Databricks — Confusions, Labs, Gotchas & Mock Interview
💡 Interview Tip
Goal: After this page, you should NEVER struggle with Databricks-specific interview questions.
Where to run labs: Databricks Community Edition (free) OR any Databricks workspace.
Memory Map
🧠 DATABRICKS MASTERY → CLUSTER-UC-WORKFLOW
DATABRICKS MASTERYCLUSTER-UC-WORKFLOW
────────────────────────────────────────
CCluster types (All-Purpose / Job / SQL / Serverless)
LLanguages + notebooks (magic commands, widgets)
UUnity Catalog (3-level namespace, governance)
WWorkflows (Jobs vs DLT pipelines)
FFile systems (DBFS vs workspace vs Volumes)
IIntegrations (DLT, Photon, Auto Loader, MLflow)
XCost Xplained (DBU, autoscaling, spot)
SECTION 0: TOP 8 DATABRICKS CONFUSIONS — Cleared Forever
Confusion 1: All-Purpose vs Job vs SQL Warehouse Cluster
| Type | Purpose | Cost | Auto-terminate | When to use |
|---|---|---|---|---|
| All-Purpose | Interactive notebooks | 💰💰💰 high | Yes (configurable) | Development, ad-hoc analysis |
| Job Cluster | Scheduled ETL jobs | 💰 low (up to 50% cheaper) | YES — dies when job ends | Production batch jobs |
| SQL Warehouse | BI / SQL queries | 💰💰 medium | Yes | Power BI, Tableau, SQL analysts |
| Serverless | Instant start, no infra | 💰💰 medium | Auto | Fast startup, no cluster management |
Memory trick:
All-Purpose = "Playground" — exploration, multi-user
Job Cluster = "Assembly Line" — scheduled, dies after use
SQL Warehouse = "BI Gateway" — always-available for dashboards
Serverless = "Rent-a-cluster" — Databricks manages everything
Interview trap: "Why use Job Cluster over All-Purpose for production?" Answer: "Job clusters terminate after the job finishes, so you pay only for compute actually used. They also run a SINGLE job, which isolates resources — no noisy neighbors. Pricing is also lower per DBU for Jobs workloads."