We all know log files. We all use log data. At a minimum, every admin and developer knows how to fire up tail –f and use an arsenal of command line tools to dig into a system's log files. But the days where those practices would suffice for operational troubleshooting are long gone. Today, you need a solid log strategy.

Log data has become big data and is more relevant to your success than ever before. Not being able to manage it and make meaningful use of it can, in the worst case, kill your business.

You might have implemented good application monitoring, but that only tells you that something is happening, not why. The information needed to understand the why, and the ability to predict and prevent it, is in your log data. And log data has exploded in volume and complexity.

Why this explosion? The commoditization of cloud technology: one of the greatest paradigm shifts the tech industry has seen over the recent years. Cloud services like Amazon's AWS, Microsoft's Azure, or Rackspace have made it affordable even for small- and medium-sized businesses to run complex applications on elastic virtual server farms. Containers running microservices are the next step in this move toward distributed and modularized systems.

The downside is that complexity is multiplied in those environments. Running tens or hundreds of machines with many different application components increases the risk that one of them will start malfunctioning.

To allow for troubleshooting, each of these many components typically (and hopefully) writes log data. Not only do you have to deal with a staggering number of large log files, but they're also scattered all over your network(s).

To make things a bit more interesting, some components, like VMs or containers, are ephemeral. They're launched on demand, and take their log data with them once they're terminated. Maybe the root cause slowing down or crashing your web store was visible in exactly one of those lost log files.

If that's still not complex enough, add in that people mix different technologies – for example, hybrid clouds that keep some systems on-premise or in a colo. You might run containers inside of VMs or use a container to deploy a hypervisor. Or you also could need to collect data from mobile applications and IoT devices.

Your log management solution needs to be able to receive and aggregate the logs from all your systems and components and store them in one central, accessible place. Leave no log behind – including the ones from ephemeral systems.

Some log management solutions require installing agents to accomplish this, while others are agentless and use de-facto standards like syslog, which are part of copious systems that allow sending logs over the network. Using agents means it's vital to make sure they are available for all operating systems, devices, and other components. You'll also need strategy to keep the agents updated and patched.

Fortunately, there is a checklist of the must-haves when it comes to log management to help you choose and sustain the best practices for your data, which I'll be sharing in my next post.

Read The No-BS Guide to Logging - Part 2

Sven Dummer is Senior Director of Product Marketing at Loggly.