“We’ve been unable to VACUUM for awhile.”

If you received this notification from us, it means that Stitch hasn’t been able to successfully perform VACUUM on some tables in your data warehouse for more than 10 days.

To keep things tidy in your data warehouse, Stitch will occasionally execute a VACUUM command after tables that use Full Table Replication have finished replicating. The purpose of VACUUMing is to reclaim disk space and reduce storage. You can read more about vacuuming in Redshift here.

Stitch needs to be able to execute this command to keep these tables from growing too large with deleted rows.

Resolving VACUUM Errors

Typically these errors are caused by insufficient permissions associated with the Stitch Redshift user.

The other cause has to do with the number of tables that use Full Table Replication and how often they’re replicated.

Verify the Stitch Redshift User’s Permissions

The Redshift user that Stitch uses to access your data warehouse must be the owner of all the integration schemas and tables created by Stitch. This will ensure that Stitch can successfully run the VACUUM command.

Verify that the Stitch user is the owner of the tables listed in the notification you received. If Stitch isn’t the owner, update the user’s permissions.

Manually Vacuuming Tables

If a large number of tables are set to Full Table Replication and these tables are replicated frequently, Stitch may be unable to execute the VACUUM command due to how the timing works out in the replication process.

For example: if a large number of fully replicated tables are replicated back-to-back, there might not be a break in the process during which Stitch can VACUUM. The likelihood of this occurring will increase with the higher number of tables that use Full Table Replication and more frequent replication attempts.

Stitch will wait 10 days to notify you about the issue if this is the root cause - we do this to give things time to work themselves out. If you receive the VACUUM error AND the permissions are correct, we recommend that you VACUUM the affected tables manually to ensure they stay at a reasonable size.

Questions? Feedback?

Did this article help? If you have questions or feedback, feel free to submit a pull request with your suggestions, open an issue on GitHub, or reach out to us.