Calculate Lambda invocation duration from logs

Problem / use case

You want to measure how long each AWS Lambda invocation took by calculating the time difference between the first and last log entry per invocation. This is useful for debugging, performance tuning, or anomaly detection.

Query

source logs
| filter cx_metadata.faas_name != null
| groupby cx_metadata.faas_name as lambda, cx_metadata.faas_execution as invocation_id agg
   min($m.timestamp) as start,
   max($m.timestamp) as end
| create duration from end.diffTime(start)

Expected output

Variations

Aggregate durations per Lambda function

Once you’ve calculated the duration of each Lambda invocation, you might want to summarize performance across all invocations of a given function. This helps answer questions like:

• Which Lambda runs the longest on average?
• What’s the total compute time per function?
• Are there outliers or performance spikes?

Add this to the end of the query:

| groupby lambda aggregate
   sum(duration) as total_duration,
   avg(duration) as avg_duration,
   percentile(0.95, duration) as p95_duration

TL;DR

Use diffTime on grouped timestamps to calculate Lambda execution duration. Optionally roll up by function name for summaries.

Previous Bucket longtask durations into performance ranges

Next API queries

Theme

Light

light

dark