Deploy a DM Cluster Using TiUP

TiUP is a cluster operation and maintenance tool introduced in TiDB 4.0. TiUP provides TiUP DM, a cluster management component written in Golang. By using TiUP DM, you can easily perform daily TiDB Data Migration (DM) operations, including deploying, starting, stopping, destroying, scaling, and upgrading a DM cluster, and manage DM cluster parameters.

TiUP supports deploying DM v2.0 or later DM versions. This document introduces how to deploy DM clusters of different topologies.

Use TiUP (Recommended) - 图1

Note

If your target machine’s operating system supports SELinux, make sure that SELinux is disabled.

Prerequisites

When DM performs a full data replication task, the DM-worker is bound with only one upstream database. The DM-worker first exports the full amount of data locally, and then imports the data into the downstream database. Therefore, the worker’s host space must be large enough to store all upstream tables to be exported. The storage path is specified later when you create the task.

In addition, you need to meet the hardware and software requirements when deploying a DM cluster.

Step 1: Install TiUP on the control machine

Log in to the control machine using a regular user account (take the tidb user as an example). All the following TiUP installation and cluster management operations can be performed by the tidb user.

  1. Install TiUP by executing the following command:

    1. curl --proto '=https' --tlsv1.2 -sSf https://tiup-mirrors.pingcap.com/install.sh | sh

    After the installing, ~/.bashrc has been modified to add TiUP to PATH, so you need to open a new terminal or redeclare the global environment variables source ~/.bashrc to use it.

  2. Install the TiUP DM component:

    1. tiup install dm dmctl

Step 2: Edit the initialization configuration file

According to the intended cluster topology, you need to manually create and edit the cluster initialization configuration file.

You need to create a YAML configuration file (named topology.yaml for example) according to the configuration file template. For other scenarios, edit the configuration accordingly.

You can use the command tiup dm template > topology.yaml to generate a configuration file template quickly.

The configuration of deploying three DM-masters, three DM-workers, and one monitoring component instance is as follows:

  1. # The global variables apply to all other components in the configuration. If one specific value is missing in the component instance, the corresponding global variable serves as the default value.
  2. global:
  3.   user: "tidb"
  4.   ssh_port: 22
  5.   deploy_dir: "/dm-deploy"
  6.   data_dir: "/dm-data"
  8. server_configs:
  9.   master:
  10.     log-level: info
  11.     # rpc-timeout: "30s"
  12.     # rpc-rate-limit: 10.0
  13.     # rpc-rate-burst: 40
  14.   worker:
  15.     log-level: info
  17. master_servers:
  18.   - host: 10.0.1.11
  19.     name: master1
  20.     ssh_port: 22
  21.     port: 8261
  22.     # peer_port: 8291
  23.     # deploy_dir: "/dm-deploy/dm-master-8261"
  24.     # data_dir: "/dm-data/dm-master-8261"
  25.     # log_dir: "/dm-deploy/dm-master-8261/log"
  26.     # numa_node: "0,1"
  27.     # The following configs are used to overwrite the `server_configs.master` values.
  28.     config:
  29.       log-level: info
  30.       # rpc-timeout: "30s"
  31.       # rpc-rate-limit: 10.0
  32.       # rpc-rate-burst: 40
  33.   - host: 10.0.1.18
  34.     name: master2
  35.     ssh_port: 22
  36.     port: 8261
  37.   - host: 10.0.1.19
  38.     name: master3
  39.     ssh_port: 22
  40.     port: 8261
  41. # If you do not need to ensure high availability of the DM cluster, deploy only one DM-master node, and the number of deployed DM-worker nodes must be no less than the number of upstream MySQL/MariaDB instances to be migrated.
  42. # To ensure high availability of the DM cluster, it is recommended to deploy three DM-master nodes, and the number of deployed DM-worker nodes must exceed the number of upstream MySQL/MariaDB instances to be migrated (for example, the number of DM-worker nodes is two more than the number of upstream instances).
  43. worker_servers:
  44.   - host: 10.0.1.12
  45.     ssh_port: 22
  46.     port: 8262
  47.     # deploy_dir: "/dm-deploy/dm-worker-8262"
  48.     # log_dir: "/dm-deploy/dm-worker-8262/log"
  49.     # numa_node: "0,1"
  50.     # The following configs are used to overwrite the `server_configs.worker` values.
  51.     config:
  52.       log-level: info
  53.   - host: 10.0.1.19
  54.     ssh_port: 22
  55.     port: 8262
  57. monitoring_servers:
  58.   - host: 10.0.1.13
  59.     ssh_port: 22
  60.     port: 9090
  61.     # deploy_dir: "/tidb-deploy/prometheus-8249"
  62.     # data_dir: "/tidb-data/prometheus-8249"
  63.     # log_dir: "/tidb-deploy/prometheus-8249/log"
  65. grafana_servers:
  66.   - host: 10.0.1.14
  67.     port: 3000
  68.     # deploy_dir: /tidb-deploy/grafana-3000
  70. alertmanager_servers:
  71.   - host: 10.0.1.15
  72.     ssh_port: 22
  73.     web_port: 9093
  74.     # cluster_port: 9094
  75.     # deploy_dir: "/tidb-deploy/alertmanager-9093"
  76.     # data_dir: "/tidb-data/alertmanager-9093"
  77.     # log_dir: "/tidb-deploy/alertmanager-9093/log"

Use TiUP (Recommended) - 图2

Note

  • It is not recommended to run too many DM-workers on one host. Each DM-worker should be allocated at least 2 core CPU and 4 GiB memory.

  • Make sure that the ports among the following components are interconnected:

    • The peer_port (8291 by default) among the DM-master nodes are interconnected.
    • Each DM-master node can connect to the port of all DM-worker nodes (8262 by default).
    • Each DM-worker node can connect to the port of all DM-master nodes (8261 by default).
    • The TiUP nodes can connect to the port of all DM-master nodes (8261 by default).
    • The TiUP nodes can connect to the port of all DM-worker nodes (8262 by default).

For more master_servers.host.config parameter description, refer to master parameter. For more worker_servers.host.config parameter description, refer to worker parameter.

Step 3: Execute the deployment command

Use TiUP (Recommended) - 图3

Note

You can use secret keys or interactive passwords for security authentication when you deploy TiDB using TiUP:

  • If you use secret keys, you can specify the path of the keys through -i or --identity_file;
  • If you use passwords, add the -p flag to enter the password interaction window;
  • If password-free login to the target machine has been configured, no authentication is required.
  1. tiup dm deploy ${name} ${version} ./topology.yaml -u ${ssh_user} [-p] [-i /home/root/.ssh/gcp_rsa]

The parameters used in this step are as follows.

ParameterDescription
${name}The name of the DM cluster, eg: dm-test
${version}The version of the DM cluster. You can see other supported versions by running tiup list dm-master.
./topology.yamlThe path of the topology configuration file.
-u or —userLog in to the target machine as the root user or other user account with ssh and sudo privileges to complete the cluster deployment.
-p or —passwordThe password of target hosts. If specified, password authentication is used.
-i or —identity_fileThe path of the SSH identity file. If specified, public key authentication is used (default “/root/.ssh/id_rsa”).

At the end of the output log, you will see Deployed cluster `dm-test` successfully. This indicates that the deployment is successful.

Step 4: Check the clusters managed by TiUP

  1. tiup dm list

TiUP supports managing multiple DM clusters. The command above outputs information of all the clusters currently managed by TiUP, including the name, deployment user, version, and secret key information:

  1. Name  User  Version  Path                                  PrivateKey
  2. ----  ----  -------  ----                                  ----------
  3. dm-test  tidb  ${version}  /root/.tiup/storage/dm/clusters/dm-test  /root/.tiup/storage/dm/clusters/dm-test/ssh/id_rsa

Step 5: Check the status of the deployed DM cluster

To check the status of the dm-test cluster, execute the following command:

  1. tiup dm display dm-test

Expected output includes the instance ID, role, host, listening port, and status (because the cluster is not started yet, so the status is Down/inactive), and directory information.

Step 6: Start the DM cluster

  1. tiup dm start dm-test

If the output log includes Started cluster `dm-test` successfully, the start is successful.

Step 7: Verify the running status of the DM cluster

Check the DM cluster status using TiUP:

  1. tiup dm display dm-test

If the Status is Up in the output, the cluster status is normal.

Step 8: Managing migration tasks using dmctl

dmctl is a command-line tool used to control DM clusters. You are recommended to use dmctl via TiUP.

dmctl supports both the command mode and the interactive mode. For details, see Maintain DM Clusters Using dmctl.