tiup集群缩小

tiup cluster scale-in command is used to scale down the cluster, which means that the service goes offline. The specified node will eventually be removed from the cluster and the remaining related files will be deleted.

Offline Special Handling

Because the offline of TiKV, TiFlash and TiDB Binlog components is asynchronous (the removal operation needs to be performed through the API first) and the offline process takes a long time (it needs to continuously observe whether the node has been successfully offline), so TiKV, TiFlash and TiDB Binlog components do special treatment:

• Operations to TiKV, TiFlash and TiDB Binlog components:
  • tiup-cluster exits directly after taking it offline through the API without waiting for the offline to complete
  • Execute tiup cluster display to check the status of the offline node and wait for its status to change to Tombstone
  • Execute the tiup cluster prune command to clean up the Tombstone nodes, which does the following:
    • Stop the service of the node that has gone offline
    • Clean up related data files of nodes that have been offline
    • Update the topology of the cluster and remove nodes that have been offline
• Operations to other components
  • When the PD component is offline, the specified node will be deleted from the cluster through the API (this process is very fast), and then the service of the specified PD will be stopped and the relevant data files of the node will be cleared
  • When other components are offline, directly stop and clear the related data files of the node

Syntax

tiup cluster scale-in [flags]

is the name of the cluster to be operated. If you forget the cluster name, you can pass Cluster ListView.

options

-N, --node (strings, no default value, must be non-null)

Select the node to be scaled down. If multiple nodes are scaled down, separate them with commas.

--force

• In some cases, the host of the node that may be scaled down has been down, resulting in the inability to connect to the node through SSH for operation. At this time, you can use the --force option to force itRemove from the cluster.
• Data type: BOOLEAN
• This option is disabled by default, and the default value is false.Add this option to the command and pass in the true value or no value to enable this function.

Warning

When using this option to forcefully remove TiKV / TiFlash nodes that are in service and offline, these nodes will be deleted directly without waiting for data scheduling to complete, so there is a high risk of data loss in this scenario.This option is not recommended for nodes that are not down.

--transfer-timeout (uint, default 300)

When shrinking PD or TiKV, the leader of the shrinking node will be migrated to other nodes first. The migration process will take a certain time. You can set the longest time by setting --transfer-timeoutWaiting time (in seconds). After the timeout, waiting for the direct scaling service will be skipped.

Note

If there is a situation of skipping and waiting for direct scaling, the service performance may experience jitter.

-h, --help

• Output help information.
• Data type: BOOLEAN
• This option is disabled by default, and the default value is false.Add this option to the command and pass in the true value or no value to enable this function.

output

Scale down log