Move from Datadog to ClickStack without a monitoring gap

The datadog receiver is an alpha component with no real /api/v1/validate route: the catch-all handler answers validation calls with an empty 200 while telemetry flows normally. Replay recorded agent traffic through the bridge before rolling it out to a fleet.

Only traces, metrics, and logs have an intake here. Live Process, Network Performance Monitoring, and orchestrator traffic have nowhere to land, so leave those agents pointed at Datadog until you stop paying for them.

Datadog Agent logs only reach this receiver over HTTP; set logs_config.force_use_http: true on the agent or the log stream silently stays on the old path. Agents must also disable the v3 metrics API (use_v3_api.series.enabled: false): the receiver has no v3 series intake and answers those payloads with a 200 while dropping them, so agents on 7.81+ defaults lose metrics silently. ClickHouse's migration guide pins Agent 7.59 or newer.

At this collector version the datadog exporter's retry_on_failure applies to logs only. Traces are attempted once and metrics rely on the Datadog client's internal retry, so a long Datadog outage can drop queued traces even though the queue is disk-backed. ClickStack delivery is unaffected because each exporter owns its own queue.

Agent-computed APM trace stats pass through the metrics pipeline to Datadog (the receiver hands them to the exporter as a dd.internal.stats.payload metric), so keep the datadog receiver and exporter in the same metrics pipeline. Services already on OTel SDKs produce no Datadog APM stats through the bridge, because the exporter's own stats computation is off by default at this version; add the datadog connector if you still need their APM numbers during the migration.

Datadog Agents send delta metrics and OTel SDKs send cumulative; ClickStack stores each point's temporality as data, and the datadog exporter converts cumulative sums itself, which is why no pipeline runs a temporality processor. Datadog's metrics intake is the least-tested signal on the receiver, and ClickHouse's migration guide says to migrate metrics last.

The clickhouse exporter's metrics support is alpha at this version (traces and logs are beta). create_schema: true makes the first run create the otel_* tables ClickStack reads; in production set it to false and manage the schema yourself so TTL changes are deliberate rather than applied by whichever collector starts first.

HyperDX keys everything off service.name: records that arrive without one land under a blank service and cannot be filtered by service. The receiver promotes the agent's service tag automatically, but check the OTLP side before trusting the comparison.

Both backends receive a full copy of everything, so egress and Datadog ingest cost roughly double for as long as the bridge runs.