Scale the TiDB Cluster Using TiDB Ansible

The capacity of a TiDB cluster can be increased or decreased without affecting the online services.

Warning:

In decreasing the capacity, if your cluster has a mixed deployment of other services, do not perform the following procedures. The following examples assume that the removed nodes have no mixed deployment of other services.

Assume that the topology is as follows:

Increase the capacity of a TiDB/TiKV node

For example, if you want to add two TiDB nodes (node101, node102) with the IP addresses 172.16.10.101 and 172.16.10.102, take the following steps:

1. Edit the inventory.ini file and the hosts.ini file, and append the node information.

   • Edit the inventory.ini file:

     [tidb_servers]
     172.16.10.4
     172.16.10.5
     172.16.10.101
     172.16.10.102
     
     [pd_servers]
     172.16.10.1
     172.16.10.2
     172.16.10.3
     
     [tikv_servers]
     172.16.10.6
     172.16.10.7
     172.16.10.8
     172.16.10.9
     
     [monitored_servers]
     172.16.10.1
     172.16.10.2
     172.16.10.3
     172.16.10.4
     172.16.10.5
     172.16.10.6
     172.16.10.7
     172.16.10.8
     172.16.10.9
     172.16.10.101
     172.16.10.102
     
     [monitoring_servers]
     172.16.10.3
     
     [grafana_servers]
     172.16.10.3

     Now the topology is as follows:

     Name	Host IP	Services
     node1	172.16.10.1	PD1
     node2	172.16.10.2	PD2
     node3	172.16.10.3	PD3, Monitor
     node4	172.16.10.4	TiDB1
     node5	172.16.10.5	TiDB2
     node101	172.16.10.101	TiDB3
     node102	172.16.10.102	TiDB4
     node6	172.16.10.6	TiKV1
     node7	172.16.10.7	TiKV2
     node8	172.16.10.8	TiKV3
     node9	172.16.10.9	TiKV4

   • Edit the hosts.ini file:

     [servers]
     172.16.10.1
     172.16.10.2
     172.16.10.3
     172.16.10.4
     172.16.10.5
     172.16.10.6
     172.16.10.7
     172.16.10.8
     172.16.10.9
     172.16.10.101
     172.16.10.102
     [all:vars]
     username = tidb
     ntp_server = pool.ntp.org

2. Initialize the newly added node.

   1. Configure the SSH mutual trust and sudo rules of the deployment machine on the central control machine:

      Copy

      ansible-playbook -i hosts.ini create_users.yml -l 172.16.10.101,172.16.10.102 -u root -k

   2. Install the NTP service on the deployment target machine:

      Copy

      ansible-playbook -i hosts.ini deploy_ntp.yml -u tidb -b

   3. Initialize the node on the deployment target machine:

      Copy

      ansible-playbook bootstrap.yml -l 172.16.10.101,172.16.10.102

   Note:

   If an alias is configured in the inventory.ini file, for example, node101 ansible_host=172.16.10.101, use -l to specify the alias when executing ansible-playbook. For example, ansible-playbook bootstrap.yml -l node101,node102. This also applies to the following steps.

3. Deploy the newly added node:

   ansible-playbook deploy.yml -l 172.16.10.101,172.16.10.102

4. Start the newly added node:

   ansible-playbook start.yml -l 172.16.10.101,172.16.10.102

5. Update the Prometheus configuration and restart the cluster:

   ansible-playbook rolling_update_monitor.yml --tags=prometheus

6. Monitor the status of the entire cluster and the newly added node by opening a browser to access the monitoring platform: http://172.16.10.3:3000.

You can use the same procedure to add a TiKV node. But to add a PD node, some configuration files need to be manually updated.

Increase the capacity of a PD node

For example, if you want to add a PD node (node103) with the IP address 172.16.10.103, take the following steps:

1. Edit the inventory.ini file and append the node information to the end of the [pd_servers] group:

   [tidb_servers]
   172.16.10.4
   172.16.10.5
   
   [pd_servers]
   172.16.10.1
   172.16.10.2
   172.16.10.3
   172.16.10.103
   
   [tikv_servers]
   172.16.10.6
   172.16.10.7
   172.16.10.8
   172.16.10.9
   
   [monitored_servers]
   172.16.10.4
   172.16.10.5
   172.16.10.1
   172.16.10.2
   172.16.10.3
   172.16.10.103
   172.16.10.6
   172.16.10.7
   172.16.10.8
   172.16.10.9
   
   [monitoring_servers]
   172.16.10.3
   
   [grafana_servers]
   172.16.10.3

   Now the topology is as follows:

   Name	Host IP	Services
   node1	172.16.10.1	PD1
   node2	172.16.10.2	PD2
   node3	172.16.10.3	PD3, Monitor
   node103	172.16.10.103	PD4
   node4	172.16.10.4	TiDB1
   node5	172.16.10.5	TiDB2
   node6	172.16.10.6	TiKV1
   node7	172.16.10.7	TiKV2
   node8	172.16.10.8	TiKV3
   node9	172.16.10.9	TiKV4

2. Initialize the newly added node:

   ansible-playbook bootstrap.yml -l 172.16.10.103

3. Deploy the newly added node:

   ansible-playbook deploy.yml -l 172.16.10.103

4. Login the newly added PD node and edit the starting script:

   {deploy_dir}/scripts/run_pd.sh

   1. Remove the --initial-cluster="xxxx" \ configuration.

      Note:

      You cannot add the # character at the beginning of the line. Otherwise, the following configuration cannot take effect.

   2. Add --join="http://172.16.10.1:2379" \. The IP address (172.16.10.1) can be any of the existing PD IP address in the cluster.

   3. Start the PD service in the newly added PD node:

      {deploy_dir}/scripts/start_pd.sh

      Note:

      Before start, you need to ensure that the health status of the newly added PD node is "true", using PD Control. Otherwise, the PD service might fail to start and an error message ["join meet error"] [error="etcdserver: unhealthy cluster"] is returned in the log.

   4. Use pd-ctl to check whether the new node is added successfully:

      ./pd-ctl -u "http://172.16.10.1:2379"

      Note:

      pd-ctl is a command used to check the number of PD nodes.

5. Start the monitoring service:

   ansible-playbook start.yml -l 172.16.10.103

   Note:

   If you use an alias (inventory_name), use the -l option to specify the alias.

6. Update the cluster configuration:

   ansible-playbook deploy.yml

7. Restart Prometheus, and enable the monitoring of PD nodes used for increasing the capacity:

   ansible-playbook stop.yml --tags=prometheus
   ansible-playbook start.yml --tags=prometheus

8. Monitor the status of the entire cluster and the newly added node by opening a browser to access the monitoring platform: http://172.16.10.3:3000.

Note:

The PD Client in TiKV caches the list of PD nodes. Currently, the list is updated only if the PD leader is switched or the TiKV server is restarted to load the latest configuration. To avoid TiKV caching an outdated list, there should be at least two existing PD members in the PD cluster after increasing or decreasing the capacity of a PD node. If this condition is not met, transfer the PD leader manually to update the list of PD nodes.

Decrease the capacity of a TiDB node

For example, if you want to remove a TiDB node (node5) with the IP address 172.16.10.5, take the following steps:

1. Stop all services on node5:

   ansible-playbook stop.yml -l 172.16.10.5

2. Edit the inventory.ini file and remove the node information:

   [tidb_servers]
   172.16.10.4
   #172.16.10.5  # the removed node
   
   [pd_servers]
   172.16.10.1
   172.16.10.2
   172.16.10.3
   
   [tikv_servers]
   172.16.10.6
   172.16.10.7
   172.16.10.8
   172.16.10.9
   
   [monitored_servers]
   172.16.10.4
   #172.16.10.5  # the removed node
   172.16.10.1
   172.16.10.2
   172.16.10.3
   172.16.10.6
   172.16.10.7
   172.16.10.8
   172.16.10.9
   
   [monitoring_servers]
   172.16.10.3
   
   [grafana_servers]
   172.16.10.3

   Now the topology is as follows:

   Name	Host IP	Services
   node1	172.16.10.1	PD1
   node2	172.16.10.2	PD2
   node3	172.16.10.3	PD3, Monitor
   node4	172.16.10.4	TiDB1
   node5	172.16.10.5	TiDB2 removed
   node6	172.16.10.6	TiKV1
   node7	172.16.10.7	TiKV2
   node8	172.16.10.8	TiKV3
   node9	172.16.10.9	TiKV4

3. Update the Prometheus configuration and restart the cluster:

   ansible-playbook rolling_update_monitor.yml --tags=prometheus

4. Monitor the status of the entire cluster by opening a browser to access the monitoring platform: http://172.16.10.3:3000.

Decrease the capacity of a TiKV node

For example, if you want to remove a TiKV node (node9) with the IP address 172.16.10.9, take the following steps:

1. Remove the node from the cluster using pd-ctl:

   1. View the store ID of node9:

      ./pd-ctl -u "http://172.16.10.1:2379" -d store

   2. Remove node9 from the cluster, assuming that the store ID is 10:

      ./pd-ctl -u "http://172.16.10.1:2379" -d store delete 10

2. Use pd-ctl to check whether the node is successfully removed:

   ./pd-ctl -u "http://172.16.10.1:2379" -d store 10

   Note:

   It takes some time to remove the node. If the status of the node you remove becomes Tombstone, then this node is successfully removed.

3. After the node is successfully removed, stop the services on node9:

   ansible-playbook stop.yml -l 172.16.10.9

4. Edit the inventory.ini file and remove the node information:

   [tidb_servers]
   172.16.10.4
   172.16.10.5
   
   [pd_servers]
   172.16.10.1
   172.16.10.2
   172.16.10.3
   
   [tikv_servers]
   172.16.10.6
   172.16.10.7
   172.16.10.8
   #172.16.10.9  # the removed node
   
   [monitored_servers]
   172.16.10.4
   172.16.10.5
   172.16.10.1
   172.16.10.2
   172.16.10.3
   172.16.10.6
   172.16.10.7
   172.16.10.8
   #172.16.10.9  # the removed node
   
   [monitoring_servers]
   172.16.10.3
   
   [grafana_servers]
   172.16.10.3

   Now the topology is as follows:

   Name	Host IP	Services
   node1	172.16.10.1	PD1
   node2	172.16.10.2	PD2
   node3	172.16.10.3	PD3, Monitor
   node4	172.16.10.4	TiDB1
   node5	172.16.10.5	TiDB2
   node6	172.16.10.6	TiKV1
   node7	172.16.10.7	TiKV2
   node8	172.16.10.8	TiKV3
   node9	172.16.10.9	TiKV4 removed

5. Update the Prometheus configuration and restart the cluster:

   ansible-playbook rolling_update_monitor.yml --tags=prometheus

6. Monitor the status of the entire cluster by opening a browser to access the monitoring platform: http://172.16.10.3:3000.

Decrease the capacity of a PD node

For example, if you want to remove a PD node (node2) with the IP address 172.16.10.2, take the following steps:

1. Remove the node from the cluster using pd-ctl:

   1. View the name of node2:

      ./pd-ctl -u "http://172.16.10.1:2379" -d member

   2. Remove node2 from the cluster, assuming that the name is pd2:

      ./pd-ctl -u "http://172.16.10.1:2379" -d member delete name pd2

2. Use Grafana or pd-ctl to check whether the node is successfully removed:

   ./pd-ctl -u "http://172.16.10.1:2379" -d member

3. After the node is successfully removed, stop the services on node2:

   ansible-playbook stop.yml -l 172.16.10.2

   Note:

   In this example, you can only stop the PD service on node2. If there are any other services deployed with the IP address 172.16.10.2, use the -t option to specify the service (such as -t tidb).

4. Edit the inventory.ini file and remove the node information:

   [tidb_servers]
   172.16.10.4
   172.16.10.5
   
   [pd_servers]
   172.16.10.1
   #172.16.10.2  # the removed node
   172.16.10.3
   
   [tikv_servers]
   172.16.10.6
   172.16.10.7
   172.16.10.8
   172.16.10.9
   
   [monitored_servers]
   172.16.10.4
   172.16.10.5
   172.16.10.1
   #172.16.10.2  # the removed node
   172.16.10.3
   172.16.10.6
   172.16.10.7
   172.16.10.8
   172.16.10.9
   
   [monitoring_servers]
   172.16.10.3
   
   [grafana_servers]
   172.16.10.3

   Now the topology is as follows:

   Name	Host IP	Services
   node1	172.16.10.1	PD1
   node2	172.16.10.2	PD2 removed
   node3	172.16.10.3	PD3, Monitor
   node4	172.16.10.4	TiDB1
   node5	172.16.10.5	TiDB2
   node6	172.16.10.6	TiKV1
   node7	172.16.10.7	TiKV2
   node8	172.16.10.8	TiKV3
   node9	172.16.10.9	TiKV4

5. Update the cluster configuration:

   ansible-playbook deploy.yml

6. Restart Prometheus, and disable the monitoring of PD nodes used for increasing the capacity:

   ansible-playbook stop.yml --tags=prometheus
   ansible-playbook start.yml --tags=prometheus

7. To monitor the status of the entire cluster, open a browser to access the monitoring platform: http://172.16.10.3:3000.

Note:

The PD Client in TiKV caches the list of PD nodes. Currently, the list is updated only if the PD leader is switched or the TiKV server is restarted to load the latest configuration. To avoid TiKV caching an outdated list, there should be at least two existing PD members in the PD cluster after increasing or decreasing the capacity of a PD node. If this condition is not met, transfer the PD leader manually to update the list of PD nodes.