Use Dumpling to Export Data

This document introduces the data export tool - Dumpling. Dumpling exports data stored in TiDB/MySQL as SQL or CSV data files and can be used to make a logical full backup or export. Dumpling also supports exporting data to Amazon S3.

You can get Dumpling using TiUP by running tiup install dumpling. Afterwards, you can use tiup dumpling ... to run Dumpling.

Dumpling is also included in the tidb-toolkit installation package and can be download here.

For detailed usage of Dumpling, use the --help option or refer to Option list of Dumpling.

When using Dumpling, you need to execute the export command on a running cluster.

TiDB also provides other tools that you can choose to use as needed.

• For backups of SST files (key-value pairs) or backups of incremental data that are not sensitive to latency, refer to BR.
• For real-time backups of incremental data, refer to TiCDC.
• All exported data can be imported back to TiDB using TiDB Lightning.

Compared to Mydumper, Dumpling has the following improvements:

• Support exporting data in multiple formats, including SQL and CSV.
• Support the table-filter feature, which makes it easier to filter data.
• Support exporting data to Amazon S3 cloud storage.
• More optimizations are made for TiDB:
  • Support configuring the memory limit of a single TiDB SQL statement.
  • Support automatic adjustment of TiDB GC time for TiDB v4.0.0 and later versions.
  • Use TiDB's hidden column _tidb_rowid to optimize the performance of concurrent data export from a single table.
  • For TiDB, you can set the value of tidb_snapshot to specify the time point of the data backup. This ensures the consistency of the backup, instead of using FLUSH TABLES WITH READ LOCK to ensure the consistency.

Export data from TiDB or MySQL

Required privileges

• SELECT
• RELOAD
• LOCK TABLES
• REPLICATION CLIENT

Export to SQL files

This document assumes that there is a TiDB instance on the 127.0.0.1:4000 host and that this TiDB instance has a root user without a password.

Dumpling exports data to SQL files by default. You can also export data to SQL files by adding the --filetype sql flag:

dumpling \
  -u root \
  -P 4000 \
  -h 127.0.0.1 \
  --filetype sql \
  --t 8 \
  -o /tmp/test \
  -r 200000 \
  -F 256MiB

In the command above:

• The -h, -p, and -u option respectively mean the address, the port, and the user. If a password is required for authentication, you can use -p $YOUR_SECRET_PASSWORD to pass the password to Dumpling.
• The -o option specifies the export directory of the storage, which supports a local file path or a URL of an external storage.
• The t option specifies the number of threads for the export. Increasing the number of threads improves the concurrency of Dumpling and the export speed, and also increases the database's memory consumption. Therefore, it is not recommended to set the number too large.
• The -r option specifies the maximum number of rows in a single file. With this option specified, Dumpling enables the in-table concurrency to speed up the export and reduce the memory usage.
• The -F option is used to specify the maximum size of a single file (the unit here is MiB; inputs like 5GiB or 8KB are also acceptable). It is recommended to keep its value to 256 MiB or less if you plan to use TiDB Lightning to load this file into a TiDB instance.

Export to CSV files

If Dumpling exports data to CSV files (use --filetype csv to export to CSV files), you can also use --sql <SQL> to export the records selected by the specified SQL statement.

For example, you can export all records that match id < 100 in test.sbtest1 using the following command:

./dumpling \
  -u root \
  -P 4000 \
  -h 127.0.0.1 \
  -o /tmp/test \
  --filetype csv \
  --sql 'select * from `test`.`sbtest1` where id < 100'

Format of exported files

• metadata: The start time of the exported files and the position of the master binary log.

  cat metadata
  

  Started dump at: 2020-11-10 10:40:19
  SHOW MASTER STATUS:
          Log: tidb-binlog
          Pos: 420747102018863124
  Finished dump at: 2020-11-10 10:40:20
  

• {schema}-schema-create.sql: The SQL file used to create the schema

  cat test-schema-create.sql
  

  CREATE DATABASE `test` /*!40100 DEFAULT CHARACTER SET utf8mb4 */;
  

• {schema}.{table}-schema.sql: The SQL file used to create the table

  cat test.t1-schema.sql
  

  CREATE TABLE `t1` (
    `id` int(11) DEFAULT NULL
  ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin;
  

• {schema}.{table}.{0001}.{sql|csv}: The date source file

  cat test.t1.0.sql
  

  /*!40101 SET NAMES binary*/;
  INSERT INTO `t1` VALUES
  (1);
  

• *-schema-view.sql、*-schema-trigger.sql、*-schema-post.sql: Other exported files

Export data to Amazon S3 cloud storage

Starting from v4.0.8, Dumpling supports exporting data to cloud storages. If you need to back up data to Amazon S3, you need to specify the Amazon S3 storage path in the -o parameter.

You need to create an Amazon S3 bucket in the specified region (see the Amazon documentation - How do I create an S3 Bucket). If you also need to create a folder in the bucket, see the Amazon documentation - Creating a folder.

Pass SecretKey and AccessKey of the account with the permission to access the Amazon S3 backend storage to the Dumpling node as environment variables.

export AWS_ACCESS_KEY_ID=${AccessKey}
export AWS_SECRET_ACCESS_KEY=${SecretKey}

Dumpling also supports reading credential files from ~/.aws/credentials. For more Dumpling configuration, see the configuration of External storages.

When you back up data using Dumpling, explicitly specify the --s3.region parameter, which means the region of the S3 storage (for example, ap-northeast-1):

./dumpling \
  -u root \
  -P 4000 \
  -h 127.0.0.1 \
  -r 200000 \
  -o "s3://${Bucket}/${Folder}" \
  --s3.region "${region}"

Filter the exported data

Use the --where option to filter data

By default, Dumpling exports all databases except system databases (including mysql, sys, INFORMATION_SCHEMA, PERFORMANCE_SCHEMA, METRICS_SCHEMA, and INSPECTION_SCHEMA). You can use --where <SQL where expression> to select the records to be exported.

./dumpling \
  -u root \
  -P 4000 \
  -h 127.0.0.1 \
  -o /tmp/test \
  --where "id < 100"

The above command exports the data that matches id < 100 from each table. Note that you cannot use the --where parameter together with --sql.

Use the --filter option to filter data

Dumpling can filter specific databases or tables by specifying the table filter with the --filter option. The syntax of table filters is similar to that of .gitignore. For details, see Table Filter.

./dumpling \
  -u root \
  -P 4000 \
  -h 127.0.0.1 \
  -o /tmp/test \
  -r 200000 \
  --filter "employees.*" \
  --filter "*.WorkOrder"

The above command exports all the tables in the employees database and the WorkOrder tables in all databases.

Use the -B or -T option to filter data

Dumpling can also export specific databases with the -B option or specific tables with the -T option.

Examples:

• -B employees exports the employees database.
• -T employees.WorkOrder exports the employees.WorkOrder table.

Improve export efficiency through concurrency

The exported file is stored in the ./export-<current local time> directory by default. Commonly used options are as follows:

• The t option specifies the number of threads for the export. Increasing the number of threads improves the concurrency of Dumpling and the export speed, and also increases the database's memory consumption. Therefore, it is not recommended to set the number too large.
• The -r option specifies the maximum number of records (or the number of rows in the database) for a single file. When it is enabled, Dumpling enables concurrency in the table to improve the speed of exporting large tables.
• The --compress gzip option can be used to compress the dump. This can help to speed up dumping of data if storage is the bottleneck or if storage capacity is a concern. The drawback of this is an increase in CPU usage. Each file is compressed individually.

With the above options specified, Dumpling can have a quicker speed of data export.

Adjust Dumpling's data consistency options

Dumpling uses the --consistency <consistency level> option to control the way in which data is exported for "consistency assurance". For TiDB, data consistency is guaranteed by getting a snapshot of a certain timestamp by default (namely, --consistency snapshot). When using snapshot for consistency, you can use the --snapshot option to specify the timestamp to be backed up. You can also use the following levels of consistency:

• flush: Use FLUSH TABLES WITH READ LOCK to temporarily interrupt the DML and DDL operations of the replica database, to ensure the global consistency of the backup connection, and to record the binlog position (POS) information. The lock is released after all backup connections start transactions. It is recommended to perform full backups during off-peak hours or on the MySQL replica database.
• snapshot: Get a consistent snapshot of the specified timestamp and export it.
• lock: Add read locks on all tables to be exported.
• none: No guarantee for consistency.
• auto: Use flush for MySQL and snapshot for TiDB.

After everything is done, you can see the exported file in /tmp/test:

ls -lh /tmp/test | awk '{print $5 "\t" $9}'

140B  metadata
66B   test-schema-create.sql
300B  test.sbtest1-schema.sql
190K  test.sbtest1.0.sql
300B  test.sbtest2-schema.sql
190K  test.sbtest2.0.sql
300B  test.sbtest3-schema.sql
190K  test.sbtest3.0.sql

Export historical data snapshots of TiDB

Dumpling can export the data of a certain tidb_snapshot with the --snapshot option specified.

The --snapshot option can be set to a TSO (the Position field output by the SHOW MASTER STATUS command) or a valid time of the datetime data type (in the form of YYYY-MM-DD hh:mm:ss), for example:

./dumpling --snapshot 417773951312461825
./dumpling --snapshot "2020-07-02 17:12:45"

The TiDB historical data snapshots when the TSO is 417773951312461825 and the time is 2020-07-02 17:12:45 are exported.

Control the memory usage of exporting large tables

When Dumpling is exporting a large single table from TiDB, Out of Memory (OOM) might occur because the exported data size is too large, which causes connection abort and export failure. You can use the following parameters to reduce the memory usage of TiDB:

• Setting -r to split the data to be exported into chunks. This reduces the memory overhead of TiDB's data scan and enables concurrent table data dump to improve export efficiency.
• Reduce the value of --tidb-mem-quota-query to 8589934592 (8 GB) or lower. --tidb-mem-quota-query controls the memory usage of a single query statement in TiDB.
• Adjust the --params "tidb_distsql_scan_concurrency=5" parameter. tidb_distsql_scan_concurrency is a session variable which controls the concurrency of the scan operations in TiDB.

Set TiDB GC when exporting a large volume of data (more than 1 TB)

When exporting data from TiDB, if the TiDB version is greater than or equal to v4.0.0 and Dumpling can access the PD address of the TiDB cluster, Dumpling automatically extends the GC time without affecting the original cluster.

In other scenarios, if the data size is very large, to avoid export failure due to GC during the export process, you can extend the GC time in advance:

update mysql.tidb set VARIABLE_VALUE = '720h' where VARIABLE_NAME = 'tikv_gc_life_time';

After your operation is completed, set the GC time back (the default value is 10m):

update mysql.tidb set VARIABLE_VALUE = '10m' where VARIABLE_NAME = 'tikv_gc_life_time';

Option list of Dumpling