Crate datafusion_tracing

DataFusion Tracing is an extension for Apache DataFusion that helps you monitor and debug queries. It uses tracing and OpenTelemetry to gather DataFusion metrics, trace execution steps, and preview partial query results.

Note: This is not an official Apache Software Foundation release.

Overview

When you run queries with DataFusion Tracing enabled, it automatically adds tracing around execution steps, records all native DataFusion metrics such as execution time and output row count, lets you preview partial results for easier debugging, and integrates with OpenTelemetry for distributed tracing. This makes it simpler to understand and improve query performance.

See it in action

Here’s what DataFusion Tracing can look like in practice:

Jaeger UI

DataDog UI

Getting Started

Installation

Include DataFusion Tracing in your project’s Cargo.toml:

[dependencies]
datafusion = "48.0.0"
datafusion-tracing = "48.0.1"

Quick Start Example

use datafusion::{
    arrow::{array::RecordBatch, util::pretty::pretty_format_batches},
    error::Result,
    execution::SessionStateBuilder,
    prelude::*,
};
use datafusion_tracing::{
    instrument_with_info_spans, pretty_format_compact_batch, InstrumentationOptions,
};
use std::sync::Arc;
use tracing::field;

#[tokio::main]
async fn main() -> Result<()> {
    // Initialize tracing subscriber as usual
    // (See examples/otlp.rs for a complete example).

    // Set up tracing options (you can customize these).
    let options = InstrumentationOptions::builder()
        .record_metrics(true)
        .preview_limit(5)
        .preview_fn(Arc::new(|batch: &RecordBatch| {
            pretty_format_compact_batch(batch, 64, 3, 10).map(|fmt| fmt.to_string())
        }))
        .add_custom_field("env", "production")
        .add_custom_field("region", "us-west")
        .build();

    let instrument_rule = instrument_with_info_spans!(
        options: options,
        env = field::Empty,
        region = field::Empty,
    );

    let session_state = SessionStateBuilder::new()
        .with_default_features()
        .with_physical_optimizer_rule(instrument_rule)
        .build();

    let ctx = SessionContext::new_with_state(session_state);

    let results = ctx.sql("SELECT 1").await?.collect().await?;
    println!(
        "Query Results:\n{}",
        pretty_format_batches(results.as_slice())?
    );

    Ok(())
}

A more complete example can be found in the examples directory.

Limitations

Recursive Queries

When using DataFusion Tracing with recursive queries (e.g., those using WITH RECURSIVE), nodes of type WorkTableExec are intentionally not instrumented. This is due to a current limitation in DataFusion: instrumenting these nodes can break recursive query execution and result in errors such as Unexpected empty work table.

As a result, while most of the recursive query plan will be traced and metrics will be collected, some internal operations related to recursion will not be visible in the trace until upstream support is available. See issue #5 for details and tracking progress on this limitation.

Macros

instrument_with_debug_spans

Constructs a new instrumentation PhysicalOptimizerRule for a DataFusion ExecutionPlan at the debug level.

instrument_with_error_spans

Constructs a new instrumentation PhysicalOptimizerRule for a DataFusion ExecutionPlan at the error level.

instrument_with_info_spans

Constructs a new instrumentation PhysicalOptimizerRule for a DataFusion ExecutionPlan at the info level.

instrument_with_spans

Constructs a new instrumentation PhysicalOptimizerRule for a DataFusion ExecutionPlan.

instrument_with_trace_spans

Constructs a new instrumentation PhysicalOptimizerRule for a DataFusion ExecutionPlan at the trace level.

instrument_with_warn_spans

Constructs a new instrumentation PhysicalOptimizerRule for a DataFusion ExecutionPlan at the warn level.

Structs

InstrumentationOptions

Configuration options for instrumented execution plans.

Functions

pretty_format_compact_batch

Formats a RecordBatch as a neatly aligned ASCII table, constraining the total width to max_width. Columns are dynamically resized or truncated, and columns that cannot fit within the given width may be dropped.