Vendor integrations
Kerniq applications export OpenTelemetry (OTLP) or expose Prometheus metrics. Every vendor below is reached through one of those; no vendor SDK is linked into the application, and switching vendors is a Collector configuration change.
Recommended topology: application → OTLP → OpenTelemetry Collector (sidecar, DaemonSet or gateway) → vendor. The Collector holds credentials, retries, fans out, samples and transforms; the application stays simple.
| Backend | Metrics | Traces | Logs | How |
|---|---|---|---|---|
| Prometheus | ✓ | scrape /metrics (Option A) or Collector prometheus / prometheusremotewrite exporter |
||
| Grafana Mimir | ✓ | Collector prometheusremotewrite or OTLP (/otlp) |
||
| Grafana Tempo | ✓ | OTLP, directly or through the Collector | ||
| Grafana Loki | ✓ | OTLP (http://loki:3100/otlp, Loki 3+) |
||
| Grafana | — | — | — | visualization only; queries the stores above |
| Jaeger | ✓ | OTLP (Jaeger accepts OTLP natively on 4317/4318) | ||
| Datadog | ✓ | ✓ | ✓ | Collector datadog exporter + datadog/connector, or the Datadog Agent's OTLP intake |
| Elastic / ELK | ✓ | ✓ | ✓ | OTLP to Elastic's managed OTLP endpoint or APM Server; or Collector elasticsearch exporter |
| Splunk | ✓ | ✓ | ✓ | Collector splunk_hec exporter (Splunk Enterprise/Cloud); Splunk Observability Cloud accepts OTLP |
| Sumo Logic | ✓ | ✓ | ✓ | Collector sumologic exporter, or OTLP/HTTP to a Sumo OTLP source |
| Azure Monitor / Application Insights | ✓ | ✓ | ✓ | Collector azuremonitor exporter (connection string) |
Ready-made Collector configurations:
observability/otel-collector/config.yaml— Prometheus, Tempo, Loki,observability/otel-collector/config-multi-exporter.yaml— the same plus Elastic, Splunk, Sumo Logic and Azure Monitor at once,observability/otel-collector/config-datadog.yaml— Datadog with APM trace metrics.
All three pass otelcol-contrib validate (0.130.0).
#Notes per vendor
Datadog. Use the datadog/connector in the traces pipeline so APM
service metrics (hits, errors, latency) are computed before sampling.
Histograms are sent as Datadog distributions. The Datadog Agent (7.x) can
also receive OTLP directly (otlp_config in datadog.yaml); point the
application's OTLP endpoint at the Agent.
Elastic. Elastic Cloud provides a managed OTLP endpoint (Authorization:
ApiKey …); self-managed deployments use APM Server, which accepts OTLP.
Logs land in logs-* data streams with trace.id for correlation.
Splunk. Splunk HEC needs a token and an index; Splunk Observability Cloud (APM / Infrastructure) ingests OTLP with an access token header.
Sumo Logic. Create a hosted collector with an OTLP/HTTP source and use
its URL; or use the sumologic exporter.
Azure Monitor. The azuremonitor exporter maps spans to requests and
dependencies, logs to traces; set the Application Insights connection
string.
Prometheus-only shops. Option A needs nothing but a scrape config. Traces and logs can be added later through OTLP without touching metrics.
#Why not vendor SDKs
- Vendor SDKs duplicate the OpenTelemetry pipeline (and its cost) inside the process and tie code to one backend.
- Credentials and endpoints would live in every service instead of one Collector.
- Every vendor above accepts OpenTelemetry, directly or through its Collector exporter.
Kerniq will add a vendor-specific adapter only if a vendor requirement cannot be met through OTLP and the Collector; none has been found.