NewtGlobal Platinum Sponsor Session: Oracle-to-PostgreSQL Migration Validation with Agentic AI
NewtGlobal Platinum Sponsor Session: Oracle-to-PostgreSQL Migration Validation with Agentic AI
Speaker: Neeta Goel (Newt Global)
🔗 https://live.pgconf.in/events/pgconf-...
PGConf India 2026 — Asia's Premier PostgreSQL Conference
March 11-13, 2026 | Sheraton Grand Hotel, Bangalore
https://pgconf.in
---
We will conduct an intermediate-level technical session titled:
“Oracle-to-PostgreSQL Migration Validation with Agentic AI”
Session Overview
Testing is the largest bottleneck in Oracle-to-PostgreSQL migrations, driving long timelines and heavy dependence on application SMEs. Many migrations stall during validation because automation stops at conversion, leaving teams to manually certify that schema objects, database code, and application transactions behave correctly on PostgreSQL.
In this session, speakers present an Agentic AI–based validation approach that automates migration certification without requiring teams to become functional experts in the application. The approach uses a multi-agent validation that plans test coverage, generates targeted checks, executes verification steps, and iteratively repairs or escalates issues—backed by deterministic compile, lint, and object-level gates to prevent “AI guesswork” from entering production.
The session will walk through an agent-orchestrated validation framework:
Agentic Validation Fabric (How it Works)
• Inventory Agent builds a dependency-aware map of Oracle and PostgreSQL objects (tables, types, constraints, indexes, views, triggers, functions, procedures) and identifies migration risk hotspots (packages, dynamic SQL, bulk operations).
• Compilation Agent validates object creation order and dependency resolution, enforcing “must compile” gates for converted PL/pgSQL and migrated DDL.
• Lint & Semantic Analysis Agent runs PostgreSQL-aware linting to catch defects that compile but are semantically incorrect (e.g., shadowed variables, unsafe casts, invalid SQL in routines, unreachable branches).
• Object-Specific Check Agents generate and execute targeted validations for Oracle constructs that require refactoring in PostgreSQL—such as packages, package global variables, GOTO statements, exception semantics, cursor patterns, and dynamic SQL rewrites—covering gaps that traditional PG_Check-style scans may miss.
• Application SQL Compatibility Agent detects and verifies application embedded SQL changes needed for PostgreSQL syntax and behavior (type mappings, naming changes, sequence semantics, NULL/empty-string differences, date/time behavior).
Transaction Tracing + Replay (End-to-End Certainty)
To validate real application behavior—not just schema correctness—the approach introduces an Application Tracing Agent that captures live transaction flows from the application: executed SQL, bind variables, execution sequence, key input parameters, row-count checkpoints, and selected result signatures. A Replay Agent then re-runs these captured “data points” against the target PostgreSQL environment and compares outcomes to expected behavior, surfacing regressions that schema-only testing cannot detect.
Outcome
By combining agentic planning and execution with hard verification gates (compile, lint, and object checks) and app-level trace/replay validation, teams can reduce SME dependency, shorten validation cycles, and materially lower migration risk—enabling smoother, more reliable migrations to Azure Database for PostgreSQL.
Level: Intermediate
Audience: QA engineers, DB architects, and platform leaders managing distributed systems and large-scale database migrations
Speakers: Neeta Goel, Himanshu