Cluster Operations

Distributed Tables

• CREATE TABLE dist ENGINE = Distributed(cluster, db, local_table, sharding_key)
• Sharding key: rand() for even distribution, cityHash64(user_id) for user affinity
• Reads: query all shards in parallel; Writes: route to correct shard or write locally

ON CLUSTER DDL

• ALTER TABLE t ON CLUSTER '{cluster}' ADD COLUMN col Type — propagate schema to all replicas
• CREATE TABLE t ON CLUSTER '{cluster}' AS template_db.template_table — clone across shards
• distributed_ddl_output_mode: throw (fail on error), null (ignore), none, active
• Check status: SELECT * FROM system.distributed_ddl_queue

Load Balancing and Read Routing

• load_balancing: random, in_order, first_or_random, nearest_hostname
• max_replica_delay_for_distributed_queries — skip lagging replicas
• fallback_to_stale_replicas_for_distributed_queries=1 — use stale when all delayed