Check whether Redshift Serverless limits stop queries
On this page
Warnings do not stop Redshift Serverless queries. Compare charged RPU-hours with capacity and RPU-hour limits, then keep a warning or choose query shutdown deliberately, because skipping this can allow usage to continue or interrupt workloads.
What you need first
- Tool
- Use a SQL client, a tool for sending database queries, connected to the target Redshift Serverless workgroup. Run the read-only SELECT against SYS_SERVERLESS_USAGE, the system view for recent Serverless resource usage.
- Access
- SYS_SERVERLESS_USAGE is visible only to database superusers. For the settings review, ask an authorized AWS colleague to provide the current capacity settings and usage-limit actions for the same workgroup.
- If you do not use that tool
- Give your Redshift administrator the target workgroup and the SQL below. Request all daily results returned for the retained seven-day window, noting any partial boundary days, plus current base and maximum capacity and every usage limit's RPU-hours, frequency, and action.
Why this is worth a look
Skipping this review can leave limits that notify without stopping query processing, so usage can continue. Turning off user queries disables processing and creates an availability risk. Redshift meters compute in RPU-seconds on a per-second basis. Max capacity limits scaling, while RPU-hour limits limit accumulated usage over a daily, weekly, or monthly period. These are usage controls, not currency budgets.
Start with this query
SQLRun this read-only query in a SQL client connected to one Serverless workgroup as a superuser. It groups the retained seven days by calendar date and returns RPU-seconds, RPU-hours, average RPUs, and average storage in MB, not currency.
SELECT
DATE_TRUNC('day', start_time) AS usage_day,
SUM(charged_seconds) AS charged_rpu_seconds,
SUM(charged_seconds) / 3600.0 AS charged_rpu_hours,
SUM(compute_seconds) AS compute_rpu_seconds,
SUM(compute_seconds) / 3600.0 AS compute_rpu_hours,
AVG(compute_capacity) AS avg_rpu_capacity,
AVG(data_storage) AS avg_data_storage_mb
FROM sys_serverless_usage
GROUP BY 1
ORDER BY 1 DESC;How to confirm it
- 01
Read recent charged usage
Run the SQL or request its results from your administrator. Use charged_rpu_hours for the workgroup's charged compute usage. Compare daily totals with daily limits, and do not treat the retained seven days as a full monthly total.
- 02
Record capacity settings
Ask the workgroup owner to confirm base capacity and maximum capacity in RPUs. Base capacity serves queries, while maximum capacity limits scaling. Do not treat either setting as a limit on accumulated RPU-hours.
- 03
Read each limit action
Open the workgroup's Limits tab in the Amazon Redshift Serverless console and choose Manage usage limits. Record each compute limit's RPU-hours, Daily, Weekly, or Monthly frequency, and action without saving changes. Ask an authorized colleague for these values if needed.
- 04
Choose warning or shutdown
Confirm with the workload owner whether each limit should log, alert, or turn off user queries. Request a separate settings change if the action conflicts with that decision. Do not lower base capacity during this review because editing it might cancel running queries.
Before making changes
Assume the SQL connection and settings refer to the same Serverless workgroup. SYS_SERVERLESS_USAGE retains seven days at one-minute granularity, so grouping by calendar day can return eight dates with partial boundary days. charged_seconds is calculated after transactions end and can be zero while they run. compute_seconds reflects account-allocated base capacity, not the same workgroup charge measure. The results describe usage, not a complete bill.