Home/Cases/PICC Group
PICC Group
Finance / PostgreSQL Smooth Upgrade

PICC PostgreSQL Domestic Migration and Smooth Upgrade Case Study

For a PostgreSQL estate with more than 1,000 instances and hundreds of clusters, Highgo provided seconds-level in-place upgrades, reverse synchronization, and financial-grade rollback paths to support a near-seamless domestic database transformation.

1,000+PostgreSQL database instances across the group
SecondsIn-place upgrade and rollback window
<30 sRTO for primary-standby failover
Background

Project Background

PICC Group grew from the People’s Insurance Company of China, founded in 1949, into a comprehensive insurance and financial group with operations spanning property insurance, life insurance, reinsurance, asset management, and insurance technology.

PICC Group and its subsidiaries extensively use PostgreSQL databases for core systems such as underwriting, claims processing, policy management, 95518 customer service, loss reporting, channel management, sales management, CIF customer data management, and master data management.

Challenges

Challenges and Requirements

Business Challenges
  • Production systems used many PostgreSQL versions, including deployments on IBM LinuxOne, so the replacement platform needed strong compatibility with PostgreSQL features, plugins, and components.
  • Core insurance systems required 24x7 operation, making the cutover window extremely small and requiring a detailed, verifiable rollback plan.
  • Traditional import and export migration would scale with data volume, making it unsuitable for a near-seamless financial production upgrade.
  • Some workloads had already moved to a domestic distributed database but suffered from cross-database join performance issues, requiring a return path to a domestic centralized database.
Core Requirements
  • Support smooth in-place upgrades from different PostgreSQL versions to Highgo Database.
  • Remain compatible with native JDBC and LIBPQ drivers, PostgreSQL extensions, replication tools, development tools, and HA scripts.
  • Provide reverse data synchronization and parallel application execution for financial-grade emergency escape paths.
  • Keep application code unchanged while maintaining production performance on the domestic platform.
Solution

Implementation Approach

Metro-ready primary-standby architecture

Highgo provided single-data-center or same-city disaster-recovery cluster patterns with at least one strictly synchronous node among the replicas, ensuring RPO 0 and VIP-based failover with RTO within 30 seconds.

Self-developed upgrade and rollback tooling

The pg2highgo and highgo2pg tools enable seconds-level in-place upgrades and rollback between PostgreSQL and Highgo Database, supplemented by logical-replication rollback paths for different scenarios.

Application-transparent migration

Highgo supports native PostgreSQL JDBC and LIBPQ drivers plus native extensions, allowing applications to move without code changes while preserving performance for workloads such as pg_bigm-based fuzzy search.

Centralized rollback from distributed systems

The centralized Highgo architecture uses single-node large-capacity capability to avoid cross-node access and distributed-transaction performance problems found in the prior distributed database path.

Integrated operations platform

Highgo supplied cluster deployment, component management, backup and recovery, monitoring, alerting, manual switchover, and failover management to simplify database operations.

Highlights

Project Highlights

  1. Self-developed database upgrade tools support in-place upgrades from multiple PostgreSQL versions to Highgo Database.
  2. In-place upgrade can complete in seconds regardless of data volume, while reverse data synchronization and parallel application execution provide a financial-grade rollback path.
  3. Highgo supports native PostgreSQL JDBC and LIBPQ drivers plus extension plugins, enabling migration without application code changes.
  4. The solution remains compatible with production tools and scripts including Debezium, DataPipeline, pglogical, customer-developed pglogical tooling, DataGrid, and Patroni.
  5. Highgo worked with PICC Property Insurance, PICC Technology, and multiple development centers to complete simulation migration testing across business systems.
Value

Customer Value

  1. Delivered seamless PostgreSQL replacement with low-disruption business migration for systems that require 24x7 continuity.
  2. Enabled rollback from problematic distributed database deployments to a domestic centralized database when cross-node joins hurt performance.
  3. Reduced migration risk through dual-track risk control, financial-grade rollback paths, and strong PostgreSQL compatibility.
  4. Kept performance degradation below 10 percent on a fully domestic platform while improving operations efficiency through integrated management.
More Cases

Related Case Studies

View All Cases
Finance

OCBC Bank Batch Processing Case Study

Highgo replaced a foreign commercial database for mission-critical banking batch workloads with metro-active and remote DR architecture.

Energy / Power

Power Grid Automation Database Localization Project

Highgo supports unattended power-station scenarios with one-click deployment, two-node HA, and double-write synchronization architectures.

Public Service

Jinan Provident Fund Core System Case Study

The project replaces Oracle and DB2 for a large provident-fund platform, balancing data scale, transactional consistency, and metro disaster recovery.