# Associate Connection Functions with distributions
<a name="associate-connection-functions"></a>

After publishing a Connection Function to the LIVE stage, you must associate it with an mTLS-enabled distribution to handle live connections. Connection functions are associated at the distribution level, unlike viewer request and viewer response functions which are associated with cache behaviors.

**Topics**
+ [Association requirements](#connection-function-association-requirements)
+ [Organizing functions with filters](#connection-function-organizing-filters)
+ [Deployment considerations](#connection-function-deployment-considerations)

## Association requirements
<a name="connection-function-association-requirements"></a>

To associate a Connection Function with a distribution:
+ The function must be in the LIVE stage
+ The distribution must have mTLS enabled
+ The distribution must have a valid trust store configured
+ You can only associate one Connection Function per distribution

## Organizing functions with filters
<a name="connection-function-organizing-filters"></a>

CloudFront provides filtering capabilities to help you organize and manage connection functions:
+ **Distribution ID filter** – Find functions associated with specific distributions
+ **Key-value store filter** – Find functions that use specific key-value stores for data lookup
+ **Stage filter** – List functions in DEVELOPMENT or LIVE stage

Use these filters when managing multiple Connection Functions across different distributions or development environments.

## Deployment considerations
<a name="connection-function-deployment-considerations"></a>

Consider these factors when deploying Connection Functions:
+ **Global deployment** – Connection functions deploy to all CloudFront edge locations worldwide, which may take several minutes
+ **Version management** – Each published version creates a new LIVE function that replaces the previous version
+ **Rollback strategy** – Plan for rollback by keeping previous working versions of your function code
+ **Testing in production** – Consider using separate distributions for staging and production environments