How do I resolve my AWS DMS CDC tasks that are paused with no errors indicated?

Short description

AWS DMS reads the source database transaction logs and translates the changes before they are passed to the sorter component. The sorter component manages incoming changes in the commit order and forwards the changes to the target apply component of the AWS DMS task. If the workload contains a high volume of CDC changes, then task spill changes take place in the sorter to disk. If the replication instance storage reaches 90% of capacity, then CDC ongoing changes are paused. You might receive the following message from the sorter component:

"Reading from source is paused. Total disk usage exceeded the limit 90%"

To resolve the CDC ongoing changes paused issue, take one of the following actions on the replication instance:

• Reduce the storage usage
• Increase the storage size

Note: To prevent this issue in the future, configure AWS Simple Notification Service (AWS SNS) notifications for AWS DMS event ID DMS-EVENT-0092.

Resolution

Reduce the storage usage

To reduce storage usage, complete the following tasks:

• Delete any large task logs, or automate the error log deletion process.
• Use Amazon CloudWatch metrics to monitor the tasks. Move the tasks with a high rate of change to a different replication instance.
• Delete unwanted ongoing replication tasks that are paused.
• Turn off detailed debug logging for AWS DMS tasks. Change your debug logging to default settings to avoid high storage usage.

Increase the storage size

Note: After you increase the allocated storage size, you can't decrease the size.

To increase the storage size of a replication instance, complete the following steps:

1. Open the AWS DMS console.
2. From the navigation pane, choose Replication instances.
3. Choose the name of your replication instance.
4. Choose Modify.
5. For the Allocated storage (GB) field, enter a new value.
6. Select Apply changes immediately, and then choose Modify.

Related information

Why is my AWS DMS replication DB instance in the storage-full status?