Set a default SQL warehouse per user

04Configure via the Databricks CLI

The Databricks CLI v0.250+ ships five subcommands under databricks warehouses for managing per-user overrides. The CLI is the fastest path for one-off configuration and scriptable rollouts.

Prerequisites

# Install
brew install databricks/tap/databricks

# Authenticate (creates profile in ~/.databrickscfg)
databricks auth login --host https://<your-workspace>.cloud.databricks.com

# Find a warehouse ID
databricks warehouses list --profile <your-profile>

Set your own override

# Create a new CUSTOM override pointing at a specific warehouse
databricks warehouses create-default-warehouse-override \
  me CUSTOM \
  --warehouse-id <warehouse-id> \
  --profile <your-profile>

The me shorthand resolves to your authenticated user. Type CUSTOM means "pin to this specific warehouse"; the alternative LAST_SELECTED uses the most recently selected warehouse instead (sticky behavior, no explicit pin).

Read your current override

databricks warehouses get-default-warehouse-override \
  default-warehouse-overrides/me \
  --profile <your-profile>

# Output:
# {
#   "default_warehouse_override_id": "<numeric-user-id>",
#   "name": "default-warehouse-overrides/<numeric-user-id>",
#   "type": "CUSTOM",
#   "warehouse_id": "<your-warehouse-id>"
# }

Update an existing override

# Change the warehouse target (uses field mask: "*" = update all fields)
databricks warehouses update-default-warehouse-override \
  default-warehouse-overrides/me "*" CUSTOM \
  --warehouse-id <new-warehouse-id> \
  --profile <your-profile>

# Or switch type from CUSTOM to LAST_SELECTED (no warehouse-id needed)
databricks warehouses update-default-warehouse-override \
  default-warehouse-overrides/me "*" LAST_SELECTED \
  --profile <your-profile>

List all overrides (admin only)

databricks warehouses list-default-warehouse-overrides \
  --profile <your-profile>

Set an override for another user (admin only)

# Replace 'me' with the target user's numeric ID
# Get user IDs from: databricks workspace-users list (or SCIM API)
databricks warehouses create-default-warehouse-override \
  <user-id> CUSTOM \
  --warehouse-id <warehouse-id> \
  --profile <admin-profile>

Delete an override

databricks warehouses delete-default-warehouse-override \
  default-warehouse-overrides/me \
  --profile <your-profile>

# After deletion, the user falls back to the workspace-level default warehouse

07Verify it works

Three layers of verification — each tighter than the last. Use whichever fits your confidence requirement.

Layer 1 — confirm the override is recorded

databricks warehouses get-default-warehouse-override \
  default-warehouse-overrides/me \
  --profile <your-profile>

Returns the override's type and warehouse_id. If warehouse_id matches your intended target, the override is stored correctly.

Layer 2 — confirm the algorithm picks it

databricks experimental aitools tools get-default-warehouse --profile <your-profile>

This CLI command resolves the same default-warehouse lookup that the managed MCP uses. The returned warehouse ID should match your override target. If it doesn't, the override is recorded but not winning the priority lookup — file a support ticket with both values.

Layer 3 — observe the routing end-to-end

Stop your target warehouse and every other warehouse the lookup might reach. Send a query through the managed MCP (or any consuming surface). The target warehouse should transition STOPPED → RUNNING, and no other warehouse should be touched.

# Stop all warehouses to baseline
for WID in $(databricks warehouses list --profile <p> -o json | jq -r '.[].id'); do
  databricks warehouses stop "$WID" --profile <p> &
done; wait

# Send a query through any default-warehouse-using surface (MCP, SQL editor, etc.)
# Then re-list warehouses and check which one woke up:
databricks warehouses list --profile <p> -o json | jq '.[] | {id, state, name}'