Log pipelines
Create a log connection in the console first (Logs). It gives you a pipeline ID and a pipeline key. Then change one output in your collector. Your applications don’t change.
OpenTelemetry Collector
# Swap one exporter endpoint.
exporters:
otlphttp/onshore:
endpoint: https://vault.yourbank.example/logs/otlp/<pipeline-id>
headers:
X-RV-Pipeline-Key: <pipeline-key>
# When the vault is unreachable the processor answers 503. These settings keep
# the batch here and retry; raw logs are never forwarded around it.
retry_on_failure: { enabled: true, max_elapsed_time: 0 }
sending_queue: { enabled: true, storage: file_storage }
service:
pipelines:
logs:
exporters: [otlphttp/onshore]OTLP over HTTP and gRPC are both accepted. Traces and metrics can go through the same pipeline.
The processor forwards cleaned logs to the destination you gave in the console.
Behaviour
| Situation | What happens |
|---|---|
| Watch only | Logs reach your vendor unchanged; identifiers are counted |
| Protecting | Identifiers become tokens, masked or dropped, per type |
| Vault unreachable, “buffer and retry” | The processor answers 503, and your collector keeps the batch and retries |
| Vault unreachable, “drop” | Lines are discarded |
| Wrong or missing key | Refused |
When protecting, unredacted logs are never forwarded, whatever the failure setting. Configure your collector to buffer to disk and retry, as above.
Keeping debugging possible
Use “same value, same token” fields: the same customer always gets the same token, so you can still search your logs for one customer’s activity.
Try a file first
rv scan <file> reads a log file offline and reports what would leave the country, without an install.
Last updated on