An Update on Heroku

Original Article ↗ Hacker News Discussion ↗

Heroku’s announcement that it is moving to a “sustaining engineering model” and ending new enterprise contracts is widely interpreted as a shift to maintenance mode and a slow wind‑down of the platform. Commenters lament the loss of what was once a pioneering, developer‑friendly PaaS, criticize Salesforce’s vague corporate messaging, and recount how pricing, reliability issues, and stalled innovation eroded trust over the years. Many are now comparing alternatives such as Render, Fly.io, Railway, DigitalOcean App Platform, and self‑hosted Heroku‑like tools, while noting that no replacement has fully matched Heroku’s original “git push to deploy” simplicity.

Interpretation of the announcement

  • “Sustaining engineering model” is widely read as “maintenance mode” / “keep the lights on”: bug fixes and operations, no meaningful new features.
  • Several commenters say this is how Heroku has effectively operated for years; the post just makes it official.
  • The language is criticized as vague “corporate nullspeech” that obscures the real message instead of clearly stating “no more new features, long tail support.”
  • Stopping new Enterprise contracts is seen as the concrete signal of a slow-motion sunset and the first step in a formal end‑of‑life process.

Customer reactions and risk assessment

  • Many argue any serious customer should immediately plan to migrate; staying now means accepting future problems as self‑inflicted.
  • Others see it as a “mature” acknowledgment of Heroku’s lifecycle: stable, not growing, but not shutting down imminently.
  • Some enterprise users report years of outages, poor support, and culture decay after the Salesforce acquisition; for them this is long overdue confirmation.
  • A few note Salesforce processes typically require explicit “end of sales” announcements as precursors to eventual EOL.

Alternatives and migration paths

  • Hosted PaaS successors frequently mentioned: Render, Railway, Fly.io, Northflank, DigitalOcean App Platform, Vercel/Netlify (for frontends), DO App Platform, plus various niche/European offerings.
  • Self‑hosted “Heroku-like” tools: Coolify, Dokku, Dokploy, CapRover, Frost, Cuber, Kamal, Canine, plus plain Kubernetes or VPS setups (often on Hetzner/DO).
  • Debate over suitability: some insist self‑hosting defeats Heroku’s value prop (“nothing to install”), others say modern self‑hosted UIs replicate the experience well enough.
  • Postgres migration is called out as the hard part; suggestions include Crunchy Bridge and logical replication to minimize downtime.

Pricing, trust, and business dynamics

  • A recurring theme: Heroku became too expensive (“Heroku tax”), especially after free tiers and cheap dynos disappeared; many teams left primarily for cost.
  • Confusing or “shady” billing language and the removal of the free tier eroded trust, especially for MVP/PoC use cases.
  • Some ex‑insiders blame sales incentives: revenue targets could be hit by upselling existing customers instead of winning new ones, removing pressure to innovate.

Heroku’s legacy and technical history

  • Strong nostalgia: the “git push and you’re live” experience is described as magical, career‑making, and still unmatched in polish.
  • Heroku is remembered as a pioneer of developer‑centric UX, buildpacks, and turnkey pipelines that inspired many later platforms.
  • Former employees in the thread describe massive growth after the Salesforce acquisition, followed by crippling tech debt, AWS outages, and a shift from product velocity to reliability and process.
  • Several see this as a classic arc: visionary founding team, rapid growth, tech debt, founders leaving, bureaucracy rising, and eventual stagnation.

Broader industry reflections

  • The situation is framed as part of a larger pattern: big‑company acquisitions (Salesforce, Microsoft, etc.) diluting once‑admired engineering cultures and products.
  • Some argue PaaS itself has been commoditized (containers, “just run my Docker image” on major clouds), making Heroku’s original model less economically compelling.
  • Others counter that no one has truly recreated the original Heroku experience end‑to‑end, which is why so many “new Herokus” keep appearing.