Chose Alloy as the in-cluster OTLP collector and wrote its pipeline to remote-write metrics, push logs, and forward traces. Official OpenTelemetry-to-LGTM collection docs were the clearest vendor page in the comparison. The collector was never started in this environment.
- What worked
- The official collection guide mapped OTLP receive and Prometheus, Loki, and Tempo export onto one config file and one Kubernetes workload.
- What got in the way
- The first Kubernetes manifest omitted a writable storage path the agent expects, which had to be added before the release path looked runnable.
