Can One PostgreSQL Instance Power 800 Million Users? OpenAI’s Trade‑offs Unpacked

OpenAI runs a single primary PostgreSQL with about 50 replicas to serve roughly 800 million ChatGPT users, employing PgBouncer connection pooling, cache‑lock mechanisms, query and schema optimizations, workload isolation, multi‑layer throttling, and cascade replication, while acknowledging the architecture’s limits and the complexity of full sharding.

Azure CosmosDBCascade ReplicationOpenAI

0 likes · 8 min read

Can One PostgreSQL Instance Power 800 Million Users? OpenAI’s Trade‑offs Unpacked

Aikesheng Open Source Community

Apr 16, 2024 · Databases

MySQL Cascade Replication Failure Caused by Duplicate server_id and Its Resolution

The article analyzes a MySQL cascade replication issue where the new slave failed to sync due to duplicate server_id values, explains the root cause stemming from automated server_id assignment, and provides step‑by‑step remediation to ensure unique server_id settings and reliable replication.

Cascade Replicationmysqlserver_id

0 likes · 6 min read

MySQL Cascade Replication Failure Caused by Duplicate server_id and Its Resolution

Aikesheng Open Source Community

Apr 18, 2023 · Databases

Schema Change for a 400‑Million‑Row MySQL Table Using Cascade Replication

This article details a step‑by‑step strategy for expanding a VARCHAR column from 50 to 500 characters on a 4‑hundred‑million‑row MySQL 8.0 table, evaluating online DDL, gh‑ost, pt‑osc, and ultimately implementing a cascade replication workflow to meet a four‑hour downtime window.

Cascade ReplicationOnline DDLSchema Change

0 likes · 10 min read

Schema Change for a 400‑Million‑Row MySQL Table Using Cascade Replication

Can One PostgreSQL Instance Power 800 Million Users? OpenAI’s Trade‑offs Unpacked

MySQL Cascade Replication Failure Caused by Duplicate server_id and Its Resolution

Schema Change for a 400‑Million‑Row MySQL Table Using Cascade Replication

Can One PostgreSQL Instance Power 800 Million Users? OpenAI’s Trade‑offs Unpacked