When running multiple services on AWS ECS, all telemetry can collapse under `service.name = aws_ecs` because the OTel Collector's resource detection sets `cloud.platform` but not `service.name`, and some backends fall back to `cloud.platform` when `service.name` is missing. Two fix patterns are covered: for EC2-hosted ECS, use `resourcedetection/ec2` to read instance tags and an OTTL `transform` to promote `ec2.tag.service_name` to `service.name`; for Fargate, use `resourcedetection: [env, ecs]` combined with `OTEL_RESOURCE_ATTRIBUTES` in the task definition. Two additional pitfalls are addressed: resource-level attributes being invisible to log-record-level query filters (fix: copy resource attributes onto each log record via OTTL), and server-side remap rules silently overwriting a correctly-set `service.name` with `cloud.platform`. Production-tested YAML configs for both deployment shapes are included, along with verification steps using the debug exporter.
Sort: