Rolling Back Disk Data from a Snapshot

During routine data management, EVS disk data may be accidentally lost or corrupted, which can directly affect the service running. If a disk has snapshots, you can restore it to the state captured at a snapshot, enabling rapid data recovery and ensuring service continuity.

Snapshot rollback uses lazy loading. Once the rollback is complete, the disk is available for use without waiting for all snapshot data to be loaded. Snapshot data is loaded to the disk on demand in the background.

Customer impact during lazy loading:

• Impact on read/write operations

  First read (data not yet loaded): When data is accessed for the first time, the system reads it directly from snapshot storage, which results in higher latency than the disk's rated performance. Subsequent reads of the same data blocks return to normal performance.

  First write (data not yet loaded): When writing to an unloaded data block, the system must first load the corresponding snapshot data before completing the write, which may increase initial write latency. Write performance for already loaded or previously written data blocks is not affected.

  Data loading completed: The disk's read/write performance returns to the normal level for its disk type. Lazy loading does not affect the final performance, capacity, or integrity of the rolled-back data.

• Typical impact on workloads
  • If workloads resume immediately after rollback, many first-time disk accesses may occur within a short period, such as when a database starts up or an application reads large numbers of files. During this period, overall I/O latency may increase significantly, slowing startup and increasing response times.
  • Before data restoration is complete, measured disk performance may fall below the disk's rated IOPS or throughput limits. Avoid running performance benchmarks during this period, as the results will not reflect steady-state performance.
  • Data restoration does not affect the correctness or integrity of the rolled-back data, and does not block operations such as disk attachment or file system mounting.

Usage recommendations:

• Rollback is often performed to recover from service failures. If rapid recovery is required, you can attach the disk and restart your workloads, accepting temporarily higher latency on first access. Data will continue loading in the background.
• If performance is critical, consider performing a full read of the disk after rollback completes, for example, by sequentially reading the entire disk using tools such as dd. Resume production traffic only after data loading is complete.
• Perform rollback during off-peak hours, and allow sufficient time for data loading to complete to avoid the added impact of high first-access latency during peak workload periods.
• Do not interpret elevated read/write latency during lazy loading as evidence of a disk failure or rollback failure. If performance issues persist after data loading is complete, contact technical support.

• Snapshot data can only be rolled back to source EVS disks. Rollback to a different disk is not possible.
• You can roll back disk data from a snapshot only when the snapshot is in the Available state and its source disk is in the Available or Rollback failed state.
• A snapshot whose name starts with autobk_snapshot_vbs_, manualbk_snapshot_vbs_, autobk_snapshot_csbs_, or manualbk_snapshot_csbs_ is automatically generated during backup. Such a snapshot can only be viewed. It cannot be used to roll back disk data.

• At least a snapshot has been created for the disk.
• You can only roll back disk data from a snapshot when the source disk status is Available (not attached to any server) or Rollback failed. If the source disk is attached, detach it first.

1. Sign in to the console.
2. Click in the upper left corner and choose Storage > Elastic Volume Service.

   The Elastic Volume Service page is displayed.

3. In the navigation pane on the left, choose Elastic Volume Service > Snapshots.

   The Snapshots page is displayed.

4. In the snapshot list, locate the target snapshot and click Roll Back Disk in the Operation column.

   Figure 1 Rolling back data

   
   

5. In the displayed dialog box, click Yes.

   The snapshot list is displayed. After the snapshot status changes from Rolling back to Available, the data rollback is successful.

• A rollback operation restores disk data to the time when the snapshot was taken. Any changes made concerning system configuration, software, files, and others after that time will be lost. Reconfigure them after the rollback.
• If you expand the disk capacity after creating a snapshot for a data disk, the partition and file system created on the expanded capacity will be lost if you roll back data from this snapshot. You need to expand the capacity again. For details, see Extending Disk Partitions and File Systems (Linux) and Extending Disk Partitions and File Systems (Windows).
  Note

  After the rollback, operations such as reattaching the disk and mounting the file system can proceed normally and are unaffected by lazy loading. However, the first read of file data triggers on-demand restoration of the corresponding snapshot data. As a result, response times for the first access after the service resumes may be longer than usual. This is normal, and performance returns to normal automatically once data restoration is complete.