AutoScaler FAQ

Overview

What is Datafy AutoScaler?

Datafy AutoScaler is a cloud storage management solution that automatically grows and shrinks your EBS volumes based on actual usage. It virtualizes the storage layer, presenting a large virtual volume to your applications while managing the underlying EBS volumes in the background to ensure optimal utilization. The result is reduced EBS costs without any changes to your applications or workflows.

For more information, check out how AutoScaler works.

What is the difference between Datafy Sensor and Datafy AutoScaler?

Datafy Sensor is a read-only monitoring agent that tracks and reports on your EBS volume utilization. It does not modify your volumes.

Datafy AutoScaler includes all of Sensor's monitoring capabilities, and additionally can manage your volumes directly - automatically growing and shrinking them to match your actual usage.

How does AutoScaler save me money?

EBS volumes are often over-provisioned to avoid running out of space. Datafy eliminates this waste by continuously right-sizing your volumes based on real-time usage. When volumes are underutilized, Datafy shrinks them. When they need more space, Datafy grows them. You only pay for the capacity you're actually using.

Getting Started

How do I install AutoScaler? How long will it take?

Installing Datafy AutoScaler can be done with a single command and takes less than 10 minutes. To get started, follow the steps in our installation guide.

Will installing AutoScaler cause any downtime?

No. Installing Datafy AutoScaler does not require any downtime. Your volumes remain fully accessible throughout the installation process.

Can I upgrade from Sensor to AutoScaler?

Yes. If you already have Datafy Sensor installed, you can upgrade to AutoScaler without reinstalling. Follow the steps in our upgrade guide.

Where can I install Datafy AutoScaler? Which instances and volume types are supported?

Datafy AutoScaler can be installed on EC2 and EKS instances running a supported Linux operating system. See the full list of Supported Infrastructure, which details supported OSs, filesystems, instance types, and volume types.

If AutoScaler is installed on an instance that has unsupported volumes, utilization will still be reported for these volumes, but you won't be able to activate autoscaling for these volumes.

Why aren't some of my volumes eligible for autoscaling?

A volume may not be eligible for autoscaling if its instance type, operating system, filesystem, or volume type is not currently supported. Check the Supported Infrastructure page to confirm your configuration is covered. If your infrastructure matches the supported list and volumes still aren't eligible, contact Datafy support.

Do I have to install AutoScaler on my whole account? What about my whole cluster?

The AutoScaler installation is controlled completely by you - you can install it in on any relevant sub-set of your environment.

On Kubernetes clusters AutoScaler is installed as a daemonset, and should be installed on the entire cluster whenever possible. If you wish to install on part of your cluster, contact Datafy support.

How do I activate autoscaling for the first time?

After installing AutoScaler, your volumes are automatically discovered and visible in the Datafy app. From there, you can enable autoscaling on individual volumes with the toggle in the AutoScale column, or on groups of volumes that match certain conditions with autoscaling rules.

How It Works

What happens to my volumes when autoscaling is activated? How long does it take?

When autoscaling is activated for a volume, Datafy AutoScaler creates new, smaller volumes and copies the data from your volume to the new volumes. When the copy is complete, the original volume is removed, and you start saving! From here on, AutoScaler will automatically grow or shrink the underlying volumes to match the size of the data actually on the volume.

The volume and the data on it remain accessible for the entire time. The initial activation runs in the background and is designed to have minimal impact on your workload. The duration depends on the amount of data on the volume and can range from minutes to several hours for larger volumes. For more details, see how autoscaling works.

Does autoscaling my volume cause any downtime?

No. Autoscaling volumes does not cause any downtime. Your volumes remain fully accessible throughout the initial activation and any following growing and shrinking actions.

What happens when a volume grows?

When an autoscaling volume's usage increases and it begins to fill up, AutoScaler automatically adds capacity by modifying the underlying volumes when possible, or by adding new volumes when needed. The grow operation happens in the background with zero downtime - your applications continue running without interruption. See How AutoScaler Works for more details on how growing works.

What happens if I write data to my volume during autoscaling activation?

AutoScaler continuously monitors the storage used by your volume and adjusts the size to match it, even during the initial activation. If enough data is written during the initial activation the volume will grow automatically, and complete copying your data to a larger volume size.

Will I never run out of space?

AutoScaler continuously monitors your volumes and grows them before they run out of space, so the full originally provisioned size that your application expects to see is always available to it. If you need to grow beyond the original size of your volume (the size of the filesystem), this can be configured through the Datafy API.

What happens when a volume shrinks? How long does it take?

When a volume's utilization drops below a defined threshold, AutoScaler reclaims the unused space by copying data to new, smaller volumes. The shrink operation runs in the background and is designed to have minimal impact on your workload. The duration depends on the amount of data on the volume and can range from minutes to several hours for larger volumes. For more information, see how shrink works.

Can I reverse autoscaling on a volume?

Yes. You can disable autoscaling on any volume at any time. When autoscaling is disabled, your data is copied back to a standard EBS volume matching the original configuration, and the volume is returned to its unmanaged state. This process runs in the background with no downtime. For more information see how to deactivate a volume and how it works.

What happens if Datafy's control plane is unreachable?

If the Datafy control plane becomes temporarily unavailable, your data remains fully accessible and your applications are not affected. Mission-critical actions - including data access and volume growth - continue to operate locally on the instance, independent of the control plane. Operations that require coordination with the control plane (such as shrink) will resume once connectivity is restored.

Does Datafy use AI for optimization?

No. Datafy's autoscaling decisions are deterministic and rule-based. Datafy does not use AI or machine learning for optimization actions. Volume scaling is driven by real-time usage data and configurable thresholds.

Privacy & Security

Can Datafy access my data?

No. Datafy does not access, read, or use any of the data stored in your EBS volumes. AutoScaler operates at the block storage level and only interacts with volume metadata and filesystem-reported usage.

Which AWS permissions does AutoScaler require? Why?

Datafy AutoScaler requires permissions to monitor and modify EBS volumes in your account - including creating, resizing, and attaching volumes - as well as read permissions for instance and volume metadata. The full list of required permissions is detailed in the permissions setup instructions.

Does AutoScaler upgrade automatically? Who controls agent updates?

Datafy does not push updates automatically, you control when AutoScaler is updated. Agent upgrades are performed by you, following the steps in the upgrade guide.

Compatibility & Integration

Does AutoScaler work with EBS Snapshots?

Yes. Datafy supports EBS Snapshots. See Datafy Snapshots for details on how snapshots work with autoscaled volumes.

Is AutoScaler compatible with Terraform, Pulumi, or other IaC tools?

Yes. Datafy is designed to work alongside your existing infrastructure-as-code workflows. See the IaC Reconciliation page for details on how Datafy handles reconciliation with tools like Terraform and Pulumi.

Does AutoScaler work with pod rescheduling tools like Cast AI or Karpenter?

Yes. AutoScaler works alongside pod rescheduling and node management tools like Cast AI and Karpenter. When pods are rescheduled to a different node, AutoScaler follows the volumes as they move. Note that the instance types provisioned by your node provisioner, such as Cast AI or Karpenter, must be within AutoScaler's supported instance types.

Scale and Performance

How does AutoScaler handle large-scale accounts?

All autoscaling actions (grow, shrink, data access) are performed locally on the instance, so they are not affected by account scale. Whether you have 10 volumes or 10,000, the performance of autoscaling operations on each instance remains the same. The one constraint to be aware of is that volume migration (such as during a shrink) is limited to one volume at a time per instance, to minimize resource usage. Multiple instances can perform migrations concurrently.

I have many AWS accounts. How does Datafy handle multi-account environments?

Each AWS account is associated with its own Datafy account. You can create as many accounts as you need, and view them all in the Datafy App.

How do I activate autoscaling at scale, or in a dynamic environment?

For environments where volumes are frequently created or where you want to avoid manual activation, you can use Autoscaling Rules. Rules automatically enable autoscaling on new volumes that match your defined criteria, without any manual intervention.

Is the performance of my volumes impacted?

In steady state, there is no performance impact on your volumes. During a grow operation, the process is fast and transparent to your applications. During a shrink operation, data is copied in the background. This process is throttled to use only excess I/O capacity, so the impact on your running workloads is minimal.