Short answer: The Heroku-22 stack reaches end of life on April 30, 2027, and no builds run from May 1, 2027. Apps keep running unpatched. Upgrade to Heroku-24 (supported to April 2029) or Heroku-26 (April 2031).
The Heroku-22 stack reaches end of life on April 30, 2027, according to the official Heroku-22 end-of-life FAQ, and from May 1, 2027 Heroku allows no further builds for any Heroku-22 app. Your dynos keep running after that, but they stop receiving security updates and you lose the ability to ship code until the app moves to a newer stack. Heroku's own guidance is to move to the latest stack, so our recommendation is Heroku-26 (supported to April 2031), tested first on a Review App or a throwaway copy of each app, then promoted through your pipeline; Heroku-24, the current default and supported to April 2029, is the fallback when a buildpack or dependency is not ready for 26.
Dates come from Heroku's official stack support table, the Heroku-22 FAQ and the Dev Center changelog, checked November 21, 2026.
What makes this deadline unusual is the ramp in front of it. Heroku is not waiting for April to apply pressure. It is breaking builds deliberately, on a schedule that tightens three times:
The FAQ is explicit that the May 1 builds cutoff is a hard deadline with no exceptions, and that this applies to enterprise customers too.
Heroku-22 is built on Ubuntu 22.04 LTS, and its end of life tracks the end of Canonical's standard five-year support for that release. On Heroku, the platform patches the OS layer for you. That is the service that ends on April 30, 2027. Servers, containers and CI runners you run yourself on the same base OS follow Canonical's calendar instead; our Ubuntu 22.04 end of standard support page covers them.
Heroku says existing Heroku-22 apps will not be shut down at end of life or when builds are disabled. Restarting dynos and running one-off dynos keep working. Heroku's stack update policy describes retired stacks the same way: apps continue to run indefinitely, but can no longer be deployed without first upgrading. So the failure mode is not an outage. It is a frozen app on an OS image that no one is patching.
This is the part teams underestimate. After May 1, 2027, a Heroku-22 app cannot take a new build, which means you cannot deploy a hotfix, a dependency patch or a config-driven rebuild on that stack. If a critical vulnerability lands in your own code or a library, the first step of fixing it becomes a stack upgrade you have not tested. Heroku says you can still upgrade the stack after May 1, 2027, but incident pressure is the worst time to find a buildpack incompatibility.
From November 1, 2026, a deploy that fails with a deprecation error is expected behavior, not a flaky pipeline. A retry generally gets through, because only the first build in each window fails, but by April 2027 that means the first deploy of every day breaks. Scheduled deploys, nightly rebuilds and alerting on failed builds will feel this months early. Heroku also emails owners, admins and collaborators whenever a Heroku-22 app is built, more often as the deadline nears.
SOC 2's CC7.1 criterion asks a company to detect and manage vulnerabilities. An unpatched base OS under production dynos, or an app you cannot redeploy, is hard to square with that, so the useful thing is a dated record: which apps are on Heroku-22, which stack each is moving to, and when. Security questionnaires often ask which platform and OS version each production service runs on, and "Heroku-26, supported to April 2031" is a short answer. We are engineers, not auditors, and nothing here is an attestation.
Heroku accounts accumulate apps: staging copies, demos, add-on-only apps that hold a database, and CI apps the Dashboard hides. All of them count.
To speed up the repository side, paste an app.json, Dockerfile, Gemfile.lock or package.json into the free EOL deadline radar. It returns a dated report of the platform and runtime versions it finds, so a Heroku-22 stack line and an end-of-life Ruby or Node line show up in the same place.
Heroku's stack upgrade guide says to review the notes for every intermediate stack when you jump more than one generation, and it uses Heroku-22 to Heroku-26 as its example. In practice almost all of the breaking changes sit in the Heroku-24 upgrade notes; the Heroku-26 page currently lists no known breaking OS changes from 24 to 26.
From Heroku-24, Git, GCC, Make and the system Python are available only during the build, not inside running dynos. Code that shells out to git at boot, compiles a native extension on first request, or runs the system Python from a release phase or worker will fail. Bazaar and Mercurial are gone entirely.
The system Python in the base image is reachable only through the python3 command; the transitional python symlink is removed. This mostly bites Procfile entries, scheduler jobs and helper scripts, since apps on the Python buildpack get their own interpreter.
The stack now ships only the C, C.utf8, POSIX and en_US.utf8 locales, dropping the other English variants. US/* time zone names are no longer available, so a TZ config var or library call using US/Eastern must change to a name like America/New_York. Both tend to surface in scheduled jobs, not at boot.
The heroku-buildpack-google-chrome buildpack does not work on Heroku-24; Heroku points to the Chrome for Testing buildpack instead. The Redis stunnel buildpack is also incompatible, because Redis 6 and later support TLS natively. Apps that generate PDFs or screenshots with headless Chrome, or that tunnel to Redis, need a buildpack change as part of the move.
The Ruby support table shows Ruby 3.1 built for Heroku-22 and Heroku-24 but not for Heroku-26, and Heroku's upgrade guide says apps on older Ruby versions must move to a supported one when they change stack. A Rails app still on Ruby 3.1 has a choice: upgrade Ruby in the same change, or stop at Heroku-24 for now. See our legacy Rails to Rails 8 runbook.
The Heroku-24 and Heroku-26 Docker images are multi-architecture (amd64 and arm64) and default to a non-root user named heroku. A Dockerfile built FROM heroku/heroku:26 that writes outside /home/heroku or /tmp must switch to root for those steps, and developers on ARM laptops will now pull arm64 images automatically.
Run the apps table listing for every team and build one sheet: app name, owner, pipeline, current buildpacks, language versions, whether it has an app.json, and its target stack. Decide Heroku-26 versus Heroku-24 per app based on the change list above; most apps can go straight to 26. Record response times, error rates and memory per dyno type as a baseline. If an app is a candidate to leave Heroku entirely, upgrade its stack anyway: that keeps it deployable while you plan the exit. Our Heroku to AWS runbook covers that path.
Heroku recommends two ways to test. If the app uses Review Apps, create a branch that sets "stack": "heroku-26" in app.json and open a pull request; the Review App builds on the new stack and you test it there. Otherwise, create a parallel app with heroku create --remote heroku-26 --stack heroku-26 your-app-heroku-26, copy add-ons and config vars (heroku config --shell prints them in a pasteable format), and git push heroku-26 main. Unpin buildpack URLs that carry a # reference, purge the build cache with heroku builds:cache:purge, and run tests and a smoke check against the new app. Destroy the test app with heroku apps:destroy --remote heroku-26 when done.
For apps in a pipeline, do not run stack:set on production. Heroku's guide explains that promotion copies the slug to the downstream app and overwrites its stack to match. Upgrade staging with heroku stack:set heroku-26 -a your-staging-app, push a build (an empty commit works if nothing changed), verify, then promote as usual. For standalone apps, heroku stack:set changes nothing until the next build, so schedule that build deliberately. Move internal and low-traffic apps first.
Each app's cutover is a single build, so watch it closely: release phase output, boot logs, the first scheduled job runs and anything that formats dates or calls external binaries. If something breaks, heroku releases and heroku rollback to the previous version returns you to the old slug. The FAQ confirms that rolling back to a Heroku-22 release remains possible even after end of life, but warns that after end of life the downgrade must go through rollback, not heroku stack:set.
Update every app.json so new Review Apps and Heroku CI runs use the new stack, replace FROM heroku/heroku:22 in Dockerfiles, and remove workarounds such as the Chrome or stunnel buildpacks. Re-run the stack listing until it is empty, and add a CI check that fails on any heroku-22 string.
A Heroku stack move touches every app at once, so it is wider than it looks. The work usually splits like this:
Many teams begin with a pod of 3 to 5 senior engineers (a tech lead plus engineers), and a full pod is about seven. Ryz engineers on your team start from $7,000 per month (our published range is $7,000–$15,000 per engineer per month, mostly mid-level to senior), quoted per team. Once you say go, engineers can typically start in under two weeks. Ryz engineers work on your team, reporting to your leads, in your Heroku teams and repositories. If the real question is whether to stay on Heroku at all, our cloud migration teams can scope that too. Scope your team.
The same build-versus-buy question applies to the SaaS tools around your stack. Our replace runbooks and the SaaS renewal vs custom TCO calculator both export a board-ready build-vs-buy memo you can bring to a budget review.
Your apps will keep running after April 30, 2027, and dyno restarts and one-off dynos keep working, but the stack stops receiving security updates and gets only limited technical support. From May 1, 2027 you also cannot deploy new code to it, including through Review Apps and Heroku CI. An unpatched OS with no way to ship a fix is a risk to close, not a steady state.
Heroku-26, based on Ubuntu 26.04 and supported through April 2031, for most apps. Heroku's upgrade guide says to move to the latest available stack, and it currently lists no known breaking OS changes between Heroku-24 and Heroku-26, so the extra hop costs little. Choose Heroku-24, the current default supported through April 2029, only when something you depend on is not ready for 26 yet, such as an app still on Ruby 3.1, and plan a second move later.
It depends on how many apps you have across teams, whether they sit in pipelines with Review Apps you can test on, how many pinned or retired buildpacks they use, whether any code calls build tools at run time, and whether a language version has to move at the same time. A fleet of buildpack apps on current runtimes moves quickly; apps with custom buildpacks, Docker images or old Ruby versions need individual testing.
No. Heroku's end-of-life FAQ states that it cannot grant extensions to the Heroku-22 builds deadline for security and compliance reasons, and that this applies to all users, including enterprise customers. The only path that keeps an app deployable after May 1, 2027 is upgrading its stack, which Heroku says remains possible after that date.
Ryz engineers start from $7,000 per month, and our published range is $7,000 to $15,000 per engineer per month, mostly mid-level to senior. We quote per team, so the total is the number of engineers times the months the work needs times the monthly rate. You get a scoped plan, a price and the names of the people before anything starts.
Questions we didn't answer? Email info@ryzlabs.com.
Thanks — your message has been sent. We’ll get back to you soon.
Something went wrong while sending your message. Please try again or email info@ryzlabs.com.