Why it matters
Development and test warehouses are rarely queried at night or at weekends, but on-demand nodes bill every hour.How nOps finds it
nOps reads the last 14 complete days of on-demand Redshift node billing. A cluster is flagged when at least 90% of its node hours are non-production (by environment tag, the environment nOps assigns, or its name), it was billed on all 14 days, ran at least 20 hours a day on average and ran at least 20 hours on at least 3 of the 4 weekend days. The saving is the on-demand node cost of the last 7 days, scaled to a month, times the share of the week outside a 60-hour schedule (1 minus 60/168). Clusters mostly on reserved nodes or credits, and savings under $5 a month, are left out.Before you start
Agree working hours with the cluster’s users. Scheduled actions need an IAM role that Redshift Scheduler can assume.How to implement
- Check what runs on the cluster at night and at weekends, such as ETL loads and dashboards.
- In the Redshift console, open the cluster and choose Actions → Pause and resume (or Schedules → Add schedule) to create a pause schedule for the evening and a resume schedule for the morning on weekdays.
- Or create scheduled actions with the CLI (
aws redshift create-scheduled-actionwith a pause-cluster and a resume-cluster action). - Move any overnight jobs into working hours.
Risks and considerations
- A paused cluster cannot be queried. Jobs that run outside the schedule fail.
- Storage and snapshots are still billed while paused.
- Resuming takes a few minutes.
- To roll back, delete the schedules and resume the cluster.
FAQs
What schedule does the estimate assume?
What schedule does the estimate assume?
12 hours a day on weekdays, 60 of the 168 hours in a week. A longer schedule saves less.