← ALL PROBLEMS LATENCY · WAREHOUSE STARTUP

The SQL warehouse is stuck in 'Starting', and the first query just waits

Cluster start-up delays push the first query behind a spin-up window. An engine with no warehouse to start answers right away.

STACK · SQL WAREHOUSE · SERVERLESS OR CLASSIC
FIRST QUERY OF THE DAYSTUCK: STARTING
COLD STARTQUERY

THE PROBLEM

A SQL warehouse that has scaled down sits in a Starting state when the next query arrives, so the first request waits for the cluster to resume instead of running.

THE PROBLEM ANATOMY

Who it hits

Data platform lead

Owns warehouse sizing and auto-stop, and fields complaints about the opening wait.

On-call engineer

Owns the pre-warm script that keeps the warehouse awake before work starts.

BI developer

Owns the dashboard that stalls on its first load each morning.

The situation

A warehouse auto-stops to save money, then has to resume when the next query comes in. During that resume it reports Starting, and the query that triggered it waits for the cluster to come up. The documented workaround is to fire a query on a schedule before anyone logs in, so the warehouse is already awake.

That keeps the warehouse warm by paying for idle time, which trades one cost for another. The startup itself never goes away. It is only hidden behind a timer.

The specific challenges

  • Startup blocks work: the first query waits for the cluster to resume
  • Auto-stop reopens it: every idle period restarts the same delay
  • Pre-warm burns idle: keeping the warehouse awake pays for time no one queries
  • Mornings are worst: the first load of the day always pays the full start-up

WHAT IT COSTS

PRACTICAL

  • Morning stalls: the first dashboard load waits on a resume
  • Timer scripts: engineers maintain pre-warm jobs to mask the delay

BUSINESS

  • Idle spend: warm warehouses bill for hours with no queries
  • Weak demos: a stalled first load undercuts the tool in front of stakeholders

EMOTIONAL

  • Recurring blame: the platform team owns a delay it cannot configure away
  • Low trust: users assume the system is down when it is only starting

THE WAY OUT

A lakehouse compute engine with atomic vCPU scaling

Swap in a pilot of e6data's query engine alongside the existing warehouse to:

Drop the start-up: queries run without a warehouse to resume, so the first request executes immediately. The tables and catalog stay untouched. What changes is that Starting stops being a state a query can wait in.

Stop paying for warm idle: compute scales to the query and back down, so there is no reason to keep a warehouse awake on a timer. The existing auto-stop habits stay or go. What is removed is the idle bill that pre-warming created.

Keep the stack: open tables stay where they are, access controls carry over, and adoption is an endpoint change rather than a migration, needing only fast validation of the required integration mapping. What is added is an immediate first query.

0
warehouse resume before a query
< 2s
p95 on interactive queries
1-vCPU
scaling, no start-up window

The result is a first query that runs the moment it arrives, on the tables already in place. Setup and connection details are in the query engine documentation.

See how the Query engine removes the wait Share your problem, talk with a founding team member →

Book a demo on your own workloads

Reach out to book a demo, share challenges you're facing, and tell us how this fits into what you're currently working on or thinking about.

Prefer to self-serve? Problems we're solving

An actual person replies. By submitting, you acknowledge your personal information will be processed in accordance with our Privacy Policy.