DM Cluster Performance Test

This document describes how to build a test scenario to do a performance test on the DM cluster, including the speed test and latency test regarding data migration.

Migration data flow

You can use a simple migration data flow, that is, MySQL -> DM -> TiDB, to test the data migration performance of the DM cluster.

Deploy test environment

  • Deploy the TiDB test cluster using TiUP, with all default configurations.
  • Deploy the MySQL service. Enable the ROW mode for binlog, and use default configurations for other configuration items.
  • Deploy a DM cluster, with a DM-worker and a DM-master.

Performance test

Table schema

Use tables with the following schema for the performance test:

CREATE TABLE `sbtest` ( `id` int(11) NOT NULL AUTO_INCREMENT, `k` int(11) NOT NULL DEFAULT '0', `c` char(120) CHARSET utf8mb4 COLLATE utf8mb4_bin NOT NULL DEFAULT '', `pad` char(60) CHARSET utf8mb4 COLLATE utf8mb4_bin NOT NULL DEFAULT '', PRIMARY KEY (`id`), KEY `k_1` (`k`) ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin

Full import benchmark case

Generate test data

Use sysbench to create test tables upstream and generate test data for full import. Execute the following sysbench command to generate test data:

sysbench --test=oltp_insert --tables=4 --mysql-host=172.16.4.40 --mysql-port=3306 --mysql-user=root --mysql-db=dm_benchmark --db-driver=mysql --table-size=50000000 prepare

Create a data migration task

  1. Create an upstream MySQL source and set source-id to source-1. For details, see Load the Data Source Configurations.

  2. Create a migration task (in full mode). The following is a task configuration template:

    --- name: test-full task-mode: full # Configure the migration task using the TiDB information of your actual test environment. target-database: host: "192.168.0.1" port: 4000 user: "root" password: "" mysql-instances: - source-id: "source-1" block-allow-list: "instance" mydumper-config-name: "global" loader-thread: 16 # Configure the name of the database where sysbench generates data. block-allow-list: instance: do-dbs: ["dm_benchmark"] mydumpers: global: rows: 32000 threads: 32

For details about how to create a migration task, see Create a Data Migration Task.

Get test results

Observe the DM-worker log. When you see all data files have been finished, it means that full data has been imported. In this case, you can see the time consumed to import data. The sample log is as follows:

[INFO] [loader.go:604] ["all data files have been finished"] [task=test] [unit=load] ["cost time"=52.439796ms]

According to the size of the test data and the time consumed to import data, you can calculate the migration speed of the full data.

Incremental replication benchmark case

Initialize tables

Use sysbench to create test tables in the upstream.

Create a data migration task

  1. Create the source of the upstream MySQL. Set source-id to source-1 (if the source has been created in the full import benchmark case, you do not need to create it again). For details, see Load the Data Source Configurations.

  2. Create a DM migration task (in all mode). The following is an example of the task configuration file:

    --- name: test-all task-mode: all # Configure the migration task using the TiDB information of your actual test environment. target-database: host: "192.168.0.1" port: 4000 user: "root" password: "" mysql-instances: - source-id: "source-1" block-allow-list: "instance" syncer-config-name: "global" # Configure the name of the database where sysbench generates data. block-allow-list: instance: do-dbs: ["dm_benchmark"] syncers: global: worker-count: 16 batch: 100

For details about how to create a data migration task, see Create a Data Migration Task.

Generate incremental data

To continuously generate incremental data in the upstream, run the sysbench command:

sysbench --test=oltp_insert --tables=4 --num-threads=32 --mysql-host=172.17.4.40 --mysql-port=3306 --mysql-user=root --mysql-db=dm_benchmark --db-driver=mysql --report-interval=10 --time=1800 run

Get test results

To observe the migration status of DM, you can run the query-status command. To observe the monitoring metrics of DM, you can use Grafana. Here the monitoring metrics refer to finished sqls jobs (the number of jobs finished per unit time), and other related metrics. For more information, see Binlog Migration Monitoring Metrics.