Online Scale-Out and Cloud Data Ingestion
summary
Uses Azure Blob Storage and the Citus extension for Azure Cosmos DB for PostgreSQL to ingest a compressed CSV directly into Postgres and loads ~3,000,000 rows via COPY/BlobGet. Shows Create distributed table concurrently (Citus 11.11) that avoids blocking writes, and moves a hot tenant shard (shard 102093 ~78MB → new shard 102110 ~74MB) online while inserts continue. Introduces a background rebalancer job that evens worker sizes (69/151MB → 73/146MB) without holding a client connection.
AI-written summary · about
chapters (7)
- 0:00Scale-out Citus clusters
- 1:09Azure Blob Storage ingestion
- 4:44Distributing tables without blocking writes
- 7:11Online table distribution
- 9:42Detecting skewed data distribution
- 11:52Moving large tenants non-blocking
- 13:45Rebalancing shards in Citus 11.1
related talks
24mCitus from the Customer eyes
52mBuilding Distributed OLTP with Azure Cosmos DB for PostgreSQL
32mAuto scaling Azure Cosmos DB for PostgreSQL with Citus, Grafana & Azure Serverless
31mKEYNOTE: The Distributed PostgreSQL Problem & How Citus Solves it
27mMulti-tenant Saas apps made simple on Azure Cosmos DB for PostgreSQL
44mMulti-tenant SaaS apps made simple on Azure Cosmos DB for PostgreSQL
20mLessons learned scaling our SaaS on Postgres to 8+ billion events
32m8 keys to the growth of Citus & Distributed PostgreSQL
