top of page

Snowflake resource monitors: cap BI warehouse credits with auto-suspend

1 minute ago
7 min read
Snowflake resource monitors and auto-suspend for capping BI warehouse credits, Data Studios original cover

A Snowflake resource monitor is the built-in way to put a monthly credit ceiling on a BI warehouse: it counts the credits the assigned warehouses consume, sends notifications at the percentages you choose and suspends the warehouses when a threshold is reached. Auto-suspend is the second control, and for dashboard workloads it often saves more credits than the monitor itself, because it decides how long an idle warehouse keeps billing after the last query finishes. Snowflake documents both features, with their limits, in its user guide and SQL reference, which we checked on November 21, 2026.


The two settings answer different questions. Auto-suspend reduces what a warehouse burns during a normal month, while a resource monitor stops an abnormal month from turning into an expensive one. Snowflake states plainly that resource monitors are not intended for setting precise limits on credit usage and that warehouses may take some time to suspend even with the immediate action, so a monitor needs a buffer below the number you actually want to stay under.


The scope is also narrower than many administrators assume. According to Snowflake's documentation, resource monitors work for warehouses only, and they cannot track spending on serverless features and AI services, which are tracked with budgets instead. A monthly cap on a BI warehouse therefore covers the queries your dashboards send through that warehouse, plus the cloud services credits those warehouses use, and nothing else on the bill.


··········


SNOWFLAKE RESOURCE MONITOR RULES AND LIMITS.


The table below collects the rules from Snowflake's resource monitor page and the CREATE RESOURCE MONITOR reference. They apply to the monitor object itself, whatever the warehouse size.


........


Setting

Documented rule

Who can create

ACCOUNTADMIN only

Delegation

MONITOR or MODIFY privilege on one monitor, via SQL

Account-level monitors

One per account, covers all warehouses

Warehouse-level monitors

Several per account; a warehouse joins only one

Warehouses per monitor

Up to 500

Credit quota

Credits per interval; no quota means it never reaches 100%

Frequency

DAILY, WEEKLY, MONTHLY, YEARLY or NEVER

Reset time

12:00 AM UTC

Actions

One SUSPEND, one SUSPEND_IMMEDIATE, up to five NOTIFY

Thresholds

Percent of quota; values above 100 allowed

Coverage

Warehouses and their cloud services; no serverless or AI services


........


The actions behave differently in a busy BI environment. SUSPEND notifies and suspends the assigned warehouses after running statements finish, so a long dashboard query completes before the warehouse stops. SUSPEND_IMMEDIATE cancels running statements and suspends at once. NOTIFY sends an alert and leaves the warehouse running. A monitor without at least one trigger does nothing when it reaches its quota, which is an easy mistake when a monitor is created through SQL without a TRIGGERS clause.


A suspended warehouse stays suspended until the next interval starts, the quota or the suspend threshold is raised, the warehouse is removed from the monitor, or the monitor is dropped. For a BI team this means that a monitor tripping on the 24th of the month leaves every report on that warehouse without compute until an administrator acts or the month resets at midnight UTC.


The cloud services accounting is a frequent source of confusion. Snowflake bills cloud services only when daily cloud services consumption exceeds 10% of daily warehouse usage, but the documentation says resource monitor limits do not take into account the daily 10% adjustment. A monitor can therefore show more credits used than the invoice later reflects, and it will reach its thresholds slightly earlier than a calculation based on billed credits would suggest. The monitor can track cloud services usage on a warehouse-level monitor but cannot suspend cloud services themselves.


··········


AUTO-SUSPEND SETTINGS FOR A SNOWFLAKE BI WAREHOUSE.


Auto-suspend is a warehouse property set in seconds. The ALTER WAREHOUSE reference lists a default of 600 seconds, accepts any integer of 0 or more, and treats 0 or NULL as never suspending. Snowflake adds that values below 30, or values that are not a multiple of 30, may not behave as expected because the background process that suspends warehouses runs roughly every 30 seconds. Auto-resume defaults to TRUE, so a dashboard query wakes a suspended warehouse without manual action.


The warehouse considerations page recommends a low value, giving 5 or 10 minutes or less as examples, matched to the gaps in the workload. Two side effects limit how low you should go. Each time a warehouse starts it is billed for a minimum of 60 seconds, so a warehouse that suspends and resumes many times an hour pays that minimum on every resume. The warehouse cache is also dropped when the warehouse suspends, which can slow the first dashboard queries after a resume until the cache rebuilds.


Credit consumption depends on size. For Gen1 standard warehouses, Snowflake's warehouse overview lists the following hourly rates, billed per second after the first minute. Gen2 warehouses use the rates in Snowflake's Service Consumption Table, so check the generation of your warehouse before applying these numbers.


........


Warehouse size

Credits per hour (Gen1)

X-Small

1

Small

2

Medium

4

Large

8

X-Large

16

2X-Large

32


........


The price of a credit is set in Snowflake's pricing guide and in your contract, so the worked example below stays in credits. Our Snowflake pricing reference on credits, editions and warehouse sizes covers how credit prices relate to editions.


··········


WORKED EXAMPLE: CAPPING A SMALL SNOWFLAKE WAREHOUSE FOR DASHBOARDS.


The following scenario is illustrative and every figure was computed in Python. A BI team runs its dashboards on a Small Gen1 warehouse at 2 credits per hour. On each of 22 working days, users trigger 40 bursts of activity spread across a 10-hour day, one burst every 15 minutes, and each burst keeps the warehouse busy for 3 minutes. Between bursts the warehouse sits idle for 12 minutes.


With the default auto-suspend of 600 seconds, each burst bills 3 active minutes plus 10 idle minutes before the warehouse suspends, or 13 minutes. With auto-suspend at 60 seconds, each burst bills 3 minutes plus 1 idle minute, or 4 minutes, and each run still clears the 60-second minimum.


........


Setting

Billed minutes per day

Credits per day

Credits per month (22 days)

AUTO_SUSPEND = 600

520

17.333

381.3

AUTO_SUSPEND = 60

160

5.333

117.3

Never suspends, 30 days

1,440

48

1,440


........


The shorter auto-suspend cuts the monthly estimate by a factor of about 3.25 for the same queries, because most of the default setting's cost is idle time. A warehouse left running around the clock for a 30-day month would consume 1,440 credits. With the 60-second setting in place, a monthly quota of 150 credits leaves headroom above the expected 117.3 credits for heavier weeks. The thresholds below follow Snowflake's advice to keep a buffer, with alerts well before the suspend point and an immediate stop only above the quota.


........


Trigger

Percent of quota

Credits

NOTIFY

75

112.5

NOTIFY

90

135

SUSPEND

100

150

SUSPEND_IMMEDIATE

110

165


........


The SQL follows the syntax in Snowflake's reference. The warehouse and monitor names are illustrative, and both the monitor creation and the assignment require the ACCOUNTADMIN role.


USE ROLE ACCOUNTADMIN;

ALTER WAREHOUSE bi_dashboards_wh SET
  AUTO_SUSPEND = 60
  AUTO_RESUME = TRUE;

CREATE OR REPLACE RESOURCE MONITOR bi_dashboards_monthly
  WITH CREDIT_QUOTA = 150
       FREQUENCY = MONTHLY
       START_TIMESTAMP = IMMEDIATELY
  TRIGGERS ON 75 PERCENT DO NOTIFY
           ON 90 PERCENT DO NOTIFY
           ON 100 PERCENT DO SUSPEND
           ON 110 PERCENT DO SUSPEND_IMMEDIATE;

ALTER WAREHOUSE bi_dashboards_wh SET RESOURCE_MONITOR = bi_dashboards_monthly;

Because the monitor counts cloud services credits without the 10% adjustment, the 150-credit quota acts as a slightly conservative cap on billed warehouse credits. A dedicated warehouse per monitor gives the tightest control, which Snowflake also recommends when strict limits matter.


··········


SNOWFLAKE RESOURCE MONITOR NOTIFICATIONS AND PERMISSIONS.


Notifications are disabled by default. Each user who should receive alerts must verify an email address and enable resource monitor notifications in the Snowsight profile preferences. Non-admin users can be added through NOTIFY_USERS, with a maximum of five, only on warehouse-level monitors and only by email. If any listed user lacks a verified email address, the CREATE statement fails, while a changed but unverified address fails silently, which makes a test alert worth sending after setup.


Delegation is limited. An account administrator can grant MONITOR or MODIFY on a specific monitor, and MODIFY lets another role change the quota, schedule and actions through SQL. Changing the monitor type or the list of warehouses on a warehouse-level monitor still requires ACCOUNTADMIN, and so does assigning a warehouse to a monitor.


··········


COMMON MISTAKES WITH SNOWFLAKE CREDIT CAPS.


The most common error is expecting an account-level monitor to cap the whole bill. Snowflake's documentation says an account-level monitor does not control serverless compute such as Snowpipe, automatic reclustering or materialized views, and resource monitors cannot track AI services at all. Spending on Cortex features belongs to budgets, as described in our Snowflake Cortex AI Functions pricing reference.


A second error is assuming an account-level monitor overrides warehouse-level ones. Both apply: the account monitor counts all warehouses, while a warehouse-level monitor can suspend its own warehouse earlier. A third is setting the suspend threshold at exactly the budget figure, which ignores the documented delay before warehouses stop. Teams that also run BigQuery or Databricks will find comparable controls in our pieces on BigQuery project caps for slots and concurrency and Azure Databricks Serverless SQL pricing, each with its own scope and limits.


··········


SNOWFLAKE BI WAREHOUSE CAPS: WHAT YOUR ACCOUNT SETTINGS DECIDE.


Snowflake's documentation confirms the mechanics: ACCOUNTADMIN creates and assigns monitors, a monitor holds one SUSPEND, one SUSPEND_IMMEDIATE and up to five NOTIFY triggers, quotas reset at midnight UTC, and auto-suspend defaults to 600 seconds with a 60-second minimum charge on every start. In the illustrative dashboard workload, moving auto-suspend from 600 to 60 seconds lowers the estimate from 381.3 to 117.3 credits a month, and a 150-credit monitor with 75% and 90% alerts caps the remaining risk.


What the documentation leaves to your account is the money and the edge cases. The price of each credit depends on your contract and Snowflake's pricing guide, Gen2 warehouses use different consumption rates, and the delay before a suspended warehouse stops is not quantified. Before relying on a cap, confirm the warehouse generation, check that notification emails are verified, and decide which reports may go dark if the monitor trips late in the month.


··········


FOLLOW US FOR MORE.


·····


DATA STUDIOS


·····


datastudios.org


bottom of page