Snowflake Resource Monitors and Cost Dashboards

1.What a resource monitor does and does not watch

A virtual warehouse consumes credits while it runs, and the number of credits depends on its size and how long it runs. A resource monitor sets a ceiling on that consumption. It tracks the credits used by every warehouse assigned to it within an interval. When usage reaches thresholds you define, it can send alert notifications, suspend standard warehouses, or disable Adaptive Warehouses.

The scope is narrow, and the exam tests it. Resource monitors work for warehouses only. They cannot track spending on serverless features or AI services. Snowflake's guidance for those features is to use a budget instead. Only the ACCOUNTADMIN role can create a resource monitor, but an account administrator can grant other roles privileges to view and modify one.

Within that scope, the quota counts more than warehouse compute alone. The credit quota covers credits consumed by user-managed warehouses *and* by the cloud services that support them. With a 1000-credit limit, 700 warehouse credits plus 300 cloud services credits in the same interval triggers the alert. The quota also ignores the daily 10% cloud services adjustment on your bill. Snowflake uses all cloud services consumption to decide whether a limit has been reached, even consumption that is never billed. As a result, a monitor can hit its limit before your bill would suggest it should.

Sources1

2.Credit quota and schedule

Two properties decide *how much* and *over what period*. CREDIT_QUOTA is the number of credits allocated per interval, and any number can be specified. If you leave it unset, the monitor has no quota and will never reach 100% usage. Snowflake tracks used credits across all assigned warehouses and resets that number to 0 at the end of each interval.

The default schedule starts monitoring immediately and resets used credits at the beginning of each calendar month, which is the start of the standard Snowflake billing cycle. You can customize the schedule with three properties:

- FREQUENCY: MONTHLY, DAILY, WEEKLY, YEARLY or NEVER. With NEVER, used credits never reset, so warehouses keep using credits until the quota is reached. - START_TIMESTAMP: a future timestamp, or IMMEDIATELY. If you give a date without a time, Snowflake uses the current time. If you give a time without a time zone, it defaults to UTC. - END_TIMESTAMP: the date and time when Snowflake suspends the assigned warehouses.

FREQUENCY and START_TIMESTAMP must be set together. You cannot set one without the other.

With a custom schedule, the frequency is relative to the start date rather than the calendar month. The reset still happens at 12:00 AM UTC, whatever time the start timestamp names. A WEEKLY monitor that starts on a Monday resets every Monday. If the start date is the last day of a month, Snowflake resets on the last day of every following month. A start of January 31 gives resets on February 28 (or 29), March 31, April 30 and so on.

END_TIMESTAMP is rarely used, and its behaviour can be a surprise. At the end date, all assigned warehouses are suspended even if the quota hasn't been reached. The notification that is sent reports a suspend immediate action at whatever percentage had been used, which may not be one of your thresholds.

USE ROLE ACCOUNTADMIN; CREATE OR REPLACE RESOURCE MONITOR limit1 WITH CREDIT_QUOTA=1000 FREQUENCY = MONTHLY START_TIMESTAMP = IMMEDIATELY TRIGGERS ON 100 PERCENT DO SUSPEND; ALTER WAREHOUSE wh1 SET RESOURCE_MONITOR = limit1;

USE ROLE ACCOUNTADMIN; CREATE OR REPLACE RESOURCE MONITOR limit1 WITH CREDIT_QUOTA=1000 FREQUENCY = MONTHLY START_TIMESTAMP =  ?  TRIGGERS ON 100 PERCENT DO SUSPEND;

Sources12

3.Triggers: notify, suspend, or suspend immediately

The quota on its own does nothing. Actions, also called triggers, say what happens at each threshold. Each trigger is a percentage of the quota paired with an action, and thresholds above 100 are allowed. A monitor with no actions defined does nothing when used credits reach a threshold. Each monitor can have one Suspend action, one Suspend Immediate action and up to five Notify actions.

triggerDefinition ::= ON <threshold> PERCENT DO { SUSPEND | SUSPEND_IMMEDIATE | NOTIFY }

SUSPEND has a detail that matters in a scenario question. After it fires, the warehouses cannot run new queries until the monitor's credit quota is increased. A common pattern escalates in stages, as in the example below. A Suspend at 80% lets running work finish, and a Suspend Immediate at 100% is the hard stop that cancels anything still running.

Notifications go to all account administrators who have notifications enabled. For warehouse monitors, you can also email up to five non-administrator users through NOTIFY_USERS. Each of those users needs a verified email address, or the statement fails. Non-administrators get email only and can't see the notifications in Snowsight. They cannot receive notifications from account monitors.

USE ROLE ACCOUNTADMIN; CREATE OR REPLACE RESOURCE MONITOR limit1 WITH CREDIT_QUOTA=2000 FREQUENCY = WEEKLY START_TIMESTAMP = '2019-03-04 00:00 PST' TRIGGERS ON 80 PERCENT DO SUSPEND ON 100 PERCENT DO SUSPEND_IMMEDIATE; ALTER WAREHOUSE wh1 SET RESOURCE_MONITOR = limit1; ALTER WAREHOUSE wh2 SET RESOURCE_MONITOR = limit1;

ALTER RESOURCE MONITOR my_warehouse_rm
  SET NOTIFY_USERS = (USER1, USER2);

Sources12

4.Account-level vs warehouse-level monitors

A monitor's type decides which warehouses it controls. A monitor that is neither set on the account nor assigned to any warehouse monitors nothing. It simply remains dormant.

A warehouse monitor can cover one or more warehouses, but each warehouse can be assigned to only one warehouse monitor. A warehouse can sit under its own monitor and the account monitor at the same time. The account monitor does not override the warehouse assignment: the warehouse is suspended when either limit is reached.

Take the documented example. The account monitor has 5000 credits per interval. Warehouse 3 has its own 1000-credit monitor, and Warehouses 4 and 5 share a 2500-credit monitor. Warehouses 3, 4 and 5 may stop below their own quotas if the account's 5000 is reached first.

The type also affects cloud services. A warehouse-level monitor can count cloud services usage but cannot suspend it. After a warehouse is suspended, queries against it can still incur cloud services costs. An account monitor controls only the warehouses you create. It does not control the Snowflake-provided compute behind serverless features such as Snowpipe, Automatic Clustering and materialized views.

USE ROLE ACCOUNTADMIN; CREATE RESOURCE MONITOR my_account_rm WITH CREDIT_QUOTA=10000 TRIGGERS ON 100 PERCENT DO SUSPEND; ALTER ACCOUNT SET RESOURCE_MONITOR = my_account_rm;

To check whether an account monitor exists, run SHOW RESOURCE MONITORS. The LEVEL column shows whether each monitor is set for the account or for individual warehouses. To find gaps, the query below lists warehouses with no warehouse-specific monitor. It doesn't check for an account-level monitor, so a listed warehouse may still be under the account's limit.

SHOW WAREHOUSES
  ->> SELECT "name" AS WAREHOUSE_NAME,
             "size" AS WAREHOUSE_SIZE
        FROM $1
        WHERE "resource_monitor" = 'null';

Sources1

5.Modifying monitors and who may do it

Business requirements change, and most monitor properties can change with them. You can raise or lower the quota, customize the schedule, add or remove actions, and change thresholds. You can also convert an account monitor to a warehouse monitor and back, and add or remove warehouses. In Snowsight, go to Admin » Cost management » Resource Monitors, select the monitor, and choose Edit from the More menu. In SQL, use ALTER RESOURCE MONITOR.

Two rules limit what you can change. First, changes do not affect the used credits counted so far. They apply only to usage after the change is saved. Second, a customized schedule cannot be set back to the default. To return to the default, you must drop the monitor and create a new one.

ALTER RESOURCE MONITOR limit1 SET CREDIT_QUOTA=3000;

You can't convert a warehouse monitor to an account monitor with ALTER RESOURCE MONITOR. Instead, you detach it from its warehouses and then attach it to the account. First, run SHOW WAREHOUSES and find the warehouses whose resource_monitor column names the monitor. Then unset the monitor on each of those warehouses. Finally, set the monitor on the account.

ALTER WAREHOUSE my_wh2 UNSET RESOURCE_MONITOR; ALTER WAREHOUSE my_wh3 UNSET RESOURCE_MONITOR;

ALTER ACCOUNT my_account SET RESOURCE_MONITOR = my_rm;

Sources1

6.Setting up a dashboard to monitor Snowflake costs

A resource monitor enforces a limit, but it doesn't show where the money goes. For that, Snowflake offers two routes. Snowsight provides ready-made visual dashboards of cost. Queries against the ACCOUNT_USAGE and ORGANIZATION_USAGE views let you drill down and build custom reports and dashboards. The Snowsight pages are under Admin » Cost management. Snowflake recommends selecting an X-Small warehouse to view the usage data.

Two caveats apply to every Snowsight cost page. Cost information can take up to 72 hours to appear, and it is shown in UTC rather than your local time zone.

Account Overview is the account's starting point for cost. Two filters at the top control every tile. *Date range* defaults to This month, and *Currency* defaults to credits; switching to a currency such as USD needs extra privileges. The tiles fall into three groups:

- Summary cards: total cost against the previous period, budget used, an anomaly count, and estimated savings. The budget card works only if a monthly budget is configured for the account. - Cost breakdown: a spend-over-time chart grouped by Services or by Usage type (Compute, Storage, Data transfer, AI inference). It also has a Cost attribution tile and a top-spenders list of warehouses, queries, databases or tags. - Alerts and recommendations: an anomalies table with the expected range for each day, plus optimization insights.

The Cost attribution tile handles chargeback. A cost attribution tag is any tag applied to a warehouse for chargeback. If you group the tile by a tag such as COST_CENTER, it splits spend by that tag's values. A Without selected tag row collects warehouses that are missing that tag.

The Consumption page has one access limit to know. To use all its features, the account administrator must also have ORGADMIN. An ACCOUNTADMIN without ORGADMIN sees only the current account.

For a custom dashboard, query the usage views directly. ORGANIZATION_USAGE covers every account in the organization, and ACCOUNT_USAGE gives similar data for a single account. The query below totals the last month's spend in currency for each account. It is the kind of query that would back a dashboard tile.

The Snowsight dashboards documentation in these sources carries a deprecation notice. It says Legacy Dashboards will be removed from Snowflake on June 22, 2026. These sources don't describe the replacement. Check current documentation before relying on Projects » Dashboards to host such a chart.

SELECT account_name, ROUND(SUM(usage_in_currency), 2) as usage_in_currency FROM snowflake.organization_usage.usage_in_currency_daily WHERE usage_date > DATEADD(month,-1,CURRENT_TIMESTAMP()) GROUP BY 1 ORDER BY 2 desc;

Sources345

Exam traps

Each one states something that sounds right. Open it to see what is actually true.

1. 1.A warehouse with its own resource monitor is exempt from the account-level monitor, because the more specific monitor wins.Why is that wrong?

   Both monitors apply at the same time. If either one reaches a suspend threshold, the warehouse is suspended.

   Covered in Account-level vs warehouse-level monitors

2. 2.An account-level resource monitor caps all credit spend in the account, including Snowpipe, automatic clustering and materialized views.Why is that wrong?

   Resource monitors control only warehouses you create. Serverless features and AI services need a budget instead.

   Covered in What a resource monitor does and does not watch

3. 3.The quota counts only billed cloud services credits, after the daily 10% adjustment.Why is that wrong?

   The monitor counts all cloud services consumption, including the part that is never billed. It can reach its limit sooner than the bill suggests.

   Covered in What a resource monitor does and does not watch

4. 4.Trigger thresholds must be between 1 and 100 percent of the quota.Why is that wrong?

   Thresholds above 100 are supported, so a monitor can notify or suspend after the quota has been exceeded.

   Covered in Triggers: notify, suspend, or suspend immediately

5. 5.An ALTER RESOURCE MONITOR can return a customized schedule to the default calendar-month reset.Why is that wrong?

   A customized schedule cannot be set back to the default. You must drop the monitor and create a new one.

   Covered in Modifying monitors and who may do it

Practise it for real

Put a weekly credit limit on a warehouse with staged suspension, then raise the limit

1. 1.As ACCOUNTADMIN, run CREATE OR REPLACE RESOURCE MONITOR limit1 WITH CREDIT_QUOTA=2000 FREQUENCY = WEEKLY START_TIMESTAMP = IMMEDIATELY TRIGGERS ON 80 PERCENT DO SUSPEND ON 100 PERCENT DO SUSPEND_IMMEDIATE;

   Why: Only ACCOUNTADMIN can create monitors. FREQUENCY and START_TIMESTAMP must be set together.

   You should see: The monitor exists but stays dormant, because nothing is assigned to it yet.

2. 2.Run ALTER WAREHOUSE wh1 SET RESOURCE_MONITOR = limit1;

   Why: Assigning a warehouse makes this a warehouse-level monitor.

   You should see: wh1 can now be suspended at 80% and 100% of 2000 credits.

3. 3.Run SHOW RESOURCE MONITORS and check the LEVEL column for limit1

   Why: The LEVEL column shows whether a monitor applies to the account or to individual warehouses.

   You should see: limit1 is shown as a warehouse-level monitor.

4. 4.Run ALTER RESOURCE MONITOR limit1 SET CREDIT_QUOTA=3000;

   Why: Changing the quota needs MODIFY on the monitor, not ACCOUNTADMIN.

   You should see: The new quota applies from now on, and credits already used this week are unchanged.