All Products
Search
Document Center

PolarDB:Overall process and estimated time

Last Updated:Mar 28, 2026

Database and table restoration lets you recover accidentally modified or deleted tables back to the original cluster without restoring the entire cluster. Use this feature when you need table-level or database-level granularity, and expect a restoration time measured in minutes to a few hours depending on data size and cluster configuration.

If restoring multiple terabytes of data, consider full cluster restoration from a backup set instead — that process typically completes in a few minutes. For more information, see Method 1 for full restoration: Restore data from a backup set.

The new database and table restoration workflow, which includes the Fast Import feature, is available in a phased release for PolarDB for MySQL 8.0.1 clusters with minor version 8.0.1.1.49 or later.

Choose a restore method

Point-in-time restoreRestore from a backup set
When to useTarget time does not match any existing backup setTarget time matches the creation time of a backup set
ConvenienceRequires replaying redo logs after restoring the backupMore straightforward — no log replay step

How it works

Both restore methods follow the same core sequence:

  1. A temporary node is created.

  2. Data from the target point in time is restored to the temporary node — either from a backup set, or from a backup set plus redo log replay (point-in-time restore only).

  3. The data is copied from the temporary node back to the original cluster.

The new restoration workflow, released on April 3, 2024, significantly reduces the time for step 3. Data is also automatically synchronized to Hot Standby Clusters and Global Database Network (GDN) secondary clusters, eliminating additional manual sync time.

If your cluster's revision version meets the requirements in the next section, the new workflow is used automatically.

image
Schedule restoration during off-peak hours to minimize the impact on your workload.

Supported versions

Database and table restoration is available on PolarDB Enterprise Edition and Standard Edition. The new restoration workflow requires a higher revision version than the basic feature.

To check your cluster's revision version, go to Basic Information > Configuration Information.

Edition seriesMySQL versionArchitectureBasic features (min revision version)New restoration process (min revision version)
Enterprise Edition (Cluster Edition)5.6X865.6.1.0.255.6.1.0.42
Enterprise Edition (Cluster Edition)5.7X865.7.1.0.85.7.1.0.36
Enterprise Edition (Cluster Edition)8.0.1X868.0.1.1.148.0.1.1.46
Enterprise Edition (Cluster Edition)8.0.2X868.0.2.2.08.0.2.2.26
Standard Edition5.6X865.6.1.0.425.6.1.0.42
Standard Edition5.7X865.7.1.0.305.7.1.0.30
Standard Edition8.0.1X868.0.1.1.38.28.0.1.1.38.2
Standard Edition8.0.1Yitian (ARM)8.0.1.1.418.0.1.1.41
Standard Edition8.0.2X868.0.2.2.218.0.2.2.21

Estimated time

Time per step

StepEstimated time
Create a temporary node and restore data from the backup setAbout 3 to 10 minutes
Replay incremental data from redo logs (point-in-time restore only)1.5 GB/minute
Restore data to the original clusterSee Restoration speed reference below
The preceding data is for reference only.

Restoration speed reference

The following tables show measured restoration speeds for a single 200 GB table under different configurations. "Restoration speed" refers to the rate at which data is copied back to the original cluster — it does not include the time to create the temporary node or replay redo logs.

Key factors that affect restoration speed:

  • Whether a Hot Standby Cluster is enabled

  • Primary node CPU and memory specifications

  • The innodb_io_capacity and innodb_io_capacity_max parameter values — dynamically adjustable; changes have a major effect on the new workflow and a minor effect on the old workflow

  • The restoration speed configuration (Quick, Standard, or Safe) — also has a major effect on the new workflow and a minor effect on the old workflow

Restoration speed configurations:

ConfigurationIOPS usageTrade-off
QuickHighestFastest restoration, but consumes more cluster I/O — use when minimizing downtime is the priority
StandardMediumBalances restoration speed against cluster I/O impact
SafeLowestMinimizes cluster I/O impact, but restoration takes longer
Because the 2-core, 8 GB specification is small and has high I/O fluctuations, the restoration speed configuration may not have a noticeable effect. Therefore, test results for this configuration are not listed for different speed configurations.

2 cores, 8 GB — innodb_io_capacity: 4000 / innodb_io_capacity_max: 8000

Hot standby for storageOld workflow — durationOld workflow — speedNew workflow configNew workflow — durationNew workflow — speedSpeed increase vs. old workflow
Yes3h 38m 25s1.03 GB/minStandard1h 43m 36s2.16 GB/min+110%
No2h 23m 0s1.57 GB/minStandard1h 43m 36s2.16 GB/min+38%

4 cores, 16 GB — innodb_io_capacity: 4000 / innodb_io_capacity_max: 8000

Hot standby for storageOld workflow — durationOld workflow — speedNew workflow configNew workflow — durationNew workflow — speedSpeed increase vs. old workflow
Yes3h 3m 31s1.14 GB/minQuick54m 13s3.70 GB/min+225%
NoQuick+88%
Yes3h 3m 31s1.14 GB/minStandard1h 20m2.5 GB/min+119%
NoStandard1h 45m 53s1.97 GB/min+27%
Yes3h 3m 31s1.14 GB/minSafe2h 12m1.52 GB/min+33%
NoSafe-30%

4 cores, 16 GB — innodb_io_capacity: 8000 / innodb_io_capacity_max: 16000

Hot standby for storageOld workflow — durationOld workflow — speedNew workflow configNew workflow — durationNew workflow — speedSpeed increase vs. old workflow
Yes3h 3m 15s1.14 GB/minQuick42m 18s4.76 GB/min+318%
NoQuick+142%
Yes3h 3m 15s1.14 GB/minStandard54m 16s3.70 GB/min+225%
NoStandard1h 45m 53s1.97 GB/min+88%
Yes3h 3m 15s1.14 GB/minSafe1h 20m2.5 GB/min+119%
NoSafe+27%

8 cores, 32 GB — innodb_io_capacity: 4000 / innodb_io_capacity_max: 8000

Hot standby for storageOld workflow — durationOld workflow — speedNew workflow configNew workflow — durationNew workflow — speedSpeed increase vs. old workflow
Yes2h 50m 56s1.19 GB/minQuick54m 39s3.70 GB/min+211%
NoQuick+80%
Yes2h 50m 56s1.19 GB/minStandard1h 21m2.47 GB/min+108%
NoStandard1h 38m 57s2.05 GB/min+20%
Yes2h 50m 56s1.19 GB/minSafe2h 12m1.52 GB/min+28%
NoSafe-35%

8 cores, 32 GB — innodb_io_capacity: 18000 / innodb_io_capacity_max: 36000

Hot standby for storageOld workflow — durationOld workflow — speedNew workflow configNew workflow — durationNew workflow — speedSpeed increase vs. old workflow
Yes2h 51m 5s1.19 GB/minQuick41m 48s4.88 GB/min+310%
NoQuick+273%
Yes2h 51m 5s1.19 GB/minStandard54m 43s3.70 GB/min+211%
NoStandard1h 38m 33s1.31 GB/min+182%
Yes2h 51m 5s1.19 GB/minSafe1h 21m2.47 GB/min+108%
NoSafe+89%

16 cores, 64 GB — innodb_io_capacity: 4000 / innodb_io_capacity_max: 8000

Hot standby for storageOld workflow — durationOld workflow — speedNew workflow configNew workflow — durationNew workflow — speedSpeed increase vs. old workflow
Yes2h 55m 26s1.17 GB/minQuick53m 28s3.77 GB/min+222%
NoQuick+88%
Yes2h 55m 26s1.17 GB/minStandard1h 20m2.5 GB/min+114%
NoStandard1h 42m 20s2.01 GB/min+24%
Yes2h 55m 26s1.17 GB/minSafe2h 12m1.52 GB/min+30%
NoSafe-32%

16 cores, 64 GB — innodb_io_capacity: 20000 / innodb_io_capacity_max: 40000

Hot standby for storageOld workflow — durationOld workflow — speedNew workflow configNew workflow — durationNew workflow — speedSpeed increase vs. old workflow
Yes2h 53m 49s1.19 GB/minQuick41m 1s4.88 GB/min+310%
NoQuick+138%
Yes2h 53m 49s1.19 GB/minStandard54m 5s3.70 GB/min+211%
NoStandard1h 40m 35s2.05 GB/min+80%
Yes2h 53m 49s1.19 GB/minSafe1h 20m2.5 GB/min+110%
NoSafe+22%
The test data above covers a single-table, ~200 GB scenario. Restoring many tables simultaneously also affects speed. Actual performance depends on the underlying hardware model and network conditions. All figures are for reference only.