Stop/Start Amazon Redshift Clusters

Many of Amazon Redshift Customers are wondering if they can Stop / Start their Redshift Cluster as many of other AWS instances and services to reduce incurring cost. After all idea here is that with cloud you pay for only what you use. And it true for many AWS services like RDS instances and even Snowflake  ( Redshift competitor).

When it comes to Redshift clusters, it’s not straight forward to simply stop and start clusters. It’s not because Amazon is deliberately trying to make it hard for customer to stop/start their clusters, basically Redshift architecture simply won’t allow this functionally without taking additional steps.

Basically, Redshift uses local storages on compute nodes to store data in slices and if shut them down, you will not release compute/storage for other cloud customers. Hence AWS cannot provide this functionally without incurring any charges. Amazon is releasing new features for Redshift on Regular bases and you might see this behavior change in future.    

You can still shutdown Redshift instances to reduce incurring charges if you are using ON-Demand clusters by using following steps. Basically, you will need to take a snapshot and delete the cluster then restore you cluster from snapshot when you need it back online.

 To Stop/Delete:

1. From AWS Console, select the cluster you want to stop

2. Drop down the “Cluster” menu and select “Delete”

3. Enter the Snapshot name; I use the same name as the cluster I’m deleting

To Start/Restore:

1. In the Redshift Snapshots select the snapshot of the cluster you want to restore

2. Drop down the “Actions” menu, and select “Restore From Snapshot”

3. Complete the configuration details, and click the “Restore” button at the bottom right

Important Considerations:

  1. You organization might not be like the idea of delete a production cluster with sensitive data just to safe cost.
  2. If you looking to reduce cost with Redshift, it best to opt of reserve nodes as they are offered on significantly lower cost vs. on-demand cluster nodes.
  3. It might take 30 mins to an hour to stop/start Redshift Cluster using snapshots for terabyte database.
  4. Snowflake might be a better option for customers with on-demand workload , because you can easily shutdown and start Snowflake clusters in seconds without deleting any data as it using S3 buckets to store data.