Last checked and updated on 04 May 2026
What Is New in Helm 2.7 (summary table)
| Category | Highlights |
|---|---|
| New Features |
|
| Improvements |
|
| Bug Fixes |
|
| Security |
|
| Deprecated | None reported for this release |
For the full changelog see the official Helm 2.7 release notes.
What are the key change groups in Helm 2.7?
CLI Enhancements
Two new flags landed on helm init. --node-selectors lets you pre‑define node selectors for Tiller pods, while --output lets you choose JSON or YAML output for the init command. In practice this reduces the amount of post‑install scripting you need.
Tiller Runtime Improvements
Support for CRD ordering ensures that custom resource definitions are applied before dependent resources, preventing a common source of deployment failures. The logs now contain clearer wait messages, which helps when troubleshooting long‑running releases.
Release Storage Backend
The beta secrets storage backend stores release information in Kubernetes Secret objects. This gives you an out‑of‑the‑box encrypted store without needing an external database.
Stability & Bug Fixes
Numerous edge‑case bugs were squashed: URL handling no longer appends an extra /index.yaml, semantic version parsing now respects prerelease tags, and several panics in the test suite were eliminated. These fixes make Helm more reliable in CI pipelines.
FAQ
How do I use the new --node-selectors flag?
Pass a label selector string to helm init, e.g.
helm init --node-selectors="role=master"Can I switch the release storage to the secrets backend after a cluster is already running?
Yes. Update the Tiller deployment to set --storage=secret in the command line arguments and restart Tiller. Existing releases remain in the original backend until you migrate them manually.
What output formats are supported by the new --output flag?
The flag accepts json or yaml. The chosen format is applied to the result of helm init, making it easier to consume in scripts.
Does the CRD ordering change affect existing charts?
No. Helm now automatically sorts CRDs before other resources, but chart authors do not need to modify their templates. The change is transparent to the user.
Are there any known limitations of the secrets storage backend?
It is marked as beta, so it lacks some advanced features like versioned secret rotation. It also inherits the size limits of Kubernetes secrets (1 MiB per secret).