How to Build a Scalable Log Strategy

Log volume is a compounding problem, not a linear one. Every new service, team, and piece of infrastructure generates logs, and the growth accelerates. Most bad log strategies are inherited, and they likely made sense for the teams that designed them, until scale exposed what they were never built for.

What you'll learn:

• Why most bad log strategies get inherited, not designed
• How to build a tiered log strategy with Flex as the default, not the exception
• How to make retention decisions source by source, based on real query patterns
• Why pipeline design makes or breaks tiering, search, and access
• What to evaluate before you modernize or migrate

Where to start:

• Filter the noise you already know about: source side filtering for full exclusions, exclusion filters for partial
• Flip the default to Flex: a wildcard catch-all prevents every future source from silently landing in the most expensive tier
• Get pipeline coverage to zero gaps: run datadog.pipelines:false and build pipelines for the gaps