JobViz - Databricks Job Visualizer
Visual development and debugging for Databricks Jobs. See your task DAGs as interactive graphs, edit and validate them, and investigate failed runs: what failed, what it affected, and what changed. All inside VS Code.

Features
🚨 Investigate Failed Runs
Connect to Databricks to see each run on the DAG, then work through a failure without leaving the editor:
- What failed: status, timing, run IDs, error, stack trace and task output for each task, including what a notebook printed, with earlier attempts from retries and repairs. Jump from a stack frame to the local file, with a warning if the file changed after the run started
- Which code ran: for bundle jobs, the commit the run was deployed from, and a warning if your checkout is on a different one
- What it affected: the failure count in the status bar lists failed tasks, tasks that didn't run because of them, and downstream tasks that ran anyway, using the states Databricks reported. It marks each Likely origin, a failed task with no failed task upstream; this is inferred from the graph, not reported by Databricks
- Is it new: a failed task shows how it did in up to the last 10 finished runs, such as
✓✓✗✗ failing for 2 runs, since #806. Click a mark to open that run
- Repeated errors: when several tasks or iterations fail with exactly the same error, it's called out once
- When no task failed: a run stopped at job level, such as by the concurrency limit, shows the run's own code and message
- For-each tasks: iteration counts, and the error and stack trace of each failed iteration
- Run-job tasks: the exact child run, and the real error from its failed tasks
- Fix it: run, repair or cancel from the DAG. Each action asks for confirmation first, and
jobviz.databricks.allowRunActions turns them off
⇄ Compare Runs
- Pick two runs from the history, or use Compare with last success on a failed run
- See status changes such as New failure and Recovered, and new or removed tasks
- Compare durations, with queue, setup and execution time shown separately, because a slow start can look like slow code
- See which job parameters and task settings changed, what each run ran on, its deployed commit when known, and whether it was repaired
- Mark the tasks that changed on the DAG with Show on DAG
- Click a task to see its details in that run, beside the comparison
≠ Configuration Drift
- Drift compares your local tasks and dependencies with the deployed job, for the active target
- See tasks and dependencies not deployed yet or removed locally, and changes made in the workspace
- Defaults that Databricks adds aren't reported, so you only see real differences
- Other settings in your YAML aren't compared, and Drift says so. It only reads from your workspace
📊 Runs Dashboard
- JobViz: Open Runs Dashboard shows the jobs from every bundle in your workspace, one section per bundle
- See the latest and running runs, progress and who started each one, then open a run or failed task in the DAG
- Get a notification when a watched job fails, and running and failed counts in the status bar
⏱️ Timeline
- Switch from DAG to Timeline to see when each task ran, what overlapped, and which chain of tasks took longest
- Sort by start time, duration, status or name, zoom in, and see earlier attempts as dashed bars
🔀 Interactive DAG Visualization
- Renders Databricks job task dependencies as a directed acyclic graph
- Pan, zoom, and navigate through large pipelines
- Top-to-bottom or left-to-right layout
- Click any node to see task details, double-click to jump to YAML source
✏️ Visual Editing
- Add, rename, duplicate and delete tasks, draw dependencies, and edit task and job settings from the inspector
- Changes are written back to your YAML
📦 Databricks Asset Bundle Support
- Parses multi-file bundle structures (
databricks.yml + resources)
- Resolves
include: directives and variable substitutions (${var.name})
- Supports target-based overrides (dev/staging/prod)
- Incremental re-parsing on file save for fast feedback
- Open a parent folder holding several bundles, and switch between them
🔍 Search & Filter
- Text search across task IDs, labels, and types
- Filter by task type (Notebook, Spark Python, SQL, etc.)
- Filter by validation status (errors, warnings, valid)
- Show only the tasks that failed in the run on screen
- Non-matching nodes dim to maintain context
⚠️ Validation
- Detects missing dependencies, circular references, and disconnected tasks
- Highlights errors and warnings directly on graph nodes
- Inline validation badges with error details in the details panel
- Validate on save or on demand via CodeLens
💥 What-If Failure Impact
- Select any node and see all downstream tasks that would be affected if it failed, with no run needed
- Impact path highlighted in red across the graph
- Understand blast radius before making changes
- Browse all jobs in the workspace from the activity bar
- See task counts, job structure at a glance
- Click a job to open its DAG instantly
🔗 Cross-Job References
- Detects
run_job_task references to other jobs in the bundle
- Visual indicator on nodes that reference external jobs
- One-click navigation to the referenced job's DAG, on the exact child run when it's known
- Back to <parent job> in the toolbar returns to the parent, with the task selected and a pinned run pinned again
- Going back to a DAG shows its last known status straight away, marked Updating… until the fresh copy arrives
Getting Started
Prerequisites
- VS Code 1.85.0+
- A Databricks Asset Bundle or standalone job YAML file
Installation
Search for "JobViz" in the VS Code Marketplace, or install manually:
code --install-extension jobviz-<version>.vsix
Usage
- Open a workspace containing a
databricks.yml bundle or a Databricks job YAML file
- Use the Validate Job or Open in DAG CodeLens that appears at the top of your job file
- Or run JobViz: Open DAG from the command palette (
Cmd+Shift+J / Ctrl+Shift+J)
Connecting to Databricks
Live run status uses a profile from your Databricks config file (~/.databrickscfg, or DATABRICKS_CONFIG_FILE, or the jobviz.databricks.configPath setting). If you don't have one, run JobViz: Sign in to Databricks (or click Databricks in the status bar):
- Browser login: signs in through the Databricks CLI (
databricks auth login). Needs CLI v0.200 or later.
- Personal access token: no CLI needed. Paste a token from your workspace (Settings → Developer → Access tokens); JobViz checks it and adds the profile to the config file.
When you open the repo again, JobViz reconnects with the same profile automatically (turn off with jobviz.databricks.autoConnect). Either way the profile is a normal config profile, so the CLI and databricks bundle can use it too. If a token expires or is revoked, the status bar shows Sign in required; click it to sign in again and JobViz updates that profile in place.
Good to Know
- Notebook output: JobViz reads what a notebook printed from the run's export. A cell that calls
dbutils.notebook.exit() doesn't keep what it prints, so that output won't appear
- Output of a running task: it refreshes with each status update, every 60 seconds by default, if Databricks offers it. Use Open task run to watch it live in Databricks
- Deployed commit: a bundle only records its latest deployment, so a run from before it shows no commit. Uncommitted changes that were deployed can't be seen
- For-each tasks: a failed iteration is shown by its task run ID, because Databricks doesn't say which input it had
- Parameter values are shown as they are: JobViz doesn't hide job parameter values, so take care when sharing your screen
- Repaired runs: Databricks keeps a run's original parameter values after a repair, so a repair's own values aren't shown
- Compute: run features have been tested on serverless compute. Classic cluster details have only been tested with simulated responses
Commands
| Command |
Keybinding |
Description |
JobViz: Open DAG |
Cmd+Shift+J |
Open the DAG view for the active YAML file |
JobViz: Validate Job |
— |
Run validation rules on the current job |
JobViz: Refresh |
— |
Re-parse and update the DAG |
JobViz: Show Failure Impact |
— |
Highlight downstream tasks affected by failure |
JobViz: Focus Task in DAG |
— |
Focus the DAG on the task under your cursor |
JobViz: Fit DAG to Screen |
— |
Reset zoom to fit the entire graph |
JobViz: Open Runs Dashboard |
— |
Open the dashboard of jobs across your bundles |
JobViz: Sign in to Databricks |
— |
Sign in or add a profile |
JobViz: Select Run |
— |
Choose which run the DAG shows |
JobViz: Repair Run |
— |
Rerun chosen tasks of the run on screen, after confirming |
JobViz: Cancel Run |
— |
Stop the run on screen, after confirming |
JobViz: Open Run in Databricks |
— |
Open the run on screen in your workspace |
JobViz: Switch Bundle |
— |
Switch to another bundle in the workspace |
JobViz: Toggle Edit Mode |
Cmd+E |
Turn visual editing on or off |
Configuration
| Setting |
Default |
Description |
jobviz.defaultTarget |
"" |
Default Databricks target for variable resolution |
jobviz.autoValidate |
true |
Automatically validate on file save |
jobviz.dagLayout |
"TB" |
DAG layout direction: "TB" (top-bottom) or "LR" (left-right) |
jobviz.showMinimap |
true |
Show minimap in DAG view |
jobviz.editing.enabled |
true |
Enable visual editing |
jobviz.databricks.liveStatus |
true |
Show live run status on the DAG |
jobviz.databricks.allowRunActions |
true |
Offer run now, repair and cancel in the DAG |
jobviz.databricks.pollInterval |
60 |
Seconds between status checks of a running job (10–300) |
jobviz.dashboard.backgroundPoll |
true |
Check bundle jobs every 5 minutes while the dashboard is closed |
jobviz.dashboard.failureNotifications |
true |
Notify when a watched job fails |
Supported Task Types
| Type |
Icon |
notebook_task |
📓 |
spark_python_task |
🐍 |
python_wheel_task |
📦 |
spark_jar_task |
☕ |
spark_submit_task |
⚡ |
pipeline_task |
🔀 |
run_job_task |
🔗 |
sql_task |
🗃️ |
dbt_task |
🏗️ |
condition_task |
❓ |
for_each_task |
🔄 |
License
MIT
Icons are from VS Code Codicons, licensed under CC BY 4.0.
| |