Tidb backup and recovery introduction

This document describes how to Kubernetes Upper TiDB The cluster performs data backup and data recovery . The tools used in backup and recovery are  Dumpling、TiDB Lightning  and  BR.

Dumpling  Is a data export tool , The tool can store in TiDB/MySQL The data in is exported as SQL perhaps CSV Format , It can be used to complete logical full backup or export .

TiDB Lightning  Is a data import tool , This tool can put Dumpling or CSV Data in output format can be quickly imported into TiDB in , It can be used to complete logical full recovery or import .

BR  yes TiDB Command line tools for distributed backup and recovery , Used to deal with TiDB Cluster for data backup and recovery . comparison Dumpling and Mydumper,BR It is more suitable for scenarios with a large amount of data ,BR Only support TiDB v3.1 And above . If you need incremental backups that are not sensitive to latency , see also  BR. If you need real-time incremental backup , see also  TiCDC.

Use scenarios

The data backup

If you have the following requirements for data backup , Consider using  BR  Yes TiDB Data backup ：

• The amount of data backed up is large （ Greater than 1 TB）, And it requires faster backup
• Directly backup data SST file （ Key value pair ）
• Delay insensitive incremental backup

BR Relevant use documents can be referred to ：

• Use BR Backup TiDB Cluster to compatible S3 The storage
• Use BR Backup TiDB Cluster to GCS
• Use BR Backup TiDB Cluster to Azure Blob Storage
• Use BR Backup TiDB Cluster to persistent volume

If you have the following requirements for data backup , Consider using  Dumpling  Yes TiDB Data backup ：

• export SQL or CSV Formatted data
• For single SQL Limit the memory of the statement
• export TiDB Snapshot of historical data

Dumpling Relevant use documents can be referred to ：

• Use Dumpling Backup TiDB Cluster data to compatible S3 The storage
• Use Dumpling Backup TiDB Cluster data to GCS

Data recovery

If you need from BR Backed up SST File pair TiDB Data recovery , You should use BR. Relevant use documents can be referred to ：

• Use BR Restore compatibility S3 Backup data on storage
• Use BR recovery GCS Backup data on
• Use BR recovery Azure Blob Storage Backup data on
• Use BR Restore the backup data on the persistent volume

If you need from Dumpling Exported or other format compatible SQL or CSV File pair TiDB Data recovery , You should use TiDB Lightning. Relevant use documents can be referred to ：

• Use TiDB Lightning Restore compatibility S3 Backup data on storage
• Use TiDB Lightning recovery GCS Backup data on

Backup and recovery process

In order to Kubernetes Upper TiDB Cluster data backup , Users need to create a custom  Backup Custom Resource (CR) Object to describe a backup , Or create a custom  BackupSchedule CR  Object to describe a scheduled backup .

In order to Kubernetes Upper TiDB Cluster for data recovery , Users can create a customized  Restore CR  Object to describe a recovery .

After creating the corresponding CR After the object ,TiDB Operator The backup or recovery will be performed according to the corresponding configuration and select the corresponding tool .

Delete the backed up Backup CR

The user can delete the corresponding backup by the following statement CR Or regular full backup CR.

kubectl delete backup ${name} -n ${namespace} kubectl delete backupschedule ${name} -n ${namespace}

If you use v1.1.2 And previous versions , Or use v1.1.3 And later versions will  spec.cleanPolicy  Set to  Delete  when ,TiDB Operator In the delete CR The backup files will be cleaned up at the same time .

When the above conditions are met , If you need to delete namespace, It is suggested to delete all Backup/BackupSchedule CR, And then delete namespace.

If it is deleted directly Backup/BackupSchedule CR Of namespace,TiDB Operator Will continue to try to create Job Clean up the backed up data , But because namespace be in  Terminating  State and failed to create , Which leads to namespace Stuck in this state .

At this time, you need to delete  finalizers：

kubectl patch -n ${namespace} backup ${name} --type merge -p '{"metadata":{"finalizers":[]}}'

Clean up backup files

TiDB Operator v1.2.3 And previous versions , The way to clean up backup files is ： Delete backup files circularly , Delete one file at a time .

TiDB Operator v1.2.4 And later versions , The way to clean up backup files is ： Delete backup files circularly , Delete multiple files in batch at a time . For the operation of deleting multiple files in batch each time , Depending on the type of back-end storage used for backup , There are different ways to delete .

• S3 Compatible back-end storage adopts concurrent batch deletion .TiDB Operator Start multiple Go coroutines , Every Go The interface is deleted in batches every time the coroutine calls  "DeleteObjects"  To delete multiple files .
• Other types of back-end storage adopt concurrent deletion .TiDB Operator Start multiple Go coroutines , Every Go The process Deletes one file at a time .

about TiDB Operator v1.2.4 And later versions , You can use Backup CR The following fields in control the cleanup behavior ：

• .spec.cleanOption.pageSize： Specify the number of files to be deleted in batch each time . The default value is 10000.

• .spec.cleanOption.disableBatchConcurrency： When set to true when ,TiDB Operator Will disable concurrent batch deletion , Use concurrent deletion .

  If S3 Compatible backend storage does not support  DeleteObjects  Interface , The default concurrent batch deletion will fail , This field needs to be configured as  true  To use concurrent deletion .

• .spec.cleanOption.batchConcurrency: Specify the Go Number of trips . The default value is 10.

• .spec.cleanOption.routineConcurrency: Specify the Go Number of trips . The default value is 100.