compare_sql_executions

MCP tool from MCP Apache Spark History Server by kubeflow

Compare aggregated performance metrics (stages, tasks, shuffle, spill, GC) between two SQL executions; opt-in plan-structure diff

How to use it

compare_sql_executions is exposed by the MCP Apache Spark History Server MCP server. Add the server to your MCP client (Claude Desktop, Cursor, Windsurf and others), and the compare_sql_executions tool becomes available to the model automatically. See the full listing for setup details and every tool this server provides.

Install MCP Apache Spark History Server

$uvx --from mcp-apache-spark-history-server spark-mcp

FULL MCP APACHE SPARK HISTORY SERVER LISTING

Other tools in MCP Apache Spark History Server (18)

aws_analyze_spark_workload

One-shot root cause analysis of failed/slow Spark workloads

aws_spark_code_recommendation

Code fix recommendations for identified Spark issues

compare_job_environments

Diff Spark configs between two applications

compare_job_performance

Diff performance metrics between two applications

compare_stages

Compare two stages (optionally across applications): stage metrics and per-task p25/p50/p75/max quantiles

get_environment

Spark config, JVM info, system properties, classpath; optional section filter to return a single part

get_executor_summary

Aggregate metrics across all executors

get_executor_thread_dump

JVM thread dump for a driver/executor, with state/name/blocked filters (running apps only)

get_job_bottlenecks

Identify bottlenecks across stages, tasks, and executors

get_resource_usage_timeline

Chronological executor add/remove with resource totals

get_sql_execution

SQL execution header by default; opt-in plan, node metrics, job summaries, aggregated stage metrics, and stage list

get_stage

Stage detail with attempt and task metric distributions

list_applications

List applications with optional status, date, and limit filters; pass appid for a single application's detail (status, resources, duration, attempts). Returned applications always include their attempts.

list_executors

List executors with executor-id filtering and sorting (failed-tasks/duration/gc/id)

list_jobs

List jobs with status/job-id filtering and sorting (e.g. slowest by duration)

list_sql_executions

List SQL executions as curated summaries, with status/description filters, sorting, and a default limit

list_stage_task_failures

Failed tasks of a stage with their error messages (exception/stack trace)

list_stages

List stages with status filtering and sorting (e.g. slowest by duration)