Release is stuck in pending-upgrade
An upgrade was interrupted by a CI timeout or a disconnected kubectl, leaving the release half-applied. Every subsequent attempt bails out with the same message.
Release unstuck, history intact
Running Helm?
Helm starts tidy and ends with fourteen values files per environment, a chart that imports three subcharts you cannot find the source for, and a release wedged in pending-upgrade since last quarter. We bring the chart tree back to something a new engineer can read in an afternoon, and we unstick the releases without resetting history.
What it is
Helm is the package manager for Kubernetes: it bundles related manifests into a chart, substitutes values into templates, and tracks what it has installed as a named release in the cluster. The project started at Deis in 2016 and now lives under the Cloud Native Computing Foundation. Helm 3, released in 2019, removed the server-side Tiller component and pushed everything into the client.
A chart is a directory with templates, a Chart.yaml metadata file, a default values.yaml, and optionally subcharts under charts/. Running helm install or helm upgrade renders the templates with the supplied values and applies the result to the cluster, while Helm records a release secret in the target namespace so it can roll back later. The dominant workflows today are either plain Helm from a CI job, Argo CD with Helm as the templating engine, or Flux using HelmReleases as custom resources.
The honest version
No technology is all upside. This is where Helm earns its place and where it tends to cause trouble.
Sound familiar?
The Helm problems that bring people to us, and what you have once each one is fixed.
An upgrade was interrupted by a CI timeout or a disconnected kubectl, leaving the release half-applied. Every subsequent attempt bails out with the same message.
Release unstuck, history intact
A nested value in a subchart that the parent never set, or a required field that was renamed in a later chart version. The render dies and takes the deploy with it.
Templates render on every environment
A selector or volume claim name that changed in the chart. Kubernetes rejects the patch and Helm has no built-in path to delete and recreate without data loss.
Resource migrated without downtime
Release history was trimmed below the point you wanted to roll back to, or the last successful revision was marked failed by a liveness probe. There is nowhere to roll back to.
History and probes aligned with real health
A template that produced unindented keys because a conditional branch dropped a leading space, or a multiline string that was not quoted. The chart renders but Kubernetes rejects the output.
Templates produce valid YAML every time
A deployment whose readiness probe never passes, often because of a misconfigured health check path, or a pod stuck on an image pull. helm install --wait eventually times out with no clear reason.
Readiness probes that mean what they say
A Chart.yaml edited without rerunning helm dependency update, so CI pulls one version and local renders another. Behaviour differs between environments for no obvious reason.
Lockfile and chart in agreement
An environment values file overriding a map with a scalar, or a list that gets replaced when the engineer expected it to be extended. Helm merges values with rules that are not always intuitive.
Values merge to a predictable shape
What we do
Come to us for one wedged release, or hand us the whole chart repository.
How it works
Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.
We fix the cause, not the symptom. If something is down, getting it back up comes first.
Backups, monitoring, updates and access, so the same problem can't quietly come back.
You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.
Often found together
We work on the whole stack, so the problem can be wherever it actually is.
A previous helm upgrade was interrupted mid-run, usually because a CI job timed out or a kubectl session was disconnected. The release state in the cluster is left marked pending, and subsequent upgrades refuse to proceed. The fix is to either roll back to the last deployed revision or, when that is not available, to patch the release secret directly, which is best done with a backup and a careful plan.
Helm earns its keep when you consume off-the-shelf charts or need runtime value substitution. Kustomize earns its keep when the manifests are mostly static with overlays. In practice many teams run both: Helm for third-party software, Kustomize for their own services. Picking one for everything is rarely worth the fight.
Kubernetes blocks the patch, so the migration has to delete the resource and recreate it. For stateless services that is fine with a quick rolling window. For StatefulSets, PVCs and anything with persistent data, the plan has to preserve the data, usually by snapshotting the volume, deleting the resource, and reattaching the volume to the recreated one.
Yes, Helm has supported OCI registries as a first-class chart store for several releases. Pushing to ECR, GHCR or Harbor avoids running a separate chart museum and reuses the IAM and signing you already have for images. The one catch is that older tooling may still expect an index.yaml and will not find it.
Start with a base values.yaml that carries the real defaults, and keep environment files as small as possible, ideally just the parts that actually differ. A schema file (values.schema.json) catches typos and shape drift at render time, which is where most of the sprawl pain actually bites.
An estimate is free. Describe the chart, the error and the cluster target, and we return a scoped quote before any changes are drafted.
Free estimate
Tell us what you're running and what's going wrong. A person reads it and comes back with an estimate.
Your message is with a real person, and we'll get back to you soon.
Rather talk now? Call us: