If your business depends on users completing transactions, driving blind is the fastest path to obsolescence. Here’s how to implement instrumentation, a fundamental SaaS business practice.
1. Always instrument anything that goes out of process.
It’s tempting not to instrument a call to your database because “it’s always fast,” but things that should never cause trouble can clue you into larger problems.
2. Always log unacceptable performance.
If you expect a system component to perform at a certain level, log when it falls outside that range. Think of it as logging “soft exceptions.”
3. If possible, log enough context so that you get the complete picture of what happened from a single log event.
Instead of logging the start and end of a process, log the end plus the elapsed time. This saves significant time when things go wrong.
4. Look at trends, not just data points.
Data trend tells a story, way more than any single data point. Loggly can prevent slowdowns, thanks to a Metrics class that gives us counts and elapsed times every 10 seconds for every pipeline in our system. Rumble Entertainment tracks error code trends.
Final words
Create logs that your log management solution can interpret rather than ones human can read. Machines don’t get tired eyes or carpal tunnel syndrome.
Best tip
Let us make log management less of a chore! Start a Loggly free trial.
–Jon Gifford, Chief Search Officer and co-founder, Loggly
Comments have been disabled for this post