feat(tracemetrics): Add sample_rate parameter to metric API calls #18103
Conversation
This allows for a sample_rate (0, 1.0] to be sent on a per metric basis. Refs LOGS-497
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
![sentry[bot]](https://avatars.githubusercontent.com/in/12637?s=60&v=4){kind=small}
k-fish
force-pushed
the
feat/tracemetrics/add-sample-rate-function
branch
from
393fb60 to
1d11dca
Compare
size-limit report 📦
|
node-overhead report 🧳Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.
|
| if (metric.sample_rate <= 0 || metric.sample_rate > 1.0) { | ||
| client.recordDroppedEvent('invalid_sample_rate', 'metric'); | ||
| return false; | ||
| } |
There was a problem hiding this comment.
Bug: Invalid NaN Rejection in Sample Rate Validation
The sample_rate validation doesn't properly reject NaN values. When sample_rate is NaN, the condition metric.sample_rate <= 0 || metric.sample_rate > 1.0 evaluates to false (since all comparisons with NaN return false), allowing NaN to pass validation. This causes the metric to be silently dropped during sampling (since sampleRand < NaN is always false) without recording a dropped event. The validation should explicitly check for NaN and reject it as invalid.
| setMetricAttribute(processedMetricAttributes, 'sentry._internal.replay_is_buffering', true); | ||
| const { sampled, samplingPerformed } = shouldSampleMetric(beforeMetric, currentScope); | ||
| if (!sampled) { | ||
| return; |
There was a problem hiding this comment.
Bug: Missed Tracking of Sampled-Out Metrics Drops
When a metric is dropped due to sampling (when sampled is false), the code does not record this as a dropped event using client.recordDroppedEvent('sample_rate', 'metric'). The EventDropReason type already includes 'sample_rate' as a valid drop reason, and the code properly records drops for invalid sample rates at line 90, but fails to record drops for valid sampling decisions. This means sampled-out metrics won't be tracked in client reports, making it impossible to monitor how many metrics are being dropped due to sampling.
|
Strange how this has impacted the CDN Bundles but nothing else 🤔 |
This allows for a sample_rate (0, 1.0] to be sent on a per metric basis.
Refs LOGS-497