Continuous Deployment with Heroku
Quick start videos
Watch these videos for a step-by-step walkthrough of setting up a continuous delivery pipeline with CircleCI and Heroku:
Part 1: Green builds and red builds
Part 2: Multiple environments
You can also look at the GitHub repo used in the above videos here.
You must supply or confirm some authentication settings before Heroku deployment. CircleCI dividies this setup work into three steps.
Note: Heroku authenticates using the API key set in your Heroku dashboard Account page; this API key must be set before proceeding.
Step 1: Enter the API key (if not autofilled) and click Save.
If this key has already been added to your Project Settings > Heroku settings page, CircleCI maintains security by displaying a shorten version of the existing API key and masking the key when autofilling the Set key field.
Note: Heroku’s architecture and security model require that you deploy as a particular user. A member of your project—possibly you—needs to register as that user as the second step in this process.
Step 2: Click the button to set the registered person as the Heroku deploy user and to associate a Heroku SSH key with the deploy user’s account.
Specify the app using the
Note: the Heroku
appname must not include upper case characters.
deployment: staging: branch: master heroku: appname: foo-bar-123 # appname must not include upper-case characters
You can now use the
git push command.
Upon a successful build, we’ll automatically deploy to the app in the section that matches the push, should there be one.
Note: If you are deploying several projects from CircleCI consider using different keys for each one. Heroku enforces a rate-limit of 75 git requests per hour per key.
Heroku with pre- or post-deployment steps
If you’d like to run commands before or after deploying to Heroku, you’ll have to use the ‘normal’
deployment: production: branch: production commands: - "[[ ! -s \"$(git rev-parse --git-dir)/shallow\" ]] || git fetch --unshallow" - git push email@example.com:foo-bar-123.git $CIRCLE_SHA1:refs/heads/master - heroku run rake db:migrate --app foo-bar-123: timeout: 400 # if your deploys take a long time
Note the ‘git fetch –unshallow’ command; we do shallow clones by default, which can’t always push to heroku.
Heroku's maintenance mode
When you are pushing new code that requires a database migration, your app will not be available. To be polite, you might want to treat your users to more than an enigmatic error page.
Heroku offers a built-in maintenance mode when you need to disable an app for some period of time. This mode blocks all HTTP requests and displays a generic HTML page telling users that service is unavailable because of maintenance (more about customizing this 503 error message page below).
You enable maintenance mode with the
heroku maintenance:on command.
Once your work is completed, you turn the mode off with the
Enabling and disabling maintenance mode is sufficient for most simple updating work such as adding a new graphic or making wording changes,
however, when you are performing a database migration Heroku suggests that you use the
heroku scale command to reduce the number of active Heroku’s worker dynos.
To ensure that no background jobs are processed during migration, scale down all the workers.
Here is how to tweak the
section of your
circle.yml file to employ these commands.
deployment: staging: commands: - heroku maintenance:on --app foo-bar-123 - heroku scale worker=0 --app foo-bar-123 - git push firstname.lastname@example.org:foo-bar-123.git $CIRCLE_SHA1:refs/heads/master - heroku run rake db:migrate --app foo-bar-123 - heroku scale worker=x --app foo-bar-123 - heroku maintenance:off --app foo-bar-123
Custom maintenance page
As previously mentioned, Heroku displays a default “Application Offline for Maintenance” HTML page telling users that an application is currently unavailable and to check back later. Should you wish, you can create a custom service unavailable page for your app.
Heroku and database migrations
After pushing new code, you may wish to run your database migration. Migration times will vary between apps because of data size and structure.
rake db:migrate command invokes several tasks including
` db:schema:dump`, which updates the db/schema.rb file to match the database structure.
CircleCI supports such command-driven database migrations as well as database migrations performed manually.
Before migrating, Heroku recommends that you use the
capture a snapshot
so that you can easily revert the database back to its previous state.
Here are migration-related articles that you might find useful.
Clearing caches after a deployment
If you are using a caching add-on, it is likely that you want to clear your caches after a deploy to avoid using out-of-date and possibly corrupt documents and/or data. Heroku does not clear caching by default, but you can install an add-on that assists with cache management.
Add-ons include settings, such as expiration time settings for clearing cache pages,
as well as commands that can assist with such housekeeping.
MemCachier supports a
Flush command that clears your entire cache.
You might want to write a standalone script that uses the memcache client to call
Flush and run the script at deployment.
Here are related articles that you might find useful.