Storage and Observability Backends
Speedscale publishes reference Helm charts in the speedscale-byoc repository. Install one independent collector channel for each destination you need.
| Chart | Destination | Speedscale capture signal | Optional application signals | Direct proxymock import |
|---|---|---|---|---|
fluentbit-s3 | Amazon S3 through the OTel awss3 exporter | RRPair logs | None | Yes, proxymock import s3 |
gcs | Native Google Cloud Storage | RRPair logs | None | Yes, proxymock import gcs |
fluentbit-gcs | GCS through its S3-compatible API | RRPair logs | None | Yes, native GCS or S3 interoperability |
grafana | Loki and Grafana | RRPair logs | Other application logs when configured separately | Gather script |
elasticsearch | Elasticsearch and Kibana | RRPair logs | Other application logs when configured separately | Gather script |
azureblob | Azure Blob Storage | RRPair logs | None | No; use azure-gather.py |
datadog | Datadog | RRPair logs | Application traces and metrics sent independently; the Datadog connector can derive APM trace metrics | Trace recipe |
dynatrace | Dynatrace | RRPair logs | Application traces and metrics sent independently | No |
newrelic | New Relic | RRPair logs | Application traces and metrics sent independently | No |
The fluentbit-s3 name is historical; the chart writes OTLP JSON directly with the OpenTelemetry awss3 exporter. New GCS installations should use the native gcs chart. Keep the legacy fluentbit-gcs chart only for an existing S3-interoperability and HMAC workflow.
The Forwarder itself emits RRPairs as OTLP logs. A collector can accept application traces and metrics on additional pipelines, but those signals appear only when the application or another agent sends them independently. For Datadog, review the Datadog exporter and connector guidance before customizing the reference chart; Datadog recommends OTLP/HTTP and the span_metrics connector for new custom configurations.
Collector versions and support boundary
The published reference charts pin collector images instead of following a floating tag. The current Amazon S3, native GCS, Azure Blob, Datadog, Dynatrace, and New Relic charts pin OpenTelemetry Collector Contrib 0.160.0 by image digest. The legacy fluentbit-gcs and Elasticsearch reference charts remain on 0.108.0.
The upstream awss3, google_cloud_storage, and azureblob exporters are alpha. Speedscale validates and supports the collector image and configuration pinned by each published chart. Treat an image override, unpinned upgrade, or custom collector pipeline as a custom deployment and validate retention, retries, and retrieval before relying on it for production traffic.
Choose for replayability
Choose Amazon S3 or native GCS when developers and automated agents need to pull a time-bounded capture directly into proxymock. Run an object-storage channel alongside an observability channel when you need both operational correlation and a simple replay path.
Loki, Elasticsearch, and Azure Blob require backend-specific gather scripts. Datadog has a Datadog-to-proxymock recipe for retrieving one trace. Dynatrace and New Relic do not have direct proxymock importers.
Observability guides
Azure Blob limitation
Azure Blob Storage does not expose an S3-compatible API. Changing --s3-endpoint-url to an Azure Blob URL does not make proxymock import s3, the pull_byoc_bucket MCP tool, or the proxymock web source picker compatible with it.
For manual retrieval, use scripts/azure-gather.py. Choose S3 or GCS if direct proxymock bucket import is a requirement.
Next step
Configure BYOC on Kubernetes, or use the ECS/Fargate example for a non-Kubernetes deployment.