Skip to main content

Storage and Observability Backends

Speedscale publishes reference Helm charts in the speedscale-byoc repository. Install one independent collector channel for each destination you need.

ChartDestinationSpeedscale capture signalOptional application signalsDirect proxymock import
fluentbit-s3Amazon S3 through the OTel awss3 exporterRRPair logsNoneYes, proxymock import s3
gcsNative Google Cloud StorageRRPair logsNoneYes, proxymock import gcs
fluentbit-gcsGCS through its S3-compatible APIRRPair logsNoneYes, native GCS or S3 interoperability
grafanaLoki and GrafanaRRPair logsOther application logs when configured separatelyGather script
elasticsearchElasticsearch and KibanaRRPair logsOther application logs when configured separatelyGather script
azureblobAzure Blob StorageRRPair logsNoneNo; use azure-gather.py
datadogDatadogRRPair logsApplication traces and metrics sent independently; the Datadog connector can derive APM trace metricsTrace recipe
dynatraceDynatraceRRPair logsApplication traces and metrics sent independentlyNo
newrelicNew RelicRRPair logsApplication traces and metrics sent independentlyNo

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.