## TL;DR
A release with no deployed releases failed during its first install and never reached a good state. You cannot upgrade it (there is nothing to upgrade to) and rollback has nothing to roll back to. The recovery is: diagnose the install failure from the release history, fix the underlying cause, then either uninstall and reinstall cleanly or force the install through with the fix.

## The query
```text
helm release "has no deployed releases": how to recover a failed install
```

## Use this when
- helm upgrade fails with "has no deployed releases"
- A release is stuck in failed or pending-install state
- Deciding between uninstall/reinstall vs repair
- First installs that never succeeded

## Not for when
- Chart template rendering errors (fix the chart first)
- Upgrades between working releases
- Helm repository or chart download issues

## Steps

### Step 1: Read the release history
List the release history to see the failed revisions and their statuses. The history shows what Helm attempted and where it stopped; pending-install and failed are the states you will see.
Expected output: the failed revision identified with its status.

### Step 2: Find why the install failed
Check the failed release's resources: pods that never became ready, hooks that failed, timeouts. The install failed for a concrete reason (bad image, failing probe, insufficient resources); the "no deployed releases" message is just the aftermath.
Expected output: the root cause of the original install failure.

### Step 3: Fix the underlying cause first
Fix whatever broke the install: the chart values, the image, the resource constraints. Reinstalling without fixing repeats the failure and adds another failed revision to the history.
Expected output: the cause fixed and verified (e.g. the image pulls, the probe passes).

### Step 4: Choose: uninstall then install, or repair in place
For a first install that never worked, uninstall the failed release and install fresh; it is the cleanest path. Only attempt repair-in-place if uninstall would destroy state you need (rare for a never-deployed release).
Expected output: a clean release history starting from a successful install.

### Step 5: Set install timeouts and atomicity for next time
Use atomic installs (automatic rollback on failure) or at least sane timeouts so future failures leave the release in a recoverable state instead of limbo.
Expected output: future failed installs self-clean instead of stranding the release.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_GTmrf4-OOsDGY5wQILGFAg