Managing logging costs

Similarly to tracing and metrics, logging increases the compute resources needed to run an application, the cost of running a logging pipeline (if any), and the costs associated with using (or running) an observability backend. Vendor pricing is frequently based on a combination of telemetry volume, retention time, and API calls, including queries.

We already know how to write logs efficiently, so let’s talk about pipelines and backends.

Pipelines

A logging pipeline consists of the infrastructure needed to send logs to the backend of your choice. It’s typical to do some grokking, parsing, transformations, buffering, throttling, and hardening on the way to the backend.

In a simple case, it’s all done by your vendor’s logging provider or the OpenTelemetry processors and exporter inside the process.

In many cases, we need logging pipelines to capture logs and events coming from outside – the OS, self-hosted third...