How We Built a Cloud‑Native Distributed PostgreSQL + Citus Cluster for Sentry Backend

Distributed Storage Cluster (Longhorn)

Longhorn provides persistent storage for the Citus coordinator and worker nodes.

https://longhorn.io/

Distributed PostgreSQL Cluster (Citus)

Citus shards the large Sentry event source table nodestore_node.

https://docs.citusdata.com/en/v11.1/

Read‑Write Splitting and High Availability (PgPool + repmgr)

PgPool and repmgr implement read‑write splitting and primary‑standby high availability for the Citus coordinator and worker nodes.

https://www.pgpool.net/docs/pgpool-II-4.2.3/en/html/example-kubernetes.html

https://repmgr.org/

Cluster Management (PgAdmin)

PgAdmin is used to manage the PostgreSQL cluster.

https://www.pgadmin.org/

nodestore_node Table Sharding

A distributed key is selected and the table is converted to a distributed table with 64 shards evenly distributed across six worker nodes.

# Create distributed table
SELECT create_distributed_table('nodestore_node', 'id', colocate_with => 'none', shard_count => 64);
# Rebalance shards
SELECT rebalance_table_shards();
# List shards
SELECT * FROM citus_shards;

Conclusion

Middleware cluster infrastructure involves many details and can be treated as a separate domain; the described setup illustrates one practical approach.