@Rocky when adding a new BE node, how can I know once the tablet distribution is stabilized and finished?

After adding a new BE node, StarRocks automatically schedules tablet rebalancing. You can verify the status and stabilization using the following methods: 1. Check the Cluster Balance Status The most direct way to see if the cluster has reached a balanced state is using the

SHOW PROC

command:

sql
SHOW PROC '/cluster_balance/balance_stat';

* Balanced: This column will show

true

once the cluster has finished rebalancing for that specific storage medium and type (e.g.,

inter-node tablet distribution

). * PendingTablets / RunningTablets: These columns indicate how many tablets are currently waiting or being moved. 2. Monitor Tablet Scheduling Tasks You can view the details of active data movement tasks: * Running tasks:

SHOW PROC '/cluster_balance/running_tablets';

* Pending tasks:

SHOW PROC '/cluster_balance/pending_tablets';

* History:

SHOW PROC '/cluster_balance/history_tablets';

If the running and pending lists are empty, the rebalancing process has likely finished or paused because it reached the cluster's load balance threshold. 3. Check Overall Backend Stats To see the distribution of data across all BE nodes, run:

sql
SHOW BACKENDS;

Look at the TabletNum and DataUsedCapacity columns. As the cluster rebalances, you will see these numbers gradually increase on the new BE node and decrease on the existing ones until they are roughly equal. 4. Check Global Cloning Status If you want a high-level view of how many tablets are currently being copied across the entire cluster:

sql
SHOW PROC '/statistic';

The CloningTabletNum column shows the number of tablets currently undergoing cloning/migration. When this reaches

0

, no active tablet movement is occurring. Summary Checklist Your distribution is considered stabilized when: 1.

SHOW PROC '/cluster_balance/balance_stat';

shows

Balanced = true

. 2.

SHOW PROC '/cluster_balance/running_tablets';

is empty. 3.

TabletNum

in

SHOW BACKENDS;

is distributed evenly across all nodes. References • sql-reference/sql-statements/cluster-management/nodes_processes/SHOW_PROC.md • administration/management/resource_management/Replica.md