Skip to content

Document 01 — Validation Master Plan (VMP)

Document IDVersionEffective Date (UTC)System NameRegulatory ScopeStatus
VMP-GXPHISTORIAN-2026-0011.02026-10-17T20:21:45ZGxPHistorian GxP Industrial Historian (ClickHouse + PostgreSQL Part 11 Control Plane)21 CFR Part 11, EU GMP Annex 11, ISPE GAMP 5 (2nd Ed.), FDA CSAAPPROVED

1. Purpose & Regulatory Scope

This Validation Master Plan (VMP) establishes the Computerized System Validation (CSV) and Computer Software Assurance (CSA) governance framework for the GxPHistorian GxP Industrial Historian deployed at BioCorp Fremont Biomanufacturing Operations. The platform replaces legacy proprietary OSIsoft PI System components with a 4-Pillar open-source architecture engineered for 21 CFR Part 11, EU EudraLex Volume 4 Annex 11, and ALCOA+ data integrity compliance.

2. System Classification (ISPE GAMP 5 Second Edition)

Subsystem PillarComponentGAMP 5 CategoryValidation Strategy
Pillar 1: Edge CollectorTelegraf 1.32 + Disk WAL + Python Deadband WorkerCategory 4 / Category 5Automated IQ + Scripted OQ on Stage 1 Exception/Deadband logic & WAL persistence
Pillar 2: Telemetry ArchiveClickHouse Server 24.8 LTS (MergeTree / AggregatingMergeTree)Category 4Automated IQ on lossless codecs (DoubleDelta, Gorilla) + Scripted OQ on quality >= 192 filtering, LTTB, and interpolation
Pillar 3: Control PlanePostgreSQL 16 + FastAPI (historian-api)Category 4 / Category 5Scripted OQ on PL/pgSQL Part 11 WORM audit triggers, SHA-256 hash chaining, and §11.200 electronic signatures
Pillar 4: Operator HMIGrafana OSS 11.2 + Excel DataLink UDF/Power QueryCategory 4Automated IQ on provisioned datasources/dashboards + OQ on DataLink CSV/JSON parity

3. Roles & Responsibilities

RoleDesignated PersonnelGxP Responsibilities
System OwnerDirector of Manufacturing IT/OTAccountable for system availability, access control, backup/restore drills, and GitOps change control.
Process OwnerHead of Upstream & Downstream MSATDefines User Requirements (URS), engineering units, alarm limits (low_limit, high_limit), and batch Event Frame templates.
Quality Assurance (QA)Lead CSV / Part 11 Qualified Person (qa_approver_01)Approves VMP, URS, Risk Assessment, executes/reviews IQ/OQ evidence, and signs batch/excursion Event Frames.
Validation EngineeringSenior CSV Automation EngineerAuthors and maintains executable pytest IQ/OQ qualification suites and generates the cryptographic validation binder.

4. Validation Lifecycle (IQ → OQ → PQ)

  1. Installation Qualification (IQ): Automated verification (test_01_iq_infrastructure.py) of container health, non-default credentials, PostgreSQL extensions (ltree, btree_gist, pgcrypto), ClickHouse lossless compression codecs, and dictionary bridging.
  2. Operational Qualification (OQ): Scripted positive and adversarial negative testing (test_02_oq_part11_compliance.py & test_03_oq_historian_core.py) verifying 21 CFR Part 11 WORM immutability, SHA-256 hash chain verification, electronic signature credential challenges, bad quality (quality < 192) rejection, linear vs. LOCF step interpolation, and LTTB peak preservation.
  3. Performance Qualification (PQ): Continuous multi-protocol plant telemetry ingestion across 5 GxP tags (101, 102, 103, 301, 401), live Event Frame excursion capture, and Apache Parquet lakehouse exports.

5. Backup, Restore, Disaster Recovery & GitOps Change Control SOP

  • Database & Telemetry Backups: PostgreSQL pi_control_plane (including part11_audit_log) undergoes daily logical pg_dump and continuous WAL archiving to immutable object storage. ClickHouse pi_historian partitions (PARTITION BY toYYYYMM(ts)) are exported to Snappy-compressed Apache Parquet files and backed up via native ClickHouse BACKUP TABLE snapshots.
  • GitOps Change Control (SOP-IT-CHG-014): All infrastructure definitions (docker-compose.yml), database schemas (init-postgres/, init-clickhouse/), and API code (app/main.py) are version-controlled in Git. Any change requires a pull request, automated execution of the 15-test IQ/OQ pytest suite (100% pass gate), and QA electronic approval prior to production deployment.