Heroku Connect Maintenance Operations

Last updated August 30, 2021

This article covers Heroku Connect maintenances. It also covers how to change your Heroku Postgres plan or upgrade its version.

Heroku Connect Maintenances

From time to time, Heroku needs to take Heroku Connect offline to perform maintenance tasks. Typical tasks include updating the underlying infrastructure. This maintenance is handled automatically by Heroku. Unlike Heroku Postgres, you can’t schedule or reschedule maintenance windows for Heroku Connect.

Maintenances typically take up to 45 minutes. During maintenance, all data synchronization for Heroku Connect is unavailable. Configuring Heroku Connect is also unavailable. Existing data in the Heroku Postgres database remains accessible. Connect retains changes and resumes synchronizing after the conclusion of maintenance.

Preparing for Heroku Connect Maintenances

Maintenance windows are announced roughly two weeks before they happen on status.heroku.com. You can click Subscribe on that page to get notifications.

There is nothing you must do to prepare besides being aware of and preparing for the delay in data sync during the maintenance window. Connect resumes normal sync following the maintenance operation.

Changing the plan of your Heroku Postgres database

If you want to change the plan of your Heroku Postgres Database, you can use the heroku addons:upgrade command to do so. Heroku Connect is able to handle this operation transparently since it behaves identically to an HA failover.

While not recommended, you could use the Follower Changeover method instead. If you wish to do that then you would need to take additional steps as those documented in the section below, omitting step 4.

If you’re upgrading from a hobby tier database, you won’t be able to create a follower. Instead, use the PG Copy process instead to copy data into the new database. Simply pause Heroku Connect at the same time as you put your application into maintenance mode, and resume it again when you reactivate your application.

Upgrading the version of a Heroku Postgres database

Heroku Connect operates continuously on a single database, which requires a few steps in order to upgrade the version of that database and allow sync operations to continue as expected. The upgrade itself is done using a follower database and pg:upgrade, and these instructions illustrate how Heroku Connect fits into that process.

These steps ensure that the new database is an exact copy of your current database. Heroku Connect will not create tables or transfer data to an empty database, so be sure to follow the proper upgrade process.

  1. Create your follower database and allow it to get mostly caught up with your lead database. This will minimize the amount of downtime required for the upgrade.

  2. Pause your connection for the same reason that you put your application into maintenance mode, to prevent Heroku Connect from trying to write to your database during the upgrade.

  3. Wait for your follower to catch up. Now that Heroku Connect and your application have both stopped writing to your database, the follower can get completely caught up. You can check its status using heroku pg:info.

  4. Upgrade the follower database using pg:upgrade. This will have the follower un-follow the leader and perform the Postgres Version upgrade.

  5. You now need to promote the follower database to become the new lead database. The steps you need to take depend on whether your Heroku Connect is using a standard database URL (DATABASE_URL) or a custom database URL (e.g. HEROKU_POSTGRESQL_ROSE_URL).

    • If you’re using a standard database URL, Promote the follower to become the new lead database. This will make it accessible as DATABASE_URL.
    • If you’re using a custom database URL, you will need to update the Heroku Connect Database Settings with the new custom database URL for the promoted follower or use heroku addons:attach to attach the database to a name of your choosing.

  6. Resume Heroku Connect using the same process used to pause it in step 2, and bring your application out of maintenance mode. Heroku Connect will automatically pick up the new DATABASE_URL for your application ( or customer database URL, if you have that configured ) and resume synchronization using the new database.

Recovering synchronization after a database rollback

Heroku Connect uses a high-water mark methodology. Rolling back a synced database causes inconsistent data. Reload mappings after a database rollback to recover data consistency.

Pausing synchronization

When performing certain tasks, such as upgrading your database, you may want to pause synchronization activity in Heroku Connect. To do this navigate to the Settings tab, Manage Connection and click the Pause button.

Heroku Connect will complete any pending operations before entering the paused state. When your connection is paused then changes to data in the database will continue to be added to the trigger log and changes made in Salesforce will not be polled. Push Topic notifications from Salesforce for mappings using the streaming mode will also be ignored.

Connections can only be paused when they are in the IDLE state. If your connection remains in the POLLING_DB_CHANGES status for an extended period of time, check your logs to debug.

To resume synchronization return to the Settings tab, Manage Connection and click the Resume button. Heroku Connect will continue where it left off: entries in the trigger log will be processed and pushed to Salesforce and changes made in Salesforce will be retrieved using polling (including those mappings that would normally use the Streaming API).

Any changes made to data in the database or in Salesforce while your connection is paused should automatically be synchronized when your connection is resumed.

Recreating a connection

If you’ve reloaded a sandbox organization or you need to upgrade from a hobby database to a professional one, you’ll need to delete your existing connection and create a new one.

The following settings are manually configured:

  • Collaborators and their notifications settings
  • Beta features
  • The Write Algorithm
  • The Salesforce API version

The rest of the recreation can be automated as follows:

  1. Export your configuration from the Connect Dashboard. Click Settings, then click Import/Export Configuration, then click the Export Configuration button.
  2. If you have any read/write mappings, make sure you aren’t writing any new data to those tables, and make sure all the mappings are in DATA_SYNCED to ensure all your data is present in Salesforce. You may need to click Settings, Manage Connection and click the Pause button to stop activity on the connection.
  3. Destroy the Connection from the heroku CLI heroku addons:destroy -a appname herokuconnect-resource-name
  4. Create a new connection from https://elements.heroku.com or heroku addons:create herokuconnect from the command line.
  5. Authenticate your new connection, and import the config you exported from step 1.

Changing field definitions in Salesforce

Field definition changes can delay synchronization with Heroku Connect if not coordinated properly.

Heroku Connect does not automatically change the table schema in PostgreSQL after you’ve changed the corresponding field definition in Salesforce. This is because attempting to change the field may result in data loss or corruption.

After changing a field in Salesforce, users must edit the object mapping in Heroku Connect to trigger an update to the PostgreSQL table schema by Heroku Connect. In simpler cases, all users need to do is click Edit for the mapping and then click Save. In more complex cases, users will need to deselect the field, save the mapping, edit the mapping again, and then re-add the field. This will cause Connect to drop the column, re-add it, and load the data back into the column.

Simpler cases include extending or shortening the length of a field. More complex cases include changing the types of fields.

Feedback

Log in to submit feedback.