Go Live Plan
This document is the practical launch checklist for Harbor.
GO_LIVE_PLAN.md
Purpose
This document is the practical launch checklist for Harbor.
Use it to answer:
- what stage Harbor is in now
- what must be true before design-partner beta
- what must be true before public v1
- what is intentionally allowed to remain manual at launch
- what is explicitly post-launch work
This file should stay more concrete than the roadmap.
Use docs/BREAKWATERHARBOR_NET_DEPLOYMENT.md for the concrete first deployment target and domain layout.
Current state
As of the current repo state, Harbor is best described as:
- strong local-first alpha
- design-partner beta candidate
- not yet ready for broad public production launch
Why it is strong
- Harbor Node is real and useful
- Harbor UI is modular and coherent
- Harbor Hub, website, docs, cloud member basics, and cloud admin are real
- central management, multi-node foundations, and response pipelines are already in the product
- the product has meaningful automated coverage in its highest-risk areas
Why it is not yet public-launch ready
- browser-facing auth/session posture still needs improvement
- CI and regression coverage are not deep enough for broad public change velocity
- deployment, backup, restore, and rollback discipline need to be formalized
- operator onboarding, policy visibility, and troubleshooting still need a final pass
- at least one OAuth-backed integration needs to be clearly production-grade
- the public website still needs its under-construction landing page and gated-preview shape
Launch philosophy
Harbor should go live in stages:
- internal alpha
- design-partner beta
- public v1
- post-launch expansion
This is the right shape for a local-first product. It keeps the trust model intact while allowing support, docs, and operations to mature alongside real usage.
Current prelaunch target
The current real deployment target is:
- domain:
breakwaterharbor.net - VPS:
162.222.206.87 - public reverse proxy: Caddy
- source host: internal Gitea over Tailscale / MagicDNS
The current recommended topology is:
- main site at
breakwaterharbor.net - docs at
docs.breakwaterharbor.net - Harbor Hub at
hub.breakwaterharbor.net - cloud API at
api.breakwaterharbor.net - Harbor Admin at
admin.breakwaterharbor.net, Tailscale-only
Harbor Node remains local-first and should not be treated as a public VPS-hosted default runtime.
Stage 1: Internal alpha
Goal
Keep building and hardening with the repo as the primary source of truth.
Success signals
- core local workflows build and run reliably
- major surfaces are real and navigable
- the most important security regressions are covered by tests
- design and architecture are stable enough that docs can track reality
Already true now
- yes for Harbor Node, Harbor UI, Harbor Hub, website, docs, cloud API, and cloud admin
Stage 2: Design-partner beta
Goal
Support a small set of real users with close human involvement and fast feedback loops.
Must be true before entering this stage
Security and auth
- cloud/admin browser sessions are moved to a stronger production-grade pattern
- Harbor Node production environment requirements are documented clearly
- secret-at-rest expectations are documented and enforced
CI and quality
- CI runs on the pinned Node toolchain
- CI runs package tests in addition to typecheck/build
- at least the highest-risk API surfaces have automated regression coverage
Product and UX
- Harbor UI onboarding is good enough for a new operator to get through install and first action publish
- Harbor Guard policy/approval behavior is visible enough to avoid confusion
- Harbor Hub import/update state is understandable
- docs cover install, upgrade, and troubleshooting for the supported environments
- the website matches the intended landing-page plus gated-preview deployment model
Ops and support
- there is a documented install flow
- there is a documented upgrade flow
- there is a documented backup and restore flow
- internal support knows how to troubleshoot account, enrollment, and contact flows
What may still be manual in this stage
- premium provisioning
- license operations
- design-partner onboarding
- some support escalations
- billing and subscription operations
Exit criteria
- a small set of real users can run Harbor without unsafe shortcuts
- Breakwater can support those users without relying on raw database edits or secret exposure
Stage 3: Public v1
Goal
Launch Harbor publicly with a narrow, honest, supportable promise.
Must be true before public v1
Security and auth
- cloud/admin auth uses a production-appropriate browser session model
- CSP, framing, referrer, and TLS guidance are documented and implemented where applicable
- the known high-severity audit findings are fixed or explicitly accepted with written rationale
Product quality
- Community install path is stable and understandable
- premium story is additive and clear
- at least one OAuth-backed integration is fully supportable
- manager and multi-node features are either clearly supported or clearly labeled as not public-v1-ready
Documentation
- website copy matches the actual shipped product
- docs include install, upgrade, restore, troubleshooting, and security guidance
- plan and premium messaging are accurate and not aspirational
Release engineering
- CI runs tests plus typecheck/build
- container builds are repeatable
- release notes and upgrade notes can be generated consistently
- rollback steps are documented and tested
Operations
- Harbor Node SQLite - Hub SQLite - Cloud Postgres
- monitoring and health expectations are documented
- backup and restore have been exercised for:
- contact/support flow works end to end
- admin tooling is sufficient for the first public users
What can still be manual at public v1
- premium provisioning and licensing operations
- some support escalations
- some recovery operations if they are documented and safe
- limited white-glove onboarding for early paid users
What should not still be manual at public v1
- basic install steps
- basic upgrade steps
- basic backup/restore process
- critical auth/session handling
- core Community product usage
Exit criteria
- Harbor can be presented publicly without hiding material operational gaps
- a user can install, run, and understand the Community product
- Breakwater can support the first public users reliably
Stage 4: Post-launch expansion
Goal
Deepen premium and commercial operations only after the public product is stable.
Good post-launch targets
- stronger Pro/Business multi-node and manager ergonomics
- better premium notifications and summaries
- deeper response pipeline and automation tooling
- team and organization depth
- self-serve billing and subscription management
- broader integration set
Explicitly post-launch
- large connector expansion
- marketplace economics
- broad enterprise RBAC
- full Fleet orchestration as the default Harbor mode
Go-live workstreams
1. Security and auth
Must ship:
- stronger cloud/admin browser session model
- documented production secrets and env requirements
- deployment header and TLS guidance
- final review of remaining high-severity issues
2. Product polish
Must ship:
- Harbor UI onboarding
- policy visibility and approval clarity
- Harbor Hub import/update clarity
- auth-readiness and troubleshooting cues
- public landing page plus gated-preview website behavior
3. Release engineering
Must ship:
- CI on pinned Node version
- test execution in CI
- repeatable container builds
- release and rollback notes
4. Ops and recovery
Must ship:
- health and monitoring expectations
- backup and restore procedures
- upgrade procedures
- disaster-recovery basics for the supported deployment model
5. Documentation and support
Must ship:
- docs aligned with real product behavior
- install/get-started path that does not require source reading
- support/admin workflows documented for the first customers
6. Commercial readiness
Must ship:
- clear Community versus premium packaging
- manual premium operations that are documented and supportable
- honest go-to-market promise about what Harbor does today
Suggested sequence from here
- finish production auth/session hardening
- upgrade CI and expand regression coverage
- finish deployment, backup, restore, and rollback documentation
- finish Harbor UI onboarding, policy visibility, and Harbor Hub clarity
- finish Gmail as the flagship OAuth-backed integration
- run design-partner beta
- fix beta feedback and close launch blockers
- publish public v1
Launch decision checklist
Before saying "go live," confirm all of these are true:
- security posture is acceptable for production use
- docs match reality
- CI and tests are strong enough for the current change rate
- backups and restores have been exercised
- install and upgrade path are stable
- first public support workflow is ready
- Community path stands on its own
- premium path is additive, not required for trust
If any of those are still weak, Harbor should remain in design-partner beta rather than forcing a public launch.