atmos helm apply

Install or upgrade a Helm release (helm upgrade --install) for a component in a stack, using values resolved from your stack configuration. With --target, apply instead delivers the rendered manifests to a provision target such as a Git deployment repository.

Usage

atmos helm apply <component> --stack <stack> [options]

Install or upgrade the release in the cluster:

atmos helm apply monitoring -s plat-ue2-dev

Apply a one-time image override without changing stack configuration:

atmos helm apply monitoring -s plat-ue2-dev -f incident-values.yaml --set image.tag=2026.09.09

Upgrade an existing release with an explicit production readiness and recovery policy:

atmos helm apply monitoring -s plat-ue2-dev \
  --on-failure=rollback \
  --cleanup-on-failure \
  --wait=watcher \
  --wait-for-jobs \
  --timeout=30m \
  --history-max=10

Deliver rendered manifests to a provision target instead of the cluster:

atmos helm apply monitoring -s plat-ue2-dev --target deployment-repo

Apply all or affected Helm components in dependency order:

atmos helm apply --all -s plat-ue2-dev
atmos helm apply --affected --base origin/main

Apply components filtered by tags or labels (composes with --all/--affected to narrow the selected set further):

atmos helm apply --all --tags production,tier-1
atmos helm apply --affected --labels cost-center=platform

# Repeat --labels to combine multiple selectors across occurrences
atmos helm apply --affected --labels cost-center=platform --labels compliance=sox

In CI, bulk applies selected with --all or --affected write one aggregate GitHub job summary after the dependency graph completes. It lists component status and Helm release metadata in stable stack/component order and includes details for failures.

While a cluster apply is running, Atmos reports the selected install or upgrade operation, effective wait policy, timeout, and elapsed-time heartbeats. Bulk output includes the component and stack on every line so concurrent or dependency-ordered operations remain distinguishable. Progress is written through the masked UI stream; chart hook container logs are not streamed automatically.

Flags

--stack, -s (required): Atmos stack.
--target (optional): Provision target to deliver to (e.g. a Git deployment repository). Defaults to provision.default, otherwise the cluster.
--dependency-update (optional): Fetch declared chart dependencies when they are missing. This may access dependency repositories and update the chart's charts/ directory and lock file.
-f / --values, --set, --set-string, --set-file, --set-json, --set-literal (optional): Repeatable Helm-compatible value overrides for this apply. They override stack-resolved values only for this invocation. See runtime value overrides.
--dry-run (optional): Preview the install or upgrade without persisting release state or creating Kubernetes resources.
--on-failure (optional): Failure action for the selected operation: uninstall or keep for install; rollback or keep for upgrade. The explicit flag overrides stack configuration for this invocation.
--cleanup-on-failure (optional): Remove resources newly created during a failed upgrade, independently of rollback. Fails if the selected operation is install.
--wait[=strategy] (optional): Use watcher, hookOnly, or legacy. Passing --wait without a value selects watcher. Boolean values remain accepted temporarily but are deprecated.
--wait-for-jobs (optional): Wait for ordinary Jobs in the release manifest. Requires watcher or legacy.
--timeout (optional): Helm release-operation timeout, such as 10m or 1h. 0s is explicitly unbounded.
--history-max (optional): Maximum retained release revisions for an upgrade. Defaults to 10; 0 means unlimited. Fails if the selected operation is install.
--no-hooks (optional): Disable Helm chart hooks. Atmos lifecycle hooks are unaffected.
--skip-crds (optional): Skip CRD installation on a first install. Fails if the selected operation is upgrade.
--all (optional): Apply all Helm components in dependency order.
--affected (optional): Apply affected Helm components and their dependencies.
--include-dependents (optional): With --affected, include dependent Helm components.
--tags (optional): Filter by tags (comma-separated, matches any): --tags=production,tier-1. Composes with --all/--affected to narrow the selected set further; cannot be combined with a single component argument.
--labels (optional): Filter by labels (comma-separated key=value or key:value pairs within one occurrence, and/or repeated for more, matches all): --labels=cost-center=platform,compliance=sox or --labels cost-center=platform --labels compliance=sox. Composes with --all/--affected/--tags; cannot be combined with a single component argument.

Lifecycle flags apply only to direct Kubernetes delivery. Combining an explicit lifecycle flag with an external --target fails instead of implying that Atmos waited for or rolled back a GitOps deployment.

Usage​

Flags​

Usage

Flags