Heroku Developer Accredited Professional Practice Exam

Correct: B — a worker process. The Procfile declares process types, and only the process type named web receives HTTP traffic from Heroku's routing layer. Worker processes are the conventional home for background job processing and queue consumers.

Why not the others: Only web dynos receive HTTP traffic (A), so the premise is wrong. A release process (C) runs once as part of a release, before the new version goes live. One-off dynos (D) are for ad-hoc tasks and are not a durable way to run a continuous queue consumer.

Correct: C — layer multiple buildpacks. The exam guide names knowing when to layer several buildpacks to provide advanced functionality as an explicit objective. Buildpacks run in the order they are added, so a library buildpack can prepare the environment before the language buildpack builds the application.

Why not the others: Committing binaries (A) bloats the repository and defeats the build system. Installing at runtime (B) slows every dyno start and breaks the immutability of the slug. Container deploys (D) are a legitimate alternative but are not the answer when the question is about buildpack composition.

Correct: A — run more web dynos. Horizontal scaling adds dynos so more requests are served in parallel, which is the direct lever for concurrency. The exam guide frames this objective as appraising the various ways Heroku apps can be scaled during high-traffic events.

Why not the others: A follower database (B) helps read-heavy database load, a different bottleneck. Slug size (C) affects deploy and boot, not request concurrency. Adding Procfile entries (D) declares new process types rather than scaling an existing one.

Correct: D — a Heroku Button. The Elements objective names add-ons, buildpacks, and Heroku Buttons together. The Button is the one-click deploy mechanism, driven by an app manifest that declares the add-ons and configuration a new deployment needs.

Why not the others: A buildpack (A) contributes to the build process but does not provision a whole app. A Dataclip (B) shares a database query result. Pipelines with review apps (C) support the internal development lifecycle, not public one-click deployment.

Correct: B — review apps in a pipeline. Review apps create a temporary application from the code in a pull request, letting reviewers test the change in isolation before it is merged. They are part of Heroku Flow alongside pipelines and CI, which is exactly how the exam guide frames this objective.

Why not the others: Follower databases (A) are read replicas. A release phase command (C) runs during deployment, useful for migrations but unrelated to per-PR environments. Config vars (D) hold configuration and are not branch-scoped by themselves.

Correct: C — a Dataclip. Dataclips turn a SQL query into a shareable, refreshable result set accessible by URL, without handing out database credentials. The exam guide lists understanding Dataclips and how they are used as an explicit objective alongside Postgres features.

Why not the others: A fork (A) creates a full point-in-time copy of the database, which is heavyweight for sharing one query. The Key-Value Store (B) is for ephemeral key-value data, not relational query sharing. A Kafka topic (D) is a streaming channel between services.

Correct: C — Apache Kafka on Heroku. The exam guide names designing solutions that use Apache Kafka and Kafka topics to facilitate streaming communication between applications, services, and functions. Topics are the durable, replayable channel that lets consumers progress independently.

Why not the others: Polling a Postgres column (A) works at small scale but does not suit high-volume streaming. The Key-Value Store (B) can back simple queues but lacks the durability and consumer-group semantics of Kafka. Dataclips (D) share query results and are not a messaging system.

Correct: A — Automated Certificate Management. ACM issues and renews TLS certificates for custom domains automatically, removing the expiry-driven manual work that causes outages. The exam guide names understanding ACM and its importance as an explicit objective in the security area.

Why not the others: A self-signed certificate (B) is not trusted by browsers and still requires manual renewal. A buildpack (C) acts at build time and has no role in certificate lifecycle. Storing key material in config vars (D) is neither the mechanism nor good practice.

Correct: B — Heroku Connect. Heroku Connect provides managed synchronisation between Heroku Postgres and Salesforce objects. The exam guide expects familiarity with its features including which Salesforce APIs it uses and which write algorithm it supports — detail that only matters because the service does the sync for you.

Why not the others: A custom sync script (A) is the manual work Heroku Connect exists to replace. A Kafka bridge (C) can move events but does not provide table-level bidirectional mapping. Dataclips (D) query Heroku Postgres and cannot reach into a Salesforce org.

Correct: D — stored in the environment. The config factor of the Twelve-Factor methodology requires strict separation of config from code, with anything that varies between deploys held in environment variables. On Heroku this maps directly onto config vars, which is why the methodology carries a full 13% of this exam.

Why not the others: Per-environment config files in the repository (A) is the specific anti-pattern the factor rules out, since credentials end up in version control. Hard-coding (B) is worse still. Baking config into the slug (C) breaks the principle that one build artifact should run in any environment.