Running Flux CD?

Flux CD HelmReleases or Kustomizations stuck? We ship it properly.

Flux is quietly powerful and does not tell you when it has given up. A HelmRelease that paused mid upgrade, a Kustomization that drifted away from Git because somebody kubectl applied a hotfix, or a GitRepository that cannot pull a private repo all sit in a status field and age. We walk the reconciliation tree and get the cluster back in line with the repo.

  • Since 2015Fixing technology
  • FreeEstimates
  • US teamReal engineers

flux-events.log

still broken

What it is

Flux CD at a glance

Flux CD is a GitOps toolkit for Kubernetes that reconciles clusters against sources held in Git or OCI registries. The first version came out of the Weaveworks team in 2016, and Flux v2 with its toolkit style controllers reached general availability in 2022, now hosted at the CNCF as a graduated project.

The design splits responsibilities across several controllers: source-controller fetches from Git, OCI or Helm repositories, kustomize-controller applies Kustomizations, helm-controller reconciles HelmReleases, and notification-controller handles alerts and webhooks. That makes Flux composable but also means a failed reconciliation lives inside one of those controllers, and reading the status chain is a skill of its own.

Type
GitOps toolkit for Kubernetes
First released
2016 as Flux v1, Flux v2 generally available in 2022
License
Apache 2.0, open source
Maintained by
The Flux community under the CNCF
Typically used for
Kubernetes GitOps, Helm chart delivery, multi tenant platforms, pull based cluster configuration

The honest version

Flux CD pros and cons

No technology is all upside. This is where Flux CD earns its place and where it tends to cause trouble.

Where Flux CD is strong

  • Pull based model fits locked down clustersThe cluster pulls from Git rather than CI pushing credentials inward, which lines up well with air gapped and regulated environments.
  • Composable controllersGitRepository, Kustomization, HelmRelease and ImagePolicy are separate resources, so you can wire exactly the pipeline you need without swallowing a platform.
  • Native OCI supportFlux can treat an OCI registry as a source, which lets teams sign artifacts with cosign and ship manifests the same way they ship images.
  • Image automation built inImage reflector and image automation controllers watch registries and open pull requests to bump tags, which keeps the GitOps loop honest.
  • Multi tenant friendlyFlux supports tenant scoped reconciliation, service account impersonation and per tenant Git sources, so a platform team can hand each team their own slice.
  • Steady upstream cadenceReleases arrive regularly, migration notes are written for operators rather than marketers, and the community on Slack is responsive.

Where Flux CD bites

  • Status is quiet when things go wrongA failed Kustomization or a stuck HelmRelease shows up as a status field with a short reason. Without notifications wired, the cluster can drift from Git for days before anyone notices.
  • HelmRelease state machine is fragileWhen a Helm release ends up in a failed or pending-upgrade state, Flux needs manual intervention or specific annotations to roll forward. Teams find this out at 2am.
  • Drift from manual kubectl appliesFlux reconciles toward Git, but it does not stop an engineer running kubectl edit or kubectl apply from pushing the cluster away from the repo between reconciliation windows.
  • No built in UIFlux is deliberately CLI and Kubernetes native, with no server side dashboard the way Argo CD ships one. Teams that want a GUI have to layer Weave GitOps or build their own.
  • Learning curve on multi clusterFleet wide Flux with tenants, encryption, image automation and policy takes real design time. The quick start works; the production shape needs thinking.
  • Policy drift between toolsTeams that run Argo CD on some clusters and Flux on others end up maintaining two sets of reconciliation patterns, two sets of status conventions and two notification paths.

Sound familiar?

Common Flux CD problems we fix

The Flux CD problems that bring people to us, and what you have once each one is fixed.

HelmRelease upgrade retries exhausted

A Helm release is stuck in failed state

A chart upgrade that hit a template error or a timed out hook, leaving the HelmRelease unable to progress without manual intervention or a targeted annotation.

Releases roll forward cleanly

GitRepository authentication required

Flux cannot pull from the Git source

A deploy key that was removed, a token that expired, or a secret that was renamed without updating the GitRepository resource.

Sources sync, status is green

drift detected: configmap changed outside Git

The cluster drifted from the repo

A kubectl edit or an emergency hotfix that was applied directly to the cluster without being committed back to Git, leaving Flux to reconcile it away at the next pass.

Cluster and Git back in step

field is immutable

A Kustomization cannot apply a change

An update to a field Kubernetes does not allow to change on a live object, usually a selector or a StorageClass, so apply fails every cycle.

Changes land without conflict

source-controller OOMKilled

The source controller runs out of memory

A large monorepo or a Helm repo index that exceeds the controller's memory, causing reconciliation to loop on OOM kills.

Controller sized for the sources

cosign no matching signatures

OCI signature verification keeps failing

A signing key that rotated without updating the public key config, or an image pushed without the signature step in CI.

Signed artifacts verify correctly

pruned resource still exists

Garbage collection cannot remove a resource

A resource whose finalizer is waiting on an external controller, blocking Kustomization pruning and stalling the next reconciliation.

Pruning completes on schedule

ImagePolicy: no matching tag

Image automation stopped bumping tags

A regex filter on the ImagePolicy that no longer matches the tags being pushed, so automation looks healthy but never updates.

Tag policy matches real releases

What we do

Flux CD help, from stuck releases to drift control

Bring us one failing Kustomization, or hand us the GitOps platform for an end to end review.

Rescue
A HelmRelease pinned in failed, a GitRepository that cannot auth, or a controller OOMing in a loop. We stabilize the reconciliation surface before touching anything else.
Repository layout
Mono repo versus per cluster, overlays, tenant boundaries and secret management designed to fit how your team actually ships.
Helm release reliability
Values, timeouts, remediation and retry policies set so upgrades finish or roll back cleanly rather than parking in failed.
Drift detection
Notifications wired so a manual kubectl change gets noticed in minutes, not days, with a path to bring it back under Git.
Argo to Flux migration
Teams moving from Argo CD to Flux or running both, mapped onto one set of conventions so operators do not maintain two mental models.
Managed Tech
Reconciliation monitoring, upstream upgrade testing, and an engineer available when a controller error goes from quiet to loud.

How it works

From first message to fixed

  1. Look Free estimate

    Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.

  2. Fix

    We fix the cause, not the symptom. If something is down, getting it back up comes first.

  3. Harden

    Backups, monitoring, updates and access, so the same problem can't quietly come back.

  4. Hand back or stay on

    You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.

Often found together

Running Flux CD alongside something else?

We work on the whole stack, so the problem can be wherever it actually is.

Questions

Flux CD questions

Something else? Call+1 737 727 7327

Why is our HelmRelease stuck in failed?

Flux's helm-controller relies on the Helm release state machine, and once a release lands in failed or pending-upgrade, Flux needs either a manual rollback through helm CLI or the right annotations on the HelmRelease to let it recover. The usual cause is a chart upgrade that hit a template error, a hook that timed out, or a webhook that rejected a resource. Fixing the template is step one, clearing the release state is step two.

How do we stop cluster drift from manual kubectl applies?

Flux reconciles toward Git but does not block a human with cluster admin from running kubectl apply. The practical controls are to lock down admin access to break glass accounts, turn on Flux notifications so drift gets flagged the moment it happens, and treat any manual change as a commit that still has to land in the repo within the shift. Policy as code with Kyverno or OPA raises the floor further.

Should we pick Flux or Argo CD?

Both are mature and conformant GitOps tools. Flux fits teams that prefer Kubernetes native resources, OCI artifacts, image automation and minimal UI surface. Argo CD fits teams that want a strong dashboard, application sets and a GUI for developers. We pick based on what the team already uses, how many clusters the platform manages, and whether a UI is a hard requirement.

Why is our source-controller running out of memory?

The source controller fetches and caches tarballs of Git repositories and Helm chart indexes. Large monorepos, repos with a lot of binary assets, or very long chart indexes push it past its default memory. Raising the controller resource limits is a short term fix; trimming the source paths, moving large assets out of the repo and limiting chart repo scope is the durable one.

How do we migrate from Flux v1 to Flux v2 safely?

Flux v1 and v2 are architecturally different, so the migration is not an upgrade in place. The pattern is to install Flux v2 alongside v1 on a copy cluster, port the sync spec into Kustomizations and HelmReleases, verify behavior, and only then cut over by removing v1. We do it one tenant or one cluster at a time.

How much does Flux CD help cost?

Estimates come free. Describe the cluster count and where Flux is stalling, and you get a price before any resources are touched.

Free estimate

Flux CD not keeping clusters in step?

Share the controller logs or a sketch of the repo layout, and a GitOps engineer replies with a plan and a quote.

  • No charge to find out what it will take
  • If it's urgent, say so

Replies come from a person. By sending, you authorize us to email you back. No marketing.