Clusters and cluster groups#
Starburst control plane increases the workload capacity of your deployment by distributing queries across multiple Starburst Enterprise platform (SEP) clusters.
How load balancing works#
When a cluster group has two or more active and healthy SEP clusters, control plane sends each query to one of them. It does this automatically, and you do not need to change your applications.
By default, control plane holds each statement in a statement queue until a cluster in the group can run it. Control plane picks the cluster among those that are below their query limit, by using the load balancing strategy that you set:
RANDOM, the default, picks a cluster at random.QUERY_COUNT_BASEDpicks the cluster with the lowest load, which is the number of queries that run on it. If you turn off the statement queue, queries can also queue on the cluster, and control plane counts those too.
If every cluster in the group is at its limit, the statement waits in the queue. See Smart load balancing.
In standalone mode, the statement queue is off, so control plane sends each statement straight to a cluster.
A query that matches no routing rule goes to the default cluster group. If the default cluster group has no active and healthy cluster, the query fails. To send other queries to a different cluster group, define routing rules. See Routing.
Clusters pane#
Use the Clusters pane to see and manage the clusters and cluster groups that control plane routes queries to. The pane has two tabs:
The Cluster groups tab lists each cluster group with its name, its number of clusters, and the concurrency limit for each cluster. To work with cluster groups, see Manage cluster groups.
The Clusters tab lists each cluster with its status, its cluster group, and its running and queued queries. To add and manage clusters, see Manage clusters. For the status values, see Monitor cluster status.