tencent cloud

This document describes how to migrate data when removing core nodes tservers from a Kudu cluster.

Note：

Core nodes can be removed from a Kudu cluster. This feature is not available by default. If you need to use it, submit a ticket for application.

Before disconnecting tservers, you can use the rebalancing tool for data migration. Note that you can disconnect only one tserver at a time. To disconnect multiple ones, repeat the following steps.

Kudu migration based on rebalancing tool

1. Make sure that the cluster status is OK.

   /usr/local/service/kudu/bin/kudu cluster ksck 10.0.1.29:7051,10.0.1.16:7051,10.0.1.36:7051
   

2. Run the ksck command in step 1 to get the uid of the disconnected nodes.

Take the fb9afb1b2989456cac5800bf6990dfea node as an example.
3. Switch the fb9afb1b2989456cac5800bf6990dfea node to the maintenance mode.

/usr/local/service/kudu/bin/kudu tserver state enter_maintenance 10.0.1.29:7051,10.0.1.16:7051,10.0.1.36:7051 fb9afb1b2989456cac5800bf6990dfea

4. Run the following rebalance command.

   /usr/local/service/kudu/bin/kudu cluster rebalance 10.0.1.29:7051,10.0.1.16:7051,10.0.1.36:7051 --ignored_tservers fb9afb1b2989456cac5800bf6990dfea --move_replicas_from_ignored_tservers
   

After the command is executed, run the ksck command again to check whether the status is OK before proceeding.
5. Suspend the tserver process at 10.0.1.45 on the fb9afb1b2989456cac5800bf6990dfea node. Note that at this point, if you run the ksck command, you will see that the cluster is unhealthy, and you need to restart the tmasters.

6. Restart the masters in the EMR console one by one (rolling restart in the console is not recommended). Then, run the ksck command to check whether the cluster is healthy.