Half a year ago, my team at Trade Republic fully migrated our observability stack from Datadog to LGTM (Loki, Grafana, Tempo, Mimir). Operations after migration are as important as the migration itself, involving ongoing challenges such as performance and scalability issues, bugs, and incidents. In this talk, I’ll share our experiences from the past six months, detailing the challenges we faced and the valuable lessons we learned while using Grafana tools. Join us to gain insights into the practical aspects of managing and optimising an observability stack in a dynamic environment.

When deciding for a metrics tooling, one of the implicit decisions we take is about the temporality: are we capturing, exporting, and storing metrics that send only the difference between states (deltas), or are we picking a solution that works by accumulating data instead? In this talk, Tom Braack will show how to migrate from delta-style backends to cumulative ones, with a special focus on doing translations at the Collector side. Join us on this session and be equipped with knowledge about how to move from delta backends to cumulative ones, knowing what are the tradeoffs and gotchas.