Skip to content
All releases
October 2026

Snowflake Monitoring Kit

Snowflake, monitored in one install

Snowflake's resource monitors tell you when you've hit a credit quota. By then, the money's already spent. The Snowflake Monitoring Kit catches it earlier: 12 ready-made metrics for cost, performance and failures, each checked against its own 30-day baseline, live as soon as you connect.

See it in two minutes

Connect, install, get your first Slack alert

No SQL to write and nothing to tune before it goes live. Alerts land in Slack with the warehouse, the current value and the baseline it broke, so you know where to look without opening a dashboard.

What it watches

12 metrics across your whole account

Cost & credits

  • Warehouse credit spikeA runaway warehouse within the hour, and which one it is.
  • Total credits spikeSeveral services each running a little hot, which no single alert would catch.
  • AI and other services creditsCortex, Copilot, serverless tasks and auto-clustering, split per service.
  • Database storage spikeA pipeline quietly adding terabytes, before it lands on the invoice.
  • Failsafe storage spikeTables recreated instead of updated, which locks in failsafe cost.

Performance

  • Long-running queriesAny query over 30 minutes, with the warehouse and the user.
  • Query queue anomalyMore queries than usual waiting in line on a warehouse.
  • Remote disk spillQueries spilling to remote disk, a sign the warehouse is undersized.

Failures

  • Failed taskSnowflake tasks fail silently by default. You get the task, the time and the error.
  • ETL failure rate spikeFires when a pipeline is actually broken, not on the usual ad-hoc errors.

Wasted spend

  • Warehouse utilizationIdle warehouses you're paying for, and busy ones about to queue.
  • Stale tablesTables nobody has queried in 90 days, sorted by size.
Example alerts and queries for every metric

Setup

Three steps, and only the first is required

  1. 1

    Connect Snowflake. A setup script creates a read-only user, role and warehouse. All 12 metrics are created as soon as the connection test passes.

  2. 2

    Connect Slack. Optional, and you can do it later.

  3. 3

    Pick a channel. Choose where the alerts go.

One read-only grant covers everything. The kit only reads Snowflake's ACCOUNT_USAGE schema. It can't read your table data, write anything or create objects, and new warehouses, users and tables are covered automatically.

GRANT IMPORTED PRIVILEGES ON DATABASE SNOWFLAKE TO ROLE <your_role>;

Find it in the app under System → Library → Snowflake Performance Monitoring. Read the setup docs.

Install the kit and the first alerts run on their next schedule.