Snapshot a volume

Copy a volume's blocks into a snapshot in its zone, incrementally, and delete snapshots you no longer need.

Last updated 18 November 2026View as Markdown

A volume snapshot is the volume's raw blocks, cut into 4 MiB pieces and stored in the zone. To restore one, see Restore a volume snapshot. Replace the ids with your own.

Important

A snapshot of an attached volume stops its instance for the whole copy and starts it again afterwards. Connections to it drop.

Prerequisites

  • Permission for ec2:CreateSnapshot; with an attached volume also ec2:StopInstances on its instance. Deleting needs ec2:DeleteSnapshot.
  • In the console, also ec2:DescribeVolumes to open Volumes and ec2:DescribeSnapshots to open Volume snapshots (without either, its page says "Access denied").
  • The volume is available or in-use. A volume still creating or in error is refused with 409 "The volume is not ready".
  • No other snapshot or restore of the volume is running, else 409 "Snapshot … of this volume is in progress; try again when it is done" or "This volume is being restored from …; wait until it is available".
  • Room in the volume snapshots quota.

Take a snapshot

  1. Open Compute, then Volumes.
  2. In the volume's row menu, choose Create snapshot.
  3. (Optional) For Name, enter before-upgrade.
  4. If the volume is attached, under Attached instance, select Stop instance for the snapshot and start it again afterwards.
  5. Choose Create snapshot. Volume snapshots opens.

An attached volume without stopInstance is refused with 409 "The volume is attached to …. Detach it, or pass stopInstance: true to stop the instance for the snapshot and start it again afterwards." With it, the instance is stopped, the volume detached and copied, then reattached and the instance started again. A detached volume is copied without touching anything else.

Result

The answer is 202 with a snapshot snap-… in state pending. It moves to creating (with progressPercent) and then completed, or error with the reason. Volume snapshots shows the progress. While a snapshot runs, the volume cannot be attached, detached, grown or deleted.

  • Encrypted volumes are copied as they are on disk, still encrypted; the copy is made without the key. The snapshot keeps its own copy of the volume's key (wrapped like the volume's, under the same KMS key if you chose one), so it stays restorable after the volume is deleted.
  • Incremental. A piece already stored by an earlier snapshot of any volume of your organization in that zone is not uploaded or stored again; all-zero pieces are not stored at all. storedBytes is the data the snapshot references; newBytes is what it was the first to store. Every block of the volume is read each time, however little changed.
  • Cost. Snapshot storage is metered daily: each stored piece counts once, on the oldest snapshot that holds it. Where snapshot storage is not charged yet, it shows as "metered, not charged yet"; you are told before charging starts.

Measured on 2026-11-12, on a 50 GiB LUKS2-encrypted volume holding 4.3 GiB of files, snapshotted with stopInstance: the first snapshot stored 5.43 GiB in 309 seconds end to end (stopping and starting the instance included); a second one after 1 GiB of new data and one 128 MiB file rewritten uploaded 1,200 MiB in 272 seconds.

Delete a snapshot

  1. Open Compute, then Volume snapshots, and choose the snapshot in the list.
  2. Under Delete this snapshot, choose Delete snapshot.
  3. Enter the snapshot's name (its id if it has none) and choose Delete snapshot.

The answer is 202 with the snapshot deleting; the pieces no other snapshot of your organization uses are removed. A delete is refused with 409 "The snapshot is still being taken; delete it after" while it is pending or creating, and with 409 "A volume is being restored from this snapshot" during a restore.

Did this page help you?