Skip to content
Agentmart
Submit a tool

dbx-perf

Use when a Databricks or Spark job is slow, fails with out-of-memory or executor loss, costs too much, or when tuning Delta tables.

Skill by Srimukh99

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/srimukh99-dbx-perf

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Use when a Databricks or Spark job is slow, fails with out-of-memory or executor loss, costs too much, or when tuning Delta tables.

Use cases: data

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/dbx-perf

What's inside

Skill name
dbx-perf
Description
Use when a Databricks or Spark job is slow, fails with out-of-memory or executor loss, costs too much, or when tuning Delta tables.
License
MIT
Instructions
33 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# dbx-perf

## 1. Find the slow part

Open the Spark UI (or the query profile for SQL warehouses). Sort stages by duration. In the slowest stage compare max task time with the median.

| You see | Meaning | Try |
| --- | --- | --- |
| Max task far above median | Data skew | Adaptive query execution skew handling; salt the hot key; filter nulls before joining |
| Spill to disk | Partitions too big for memory | More shuffle partitions or let AQE coalesce; bigger nodes for this job |
| Huge shuffle read/write | Wide joins or aggregations | Broadcast the small side; pre-aggregate; filter earlier |
| Many tiny tasks or files | Small-file problem | `OPTIMIZE`; predictive optimization; tune write sizes |
| Python UDF stages slow | Row-by-row Python | Use built-in functions, or pandas (vectorized) UDFs |

## 2. Delta tables

- Liquid clustering on the columns most queries filter by, instead of hand-picked partitions.
- If you do partition, keep partitions large (around 1 GB or more); never partition by high-cardinality columns.
- Enable predictive optimization or schedule `OPTIMIZE` and `VACUUM`.

## 3. Code smells

`collect()` or `toPandas()` on large data, loops issuing one Spark action per row, `count()` used only for logging, caching data that is read once.

## 4. Cost

Check `system.billing.usage` for DBUs by job. Right-size clusters, use autoscaling and job clusters (or serverless) instead of always-on all-purpose clusters.

## Prove it

Report runtime and DBUs before and after (`ship`).

From Srimukh99/rubric, MIT.

More from Srimukh99/rubric

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 2, 2026 (27 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
Repository
github.com/Srimukh99/rubric
First seen
Oct 24, 2026
Last checked
Oct 24, 2026
Source
GitHubgithub-graphql

More from Srimukh99

Cite this page

dbx-perf skill. Agent Mart. https://agentmart.dev/skills/srimukh99-dbx-perf (last updated 2026-10-24)