SHFL

Shuffle I/O

Moving data between executors takes more time than computing it.

Outcome: Speed
Scope: Judged per stage
Usual fix: A Spark configuration change
CI gate: No budget flag: gate it with --fail-on-introduced against a baseline

An example finding

Where you see it

Dashboard

The SHFL tag marks it in the action region of the dashboard, among the widgets that rank a fix. Every affected stage is flagged, not only the worst.How to read it in the findings guide.

Agent (MCP)

An assistant can read this entry with the list_detectors tool and its background with get_finding_documentation. The call and the response below come from a real run of the MCP server.

Call

{
  "tool": "list_detectors",
  "arguments": {
    "type": "shuffle"
  }
}

Response

{
  "count": 1,
  "detectors": [
    {
      "type": "shuffle",
      "name": "shuffle I/O",
      "tag": "SHFL",
      "tagSharedWith": [],
      "family": "speed",
      "pitch": "Moving data between executors takes more time than computing it.",
      "region": "action",
      "scope": "stage",
      "fixEffort": "config",
      "thresholdSummary": "stage shuffle read above 50 MiB",
      "detectors": [
        {
          "detector": "shuffle",
          "version": 1,
          "thresholds": {
            "minBytes": 52428800,
            "stageFloorPct": 0.005
          }
        }
      ],
      "pages": {
        "catalog": "docs/detectors/shuffle.html",
        "guide": "docs/user-guide/understanding-findings.html#shfl"
      },
      "tuning": {
        "kind": "page",
        "anchor": "#bottleneck-shuffle",
        "path": "docs/tuning-reference/bottleneck-shuffle.html#bottleneck-shuffle"
      },
      "gate": null
    }
  ]
}

get_finding_documentation call

{
  "tool": "get_finding_documentation",
  "arguments": {
    "type": "shuffle"
  }
}

Its response, with each Markdown body replaced by its length

{
  "type": "shuffle",
  "name": "Shuffle I/O",
  "detectionDoc": {
    "tag": "SHFL",
    "title": "Shuffle I/O",
    "content": "<1252 characters of Markdown>"
  },
  "tuningDoc": {
    "anchor": "#bottleneck-shuffle",
    "title": "Shuffle I/O",
    "content": "<11215 characters of Markdown>"
  }
}

CI (CLI)

List this entry from the command line. The output below is a real run of the command. Budgets and exit codes are in the CI guide.

sparkforensics-analyze detectors --type shuffle --format md

# Detectors (1)

## Speed

| Tag | Type | Scope | Fix | Gate | What it finds |
| --- | --- | --- | --- | --- | --- |
| SHFL | `shuffle` | stage | config |  | Moving data between executors takes more time than computing it. |

Report only this finding type for a run with --type:

sparkforensics-analyze <event-log> --type shuffle --impact critical,warning,info

It has no budget flag. Fail a build on a new one by comparing against a baseline with --fail-on-introduced.

Airflow

The operator has no budget parameter for it. Use fail_on_introduced with a baseline.

check_spark_job = SparkForensicsOperator(
    task_id="check_spark_job",
    log_source=...,
    backend=SubprocessAnalyzeHook(),
    baseline_log_source=...,
    fail_on_introduced="critical",
)

The operator works per run, and its summary counts findings per impact band. See the operator for its parameters.

How it is measured

Thresholds

Stage shuffle read above 50 MiB.

shuffle version 1

Threshold	Default
`minBytes`	52428800
`stageFloorPct`	0.005

Each rule's impact bands and fixed fallbacks are in the threshold tables.

Override a default per run with --thresholds; see Tuning detector thresholds.

Docs

This finding has its own tuning page: Spark tuning reference.
Findings guide: SHFL in Understanding findings.