Troubleshooting

A key skill for anyone doing operations, is the ability to successfully troubleshoot problems.

Methodologies

Here we will got over a few steps you can take to help quickly narrow down problems to their causes.

Triaging with differential diagnosis

What is broken? First think about how it works in most basic terms. Then build on that the things which can break. Then from those, pick the ones that could cause the symptoms you see.

Example:

You cannot ping a server.

Preparing your toolbelt

You have a variety of tools at your fingertips to help work out the cause of a problem. Over time you will expand what is in your toolbelt, but to start with you must know how to use each of these:

  • top, vmstat, iostat, systat, sar, mpstat help you see the current state of the system - what is running, what is using cpu, memory? Is the disk being heavily used? There is a great deal of information, and knowing how these tools work will help you pick out the bits you should focus on.

  • tcpdump, ngrep If you suspect you have a network-related problem, tcpdump and ngrep can help you confirm it.

Walk through of a diagnosis

  • Eliminating variables

  • What changed recently?

  • Could any of the symptoms be red herrings?

  • Common culprits (is it plugged in? is it network accessible?)

  • Look through your logs

  • Communicating during an outage

  • ‘Talking Out-Loud’ (IRC/GroupChat)

  • Communicating after an outage (postmortems)

Recent changes

Often problems can be traced back to recent changes. Problems that start around the time of a change aren’t usually coincidence.

Learning common errors

Over time you may find that a small set of errors cause a large portion of the problems you have to fix. Let’s cause some of these problems and see how we identify and fix them.

Cannot bind to socket

There are two common reasons that you can’t bind to a socket: the port is already in use, or you don’t have permission. As an example, you can see what happens when I try to start a Python SimpleHTTPServer on a port that is already in use:

user@opsschool ~$ python -m SimpleHTTPServer 8080
...
socket.error: [Errno 98] Address already in use

Here’s an example of what happens when I try to bind to a privileged port without proper permissions (in Linux, ports < 1024 are privileged):

user@opsschool ~$ python -m SimpleHTTPServer 80
...
socket.error: [Errno 13] Permission denied

Permission denied reading from / writing to disk

Out of disk space

(finding large files, and also finding deleted-but-open files)

Mystery problems and SELinux

Out of inodes

Manifests as “disk full” when df claims you have disk space free.

Working effectively during a crisis

Being able to work successfully through a crisis is crucial to being a good operations person. For some it is a personality trait, but it can certainly be learned and is almost a requirement for many employers.

A very important skill to learn is the ability to remain calm in the face of disaster. It’s not always easy, especially with a client on the phone, but panicking will only make a situation worse. Yes, the most critical server in the infrastructure may have just completely failed without a backup. Instead of focusing on what will happen as a result of the crisis, focus on what needs to be done to bring the system back up. Deal with the results later, after fixing the immediate failure. The fallout of the crisis might be terrible, but it will almost certainly be worse if the immediate problem isn’t fixed. A calm mind can carefully analyze a situation to determine the best solution. Panic responses do not benefit from the same calculating rationality.

Different people will adapt to handling crisis situations in different ways. Some will adopt the detached, analytical calm of a surgeon. Others will take a few deep breaths to calm themselves before digging in to analyze the problem. The ability to stay calm in the face of disaster is more important than the method by which calm is achieved. It will take practice to reach the point of reacting to a disaster calmly.

Avoid placing blame. It doesn’t accomplish anything beyond creating animosity and tension when a team most needs cohesion and efficiency. While a good practice in general, it is even more important to resist the urge to point fingers during a crisis. It doesn’t assist in solving the problem, which is the top priority. Everything else is secondary.

The Importance of Procedure

Creating procedures for responding to disasters provides both a checklist of things to do in the given situation as well as a structured way to practice responding to the situation. The practice serves to solidify understanding of how to react, while the procedure itself provides a target of mental focus during an actual disaster. Adhering to the procedure ensures the steps taken to resolve a crisis are well-known and tested. Focus on the procedure to the exclusion of everything else.

That said, not every situation will have an associated procedure. These situations call for their own procedures. Try to create a procedure for every situation that doesn’t already have one. This diligence pays off over time, as history tends to repeat itself. In addition to this, a procedure for situations lacking a procedure provides a safety net when everything else fails. This will differ from one organization to the next, but the value is constant.

Like backups, no disaster recovery procedure is useful unless and until it is tested. Thorough testing and practicing–in a real environment if possible–quickly finds problems that will happen in the real world. Beyond having procedures for known possible failures, a procedure for situations other procedures do not cover provides a fallback for what to do in the inevitable unpredictable crisis.

In addition to the technical sector, other industries deal regularly with crisis response–fire fighters, law enforcement, paramedics. These organizations have their own procedures. These industries all predate technology, offering much to learn.

Non-technical skills

Situational Awareness (Mica Endsley) Decision Making (NDM and RPD) - Klein Communication (Common ground, Basic Compact, Assertiveness) Team Working (Joint Activity, fundamentals of coordination and collaboration) Leadership (before, during, after incidents) (Weick, Sutcliffe work on HROs) Managing Stress Coping with Fatigue Training and Assessment Methods Cognitive Psychology concerns (escalating scenarios, team-based troubleshooting)