How it works
Step 1
Spanrow ships as a small package for each language. It wraps your web framework, database drivers and HTTP client, so you don’t write tracing code by hand.
npm install @spanrow/node
export SPANROW_KEY=sk_live_7f3a91c2
node -r @spanrow/node/auto server.js
Already on OpenTelemetry? Point your exporter at collector.spanrow.io and skip this step.
Step 2
Each service that handles part of a request adds its own spans. Spanrow joins them into one trace using the header your services already pass along.
Queues and background jobs work the same way: the job’s spans hang under the request that enqueued it.
traceparent: 00-7f3a91c2e04b9d10-5c1e02a4-01
Step 3
A trace is a waterfall: one row per span, on a shared clock. Long bars are slow. Bars that sit end to end are waiting on each other.
Spanrow marks the span that cost the most time and explains why, with the file and line that did the work.
Step 4
Pick a route and the percentile you care about. Spanrow draws the target on the chart and tracks how much of this month’s budget is left.
route: POST /v2/checkout
p95: 500ms
window: 30d
Step 5
When a route misses its target for an hour, Spanrow sends one message with the slowest trace from that hour attached. No alert storms and no dashboards to watch.
Alerts go to Slack, PagerDuty, email or any webhook.