoid as many dead ends as possible.

Troubleshooting a problem can be broken down into two stages: data gathering and data analysis, although your analysis might lead you to collect additional data. The following list is a general guide for steps to take when troubleshooting an IP Telephony problem:

Production Versus Nonproduction Outages

Troubleshooting a problem can occur in one of two timeframes:

• During a scheduled outage window, such as when you're installing a new system, adding components, or upgrading for new features or functionality

• During production hours when the problem affects end users or service

Although the methodology to troubleshoot problems in either of these two situations is similar, the focus on how to resolve the problem should be different. In the case of a service-affecting problem during production hours, the focus should be to quickly restore service by either resolving the problem or finding a suitable workaround.

In contrast, when a problem is found during a new install or scheduled outage window, the focus should be on determining the root cause to ensure the problem is completely diagnosed and resolved so that it does not have the potential to become service-affecting.

For example, if users are encountering a delayed dial tone or sluggish behavior on their phones, you might discover that a high-level process on CallManager is consuming 100 percent of the CPU on one of the servers. During a new install or scheduled outage window, it's a good idea to investigate what is causing the CPU consumption to ensure that the problem does not return during production hours.

However, if this problem occurs during production hours, the best approach is to stop or restart the offending process and let the redundant systems take over to quickly restore service. After you restore service, perform a root-cause analysis to try to determine why that process was consuming the CPU. The downside of this approach is that you might not be able to further troubleshoot the problem when the process is restarted. Fortunately, CallManager provides many diagnostic traces (if they are enabled prior to the problem) that you can reference after a problem has occurred to see what was happening on CallManager at the time of the problem.

Note that although 100 percent CPU of a high-level process can cause sluggish behavior or delayed dial tone, do not infer from this that 100 percent CPU is necessarily always a bad thing. As of CallManager 3.3(1), low priority tasks (such as phone registrations) can con-sume 100 percent CPU without causing adverse effects to the ability to place or receive calls. Look at the 100 percent CPU as a possible symptom but not necessarily the root cause. In this case, you observe the symptoms of sluggish or delayed dial tone and 100 percent CPU utilization and make a correlation between the two.

If you encounter an event where you are unable to determine the root cause due to insuf-ficient information, it is a good idea to turn on the appropriate traces to ensure that if the problem reoccurs, you will have enough data to identify the root cause.

Sometimes, several service-affecting problems occur simultaneously. In fact, this is not uncommon, because multiple problems often manifest themselves as symptoms of the same root cause. When multiple problems occur simultaneously, focus on the problem that has the greatest impact on users. For example, if some users are reporting dropped calls and others are reporting occasional echo, the two problems are probably unrelated. Trouble-shoot the dropped-call problem first because keeping calls connected is more critical than removing the occasional echo on an active call.

Step 1: Gathering Data About the Problem

So you've just installed a new IP Telephony network, or you've been given the task of maintaining one—or maybe you've taken your first CallManager out of the box and are having problems getting it to run. You've encountered a problem. The first thing to do is gather as much information about the problem as possible.

Identifying and Isolating the Problem

Half the battle in troubleshooting a problem is determining which piece of the puzzle is the source of the problem. With so many different pieces composing an IP Telephony network, the first step is to isolate the problem and, if multiple problems are being reported, deter-mine which of the problems might be related to each other and which should be identified as separate problems.

You must also determine which parts of the problem are symptoms and which are the root cause of the problem. For example, if a user complains of a phone resetting itself, it might seem logical to first assume that something is wrong with the phone. However, the problem might lie with CallManager or one of the many routers and switches that make up the un-derlying data network. So although the symptom is a phone reset, the root cause could be a WAN network outage or CallManager failure. You must always remember to look at the big picture when searching for the root cause and not let the symptoms of the problem lead you in the wrong direction. To help you visualize the big picture, detailed topology information is essential.

Using Topology Information to Isolate the Problem

You can take many proactive steps to help make the troubleshooting process easier. One of the first lines of defense is possessing current topology information. One of the most im-portant pieces of topology information is a detailed network diagram (usually created using Microsoft Visio or a similar application). The network diagram should include network addressing information and the names of all the devices. It should also clearly show how the devices are interconnected and the port numbers being used for these interconnections. This information will prove invaluable when you try to isolate which components are involved in a particular problem.

For medium- to larger-sized networks, you should have a high-level overview topology that gives you a general idea of how things are connected and then several more-detailed dia-grams for each piece of the network that drill down to the interface level on your network devices.

Figure 1-1 shows a typical high-level topology diagram for a large enterprise IP Telephony network. Notice that device names and IP addresses are listed in the diagram. This makes troubleshooting easier by allowing you to quickly look up devices to access them. Because Figure 1-1 is a high-level diagram, it does not get down to the interface level of each device.

Most networks are not as large as the one shown in Figure 1-1. However, no matter the size of your network, a similar topology diagram is very useful for quickly sharing information about your network with others who might be assisting you in troubleshooting.

In addition to the network diagram, you should use some method to store information such as IP address assignments, device names, password information, and so on. For a small network, you can use something as simple as a spreadsheet or even a plain text file. For larger deployments, some kind of database or network management application such as CiscoWorks is recommended. Many customers keep all this topology information on a web server as well, making it quickly and easily accessible to others when it is needed the most. Be sure to keep this information in a secure location.

You also need documentation of your dial plan. Some deployments, especially those heavily utilizing toll-bypass, have very complex dial plans. Knowing where a call is supposed to go just by knowing the phone number and from where it is dialed helps you quickly understand a problem.

Figure 1-1 Sample High-Level Topology Diagram

When your topology information is complete, it should include all the following information:

• Interconnection information for all devices, including device names and port num-bers. If any patch panels exist between devices, the port numbers should be listed.

• IP addressing for all network devices (routers, switches, and so on)

• IP addressing for all telephony and application servers and voice gateways (including data application servers)

• IP addressing for endpoints (that is, scopes of a DHCP pool)

• WAN and PSTN service provider names and Circuit IDs for each circuit

• Spanning-tree topology, including root bridges for all VLANs and which ports should be forwarding and blocking

• Dial plan information

• Software version information for all devices

If you are troubleshooting a network you didn't design, topology is one of the first pieces of information you should obtain, if it's available. If a topology drawing is not available, it is a good idea to spend time obtaining this information from someone who is familiar with the network and then making a quick sketch. A general topological understanding of the network or at least the piece of the network in question helps when you're trying to dif-ferentiate the problem from its symptoms. It's necessary when you're trying to isolate the problem to a particular part of the network.

For example, if a user reports hearing choppy audio when making a conference call, it is essential to know exactly where in the network the conference bridge device is located in relation to the user's phone, including all the intermediate network devices. Without a net-work diagram, finding this information could waste precious time. Assume that the network you are troubleshooting looks like Figure 1-1. If the user's phone is connected to Access Switch 1A, the other conference participants are on Access Switch 1Z, and the conference bridge device is on Voice Switch 1A, you can see that the number of devices is greatly reduced from 100 or more switches and routers to four or five.

What is worse than not having topology information? Having incorrect topology infor-mation can lead to countless hours heading down the wrong path. If you're going to keep topology information (highly recommended), make sure you keep it current.

Use all the topology information you have to narrow down which pieces of the network might be involved in the problem you are trying to troubleshoot. To further isolate the problem, interview the end users who reported the problem to gather additional information.

Gathering Information from the User

Information the user provides can be vital to your ability to correct a problem. Try to gather as much detail as possible on exactly what the problem is. Often when troubleshooting a problem, you might realize that what you've been troubleshooting for hours is not really the problem the user encountered. The more detail about the problem you can gather before you begin troubleshooting, the easier it is to find a resolution—and that means less frus-tration for you. Here is some general information to collect from users:

• Details about exactly what the user experienced when the problem occurred.

• Phone numbers for all parties involved in the problematic call or calls. You can use this as search criteria if you need to look through traces.

• Actions performed by the user when the problem occurred. This includes what buttons were pressed and in what order.

• User observations. This includes text messages displayed on the phone or recorded announcements.

• Information about the user's device. For example, if the user experienced a problem while using a 7960 phone, get the phone's MAC address and IP address, along with registration information and any other statistics available from the phone.

Sometimes the information provided by an end user is not enough to even begin trouble-shooting. For example, if a user has trouble transferring calls, you should ask what steps the user took when the problem happened and, if possible, when the problem occurred so that you can examine traces. Sometimes the proper diagnostic tools are not enabled when the problem occurs, forcing you to ask the user to inform you the next time the problem occurs. Be sure to turn on tracing or debugs before making the request so that when the problem occurs again, you will have captured the data. Users can get quite irritated if you have to ask them for the same piece of information two or three times. Also point out to the user the importance of letting you know immediately after a problem occurs, as many of the diagnostic trace files overwrite themselves within several hours or days (depending on the amount of traffic on your system).

Determining the Problem's Timeframe

In addition to what the problem is, you should try to determine when the problem occurred. Determining the problem's earliest occurrence can help correlate the problem with other changes that might have been made to the system or other events that occurred around the same time. For example, assume that a regular workday begins at 9 a.m. and ends around 6 p.m. Many users report that they get a busy signal when dialing into their voice mail. It is important to know whether they are attempting to do this at 9:10 a.m., a time when the voice mail system is likely under attack from many users all trying to access the system at once. This might change the problem from a troubleshooting issue to a load-balancing or equipment-expansion issue. You check the voice mail system and notice that at the time the problem was reported, all the voice mail ports were in use. Clearly in this example you need more voice mail ports or servers to handle call volume. However, if the problem occurs at 10:30 p.m., capacity is likely not the problem, so it's time to start troubleshooting your network and voice mail system. As another example, if a user reports that her phone was not working for 10 minutes and you know there was a network outage in her part of the building at that time, you can be relatively sure that the problem was due to the network outage.

When relying on end users to give "when" information about a problem, ask them to note the time on their phone when the problem occurred. The phone's time is synchronized with the clock on the CallManager to which the phone is registered. As long as you have the time on your CallManagers and network devices synchronized, having a phone-based time from the user makes finding the proper trace files very easy.

In some cases, the information about when a problem occurred might be the only piece of information you have other than a limited description of the problem at hand. If you have information about when, you might be able to look through trace files during that timeframe to search for anything abnormal.

Step 2: Analyzing the Data Collected About the Problem

Now that you have collected data from a variety of sources, you must analyze it to find the root cause and/or workaround for your problem.

Using Deductive Reasoning to Narrow the List of Possible Causes

The next part of your fact-finding mission is to identify the various components that might be involved and to eliminate as many components as possible. The more you can isolate the problem, the easier it is to find the root cause. For example, if a user complains about choppy voice quality, consider some of the following questions to help isolate the real problem, and think about how the answer will help narrow your focus:

• Does the problem happen on only one phone? If so, you can probably eliminate hundreds or thousands of other phones as suspects. However, keep in mind a single user's perspective. He might think the problem happens only on his phone, so you'll have to ask other users to see if the problem is more widespread than a single phone.

• What numbers are being called when the problem occurs? The answer to this question helps determine which parts of the system are being used when the problem occurs. For example, if the user never experiences poor audio quality when calling certain numbers but always experiences it when calling other numbers, this is a big clue.

• Does the problem happen only between IP phones, only through one or more voice gateways, or both? The user probably won't know the answer, but you'll be able to answer this question yourself after you answer the preceding question about which numbers are being called when the problem happens.

You will find more detailed questions similar to these throughout this book when troubleshooting particular problems.

Although not all of the following apply to every problem, where applicable, you must check all of the following pieces involved in the call. Use your topology information to help obtain this information.

• CallManager nodes involved in the signaling

• Network devices that signaling and/or voice traffic traverse

• Gateways or phones involved in the call

• Other devices involved, such as conference bridges or transcoders

Concentrate your energy on the smallest subset of devices possible. For example, if all the users on a particular floor are having the same problem, concentrate on the problem a particular user is having. If you fix the problem for that one user, in most cases you fix it for all the affected users.

Verifying IP Network Integrity

One thing that people often forget is that your IP Telephony network is only as good as your IP network. A degraded network or a network outage can cause a wide range of problems, ranging from slight voice quality problems to a total inability to make or receive calls on one or more phones. The network is always a consideration when you encounter certain problems, so network health issues are covered throughout this book. Network health is especially important during the discussion of voice quality problems in Chapter 7, "Voice Quality," because most voice quality problems stem from packet delay and/or loss.

Always remember to keep the IP network in mind and look at every layer in the OSI model, starting from Layer 1. Check your physical layer connectivity (cables, patch panels, fiber connectors, and so on). Then make sure you have Layer 2 connectivity by checking for errors on ports, ensuring that Layer 2 switches are functioning properly, and so forth. Continue working your way up the stack until you reach the application layer (Layer 7). As an example, two of the most common reasons for one-way audio (where one side of the conversation cannot hear the other) are the lack of an IP route from one phone to another and the lack of a default gateway being configured on a phone. Taking the layered approach, you would first check the cabling and switches to make sure that there are no errors on the ports. You would then check Layer 3, the network layer, by ensuring that IP routing is working correctly. When you reach this layer, you discover that for some reason the IP packets from one phone are unable to reach the other phone. Upon further investigation, you might discover that there was a missing IP route on one of the routers in the network or a missing default gateway on one of the end devices (such as an IP phone or voice gateway).

Determining the Proper Troubleshooting Tool

After you narrow down the appropriate component(s) causing a problem and have detailed information from the user(s) experiencing the problem, you must select the proper tool(s) to troubleshoot the problem. Most components have multiple troubleshooting tools avail-able to help you. Chapter 3, "Understanding the Troubleshooting Tools," provides more details about some of the tools available for troubleshooting CallManager. You should use the tracing and debugging facilities available in CallManager and other devices to deter-mine exactly what is happening. Additional tools and traces are covered in the chapter associated with diagnosing certain types of problems. For example, Chapter 6, "Voice Gateways," covers debugging Cisco IOS Software voice gateways. Because CallManager is central to almost all problems, information about various portions of the CCM trace facilities appears throughout this book.

This step is the most demanding on your troubleshooting skills because you analyze the detailed information provided in the various tools and use it to search for additional clues using other tools. Sometimes the problem description you have is not detailed enough to determine which tool to use. In this case, you should try various tools in search of anything that looks out of the ordinary.

The following case study shows how this troubleshooting methodology works in a real-world scenario.