Is it over for metrics?

Metrics have traditionally played a crucial role in monitoring systems by providing quick insights into system health through pre-aggregated statistics that are cheap to store and query, particularly in environments with high data volumes. However, their limitations in handling high-cardinality data and lacking correlation with logs and traces have become apparent, often necessitating deeper dives into more detailed data sources for effective troubleshooting. As a result, there's a growing shift towards using structured events and traces, especially with advancements in columnar databases like ClickHouse, which allow for efficient storage and analysis of high-fidelity data, transforming metrics from a primary monitoring tool to a performance optimization layer. This evolution in observability tools enables more flexible investigations, allowing teams to define metrics post-incident rather than preemptively, aligning with the broader trend of moving from ETL to ELT processes, and paving the way for more dynamic, data-driven decision-making in system monitoring.