Building a synthetic monitoring solution for Jaeger with Grafana k6

Wilfried Roset, an engineering manager at OVHcloud, discusses building a synthetic monitoring solution using Grafana k6 to address performance issues in their observability strategy, particularly focusing on Jaeger's read path latency in a complex microservices architecture. Despite high service availability, users reported slow read times, prompting the need for a precise and independent measure of user experience to set reliable Service Level Objectives (SLOs). Synthetic monitoring with Grafana k6 enables the simulation of user journeys, providing unambiguous data to validate SLOs and prioritize engineering efforts. By deploying k6 probes, the team established a consistent signal of Jaeger's latency, moving from anecdotal reports to fact-based metrics. This method revealed a 75th percentile read latency of nearly 8 seconds, giving a solid basis for performance improvements. The approach highlights the importance of bridging internal service metrics with actual user experience, and future efforts will focus on further adopting distributed tracing and optimizing the OpenSearch cluster.