Grafana released their open source distribution of the OpenTelemetry Collector (OTel) on April 9, 2024, named Grafana Alloy. Alloy uses the same components, code, and concepts that were first introduced in Grafana Agent (released in 2020). However, Grafana Agent and Agent Operator are now deprecated as of April 9, 2024, and have entered into Long-Term Support (LTS) whereby they will continue to receive bug/security fixes.
Why the need for Grafana Alloy? For one it allows for native ingest of metrics, logs, traces, and profiles - but it stresses compatibility with a large ecosystem including Prometheus, OpenTelemetry, Mimir, Loki, Tempo, and Pyroscope. It also simplifies operating at scale with features such as an embedded debugging UI viewable from the Alloy HTTP server, native clustering for production, and built-in Hashicorp Vault support for enhanced security.
Alloy can collect telemetry from many sources, including OTel, and then transform/enrich/filter data in pipelines before sending to final destinations. Alloy can even send alerting rules with the data for processing.
Alloy is a continuation of Grafana's 'big tent' strategy and compliments an already robust ecosystem from the observability leader > Loki (log management), Grafana (visualization), Tempo (traces), and Mimir (metrics).
What's more, Alloy supports profile routing as well (Pyroscope).
Concerning continuous profiling, OTel is merging its profiling data model into the OTel specification and working towards a stable implementation this year (OTEP). Whether you wait for this implementation or dive in with Pyroscope, profiling has an incredible promise. Consider the following use cases:
Metrics to profiles: Move from infrastructure resource usage spikes to the pieces of the code that are consuming the resources.
Traces to profiles: Move beyond latency measurements across your services and into the code that is responsible.
Logs to profiles: Again, go beyond generic log errors into increased granularity
Track OTel (OTEPs) and Grafana closely to see where and how you can evolve your observability strategy.
