The problem Microsoft says it is solving
Microsoft's Azure Arc team frames this around scale. Platform teams may look after hundreds or thousands of clusters across Azure, on-premises datacenters, edge sites and other clouds. That figure describes the size of the challenge. It is not a documented product limit or a benchmark.
Two jobs get harder at that scale:
- Keeping clusters safely up to date.
- Deploying applications and configuration consistently across all of them.
Upgrade Insights targets the first job. Flux with Fleet Manager targets the second.
Upgrade Insights for Azure Arc-enabled Kubernetes
Microsoft describes this as a pre-upgrade readiness check across three areas:
- Deprecated Kubernetes APIs. It identifies workloads and controllers that use APIs that may be deprecated or removed in the target Kubernetes version.
- Component compatibility. It evaluates commonly used components for compatibility with the target version. Microsoft names cert-manager, Istio, Cilium and containerd.
- Upgrade readiness. It flags configuration problems, such as Pod Disruption Budgets (PDBs), that could interfere with node drains and upgrades.
The intended outcome is a central view of which clusters are ready to upgrade and which need remediation first.
It is advisory only
For Arc-enabled Kubernetes, Microsoft says Upgrade Insights is advisory. The upgrade itself stays under the control of the underlying Kubernetes platform. It is not an upgrade engine, it does not remediate anything, and it is not a guarantee that an upgrade will succeed.
Why PDBs and deprecated APIs matter
Microsoft's AKS troubleshooting documentation shows why these checks are worth running. It describes UpgradeFailed errors where a restrictive Pod Disruption Budget stops a node drain. That specific page is about AKS, not Arc, so it supports the operational risk but not how Arc's feature is implemented. The same page says the PDB's ALLOWED DISRUPTIONS value should be 1 or greater. A value of 0 blocks eviction and the drain fails. It also suggests running the portal's Upgrade readiness check on AKS before upgrading.
Anyone who has watched a drain hang on a PDB set to "never disrupt anything" knows how this goes. You can cover part of this by hand today with kubectl get pdb. At fleet scale that doesn't hold up, which is where a central assessment would help.
Related AKS work
Pixel Robots reported that Microsoft Learn has a separate article on an Upgrade Insights API for AKS. It uses az aks upgrade-insights from the aks-preview extension. When the blog was written, those commands were not in any released version. Don't confuse that AKS feature with the Arc one.
What isn't published
Microsoft's announcement does not say:
- which Kubernetes distributions and versions are supported
- how findings are surfaced
- how often assessments run
- how to enable it
- whether every named component is detected in every environment
Flux GitOps with Azure Kubernetes Fleet Manager
Microsoft splits the roles this way:
- Git or OCI is the source of truth.
- Flux reconciles desired state.
- Fleet Manager decides where resources run.
Flux runs on the Fleet Manager hub cluster. It reconciles application manifests and Fleet resources from a Git or OCI source. Fleet Manager then evaluates placement policies and distributes resources to member clusters. Application definitions therefore don't have to list every destination cluster.
What the documentation adds
Microsoft Learn says the Fleet Manager model follows the same Flux v2 extension pattern used by AKS and Azure Arc-enabled Kubernetes. The microsoft.flux extension installs these components by default:
- a source controller, which watches
GitRepository,OCIRepositoryandHelmRepositoryobjects - a Kustomize controller
- a Helm controller
- a notification controller
The documentation also states some limits and design points:
- The hub stores, it doesn't run. By design, the hub cluster doesn't run synchronized resources locally. It only stores resource definitions for distribution to member clusters.
- Portal and ARM configuration is missing. ARM and Azure portal configuration for the
fluxconfig-agentandfluxconfig-controllercomponents is not currently available for Fleet Manager. - OCI sources are supported. An
OCIRepositorycan point at an artifact in a container registry such as Azure Container Registry. AKustomizationthen references it. Microsoft's example uses a tagged artifact and a registry credentials secret. - Documented use cases. These include regional rollouts, staged releases across dev, test and production, environment-specific overlays, and shared services.
Alongside the announcement, Windows Forum coverage of the Arc Ignite updates says Fleet Manager with Flux lets platform teams push applications and configurations to AKS and Arc-enabled clusters across Azure, on-premises, AWS and Google Cloud. The same coverage says Arc member clusters are currently supported only in Azure public cloud regions. I could not confirm that regional limit in Microsoft's own documentation in the material I reviewed. Check it before planning around it.
The preview caveat
Microsoft's Fleet Manager GitOps documentation is blunt about preview status. Preview features are opt-in and self-service. They are provided "as is" and "as available", and they are excluded from SLAs and the limited warranty. Support is best-effort and partial. The page says these features aren't meant for production use.
That warning is on the Fleet Manager GitOps page. I found no evidence that it applies identically to Upgrade Insights. Still, Microsoft labels both features as public preview, so treat both as pilot material until you have tested them.
What this means for admins
The two features cover separate layers of fleet operations:
| Question | Capability | What it does not do |
|---|---|---|
| Is this cluster ready to upgrade? | Upgrade Insights (Arc-enabled Kubernetes) | Run the upgrade or fix issues |
| What should run where, and how do we keep it reconciled? | Flux + Fleet Manager | Judge upgrade readiness |
Microsoft's closing pitch treats them as a pair. As described, though, they are two previews, not one integrated workflow.
Practical steps
- Inventory which clusters are Arc-enabled and which are AKS members of a fleet. The two features apply to different cluster populations.
- Audit PDBs and deprecated API usage now with standard tooling. That keeps you from depending on a preview feature for basic hygiene.
- Trial Flux with Fleet Manager in a non-production fleet. Check that your Git or OCI source layout, namespaces and RBAC suit the hub-stores, members-run model.
- Ask Microsoft, or watch its documentation, for supported distributions and versions, pricing, and a general-availability timeline. The announcement doesn't give them.
Bottom line
This is a sensible pairing of upgrade visibility and GitOps placement for teams that run many clusters. The announcement offers no measured time savings, pricing or GA dates, and the preview terms rule out production use. The GitOps architecture is documented in useful detail. Upgrade Insights is, for now, mostly a promise with a short feature list.
References
- Simplify Kubernetes operations with Upgrade Insights and GitOps with Azure Kubernetes Fleet Manager Azure Arc Blog · 2026-11-17T18:45:00+00:00
- Simplify Kubernetes operations with Upgrade Insights and GitOps with Azure Kubernetes Fleet Manager | Microsoft Community Hub techcommunity.microsoft.com
- GitOps-based deployments with Azure Kubernetes Fleet Manager and Flux | Microsoft Learn learn.microsoft.com