Analytics
Link any Node into it. It samples that traffic and turns it into Insight.
What it does#
Analytics stands for an OpenTelemetry collector with a tracing backend: the system that records what your requests did.
Analytics is a tap. Link any Node into it, and it receives a copy of that Node's traffic, not a share of it. It samples that copy at the sample rate you set. Each sample earns Insight in the pool of its request kind. A request that fails or frustrates its user also earns Failure Insight, and it teaches more than a healthy one.
Insight grows with the square root of samples, counted per Component type and kind. So tapping many different Components teaches more than tapping one large one. Insight and observability covers the full model.
Analytics keeps a limited number of samples per second. Each size step doubles it. Beyond that, it drops samples evenly. Retention sets how long it keeps traces, and a longer window costs more Burn.
When to place it#
The Garage board starts with one Analytics Node linked to the App Server. Add links when Insight for a kind grows slowly, or when a new Component carries traffic you have not sampled yet. Resize it when its card shows it at full throughput.
How it fails#
Analytics never breaks a request, but it has costs.
Overloaded, it drops samples, and Insight slows.
A high sample rate adds CPU and latency to every tapped Node, and it multiplies storage Burn.
Without Analytics, you earn no Insight from traffic, and Research stalls.
No incident targets Analytics.
Tips#
Tap one Node of each Component type before you raise the sample rate.
Lower the sample rate on hot App Servers if p99 matters more than Insight.
Rules#
- Each further copy costs 25% more than the one before.
- Selling it refunds 50% of what its last copy cost.
- It has 8 sizes. Each step doubles its capacity and its Burn.
- It receives a copy of its parent's traffic, not a share of it.