Today we’re excited to share that Quick Attach Detach (QAD), a new Azure platform capability for accelerating managed disk attach and detach operations, is coming to Azure Kubernetes Service (AKS) and Azure Container Instances (ACI). QAD is currently in private preview, with broader availability planned as we validate performance, security, and reliability at scale.
Stateful workloads shouldn’t wait on storage
As organizations move more stateful, data-intensive, and AI-driven workloads onto Kubernetes and serverless containers, the operational reality looks like this:
- Node and pod recovery — StatefulSets need their Persistent Volumes back quickly when a pod is rescheduled after a failure.
- Planned maintenance and upgrades — node image upgrades, cluster upgrades, and scale operations all involve moving storage along with (or ahead of) compute.
- AI sandbox and agent restore — agentic and AI workloads increasingly need to resume from a previously persisted state rather than rebuilding their environment from scratch every time compute changes underneath them.
Traditional managed disk attach/detach relies on a top-down workflow, but not tuned for orchestrators that need to move disks frequently and at fleet scale. That round trip can add meaningful latency to pod startup, failover, and recovery, exactly when customers notice it most.
What is Quick Attach Detach?
QAD introduces a faster, node-driven storage transition path for approved workload orchestrators, sitting alongside — not replacing — Azure’s existing managed disk attach/detach model. Every QAD operation still goes through full Azure validation, including:
- Identity and authorization
- Azure RBAC permissions
- Disk ownership and eligibility
- Subscription boundaries
- Platform policy and VM disk limits
- Storage governance controls
In other words: it’s a faster execution path, not a bypass of Azure security. QAD requests are designed to be idempotent, with per-disk sequencing to protect against race conditions when multiple compute instances contend for the same disk — important for orchestrators managing thousands of attach/detach operations across a large fleet.
Why this matters for AKS
For AKS, QAD integrates with the Azure Disk CSI Driver, so nothing changes in how developers think about storage. You keep using the Kubernetes primitives you already know:
- StorageClasses
- Persistent Volumes and Persistent Volume Claims
- StatefulSets
- Standard pod scheduling and rescheduling
What changes is what happens underneath that experience. The node-driven attach/detach path is designed to cut the time a pod spends waiting on storage when it’s created, rescheduled, or recovered on a different node — directly improving the experience for scenarios like StatefulSet recovery, node image upgrades, cluster upgrades, planned maintenance, and scale events.
Why this matters for ACI
For Azure Container Instances, QAD is central to a new stateful restore capability for agentic and AI sandbox workloads. Instead of rebuilding an environment from zero, a container can resume from previously persisted state:
- A stateful AI sandbox or application is running with data on durable Azure storage.
- The original container is replaced or removed.
- ACI requests the stateful environment on the replacement compute.
- QAD reassigns and attaches the persistent disk to the new environment.
- The application continues from where it left off — no rehydration, no rebuild.
This is the same “compute is disposable, state persists” principle, applied to the fast-moving world of container-based AI workloads.
Performance: built for speed, validated for scale
QAD attach operations give your VM the disk within seconds.
Just as important as raw speed is what QAD does not change: QAD does not give up any of the identity, ownership, or governance guarantees they already rely on from Azure managed disks.
Security and governance, unchanged
QAD does not introduce a new trust boundary. Every accelerated operation is still authorized end-to-end — the platform confirms the calling identity has permission to use the disk, validates ownership and resource boundaries, and authorizes storage access before the operation completes. The host only ever receives what it needs to complete an already-authorized request.
What’s next
QAD for AKS and ACI is currently in limited preview. Over the coming months, the team is focused on:
- Decreasing time to attach and detach disks
- Broadening regional and storage-tenant compatibility
- Bringing this capability to all VM scenarios.
If you’re running stateful workloads on AKS or ACI and want to be part of the preview, or if you are interested in trying this for other scenarios, reach out to QAD-Vteam@microsoft.com for more information.
Quick Attach Detach is part of Azure’s ongoing investment in making infrastructure keep pace with increasingly dynamic, AI-driven workloads — so that when your compute moves, your data doesn’t have to wait to catch up.