Backup & Restore

Overview

VIGIA’s multi-tenant architecture requires careful backup and restore procedures. Each tenant database must be backed up independently, while the master database contains critical tenant registry information.

Backup Strategy

Database Hierarchy

VIGIA uses a three-tier database structure:

Master Database: Tenant registry (clientes_saas table)
Tenant Databases: Individual client data (one per tenant)
Legacy Database: Single-tenant deployments (optional)

Backup Frequency Recommendations

Database Type	Frequency	Retention	Priority
Master	Hourly	30 days	Critical
Active Tenants	Daily	90 days	High
Demo Tenants	Weekly	30 days	Medium
Legacy	Daily	90 days	High

PostgreSQL Backup Methods

pg_dump (Logical Backup)

Logical backups export database contents as SQL statements.

Backup Single Tenant

# Basic backup
pg_dump -h localhost -U postgres -d vigia_empresa1 > backup_empresa1_$(date +%Y%m%d).sql

# Compressed backup (recommended)
pg_dump -h localhost -U postgres -d vigia_empresa1 | gzip > backup_empresa1_$(date +%Y%m%d).sql.gz

# Custom format (supports parallel restore)
pg_dump -h localhost -U postgres -Fc -d vigia_empresa1 -f backup_empresa1_$(date +%Y%m%d).dump

# Include CREATE DATABASE statement
pg_dump -h localhost -U postgres -C -d vigia_empresa1 > backup_empresa1_$(date +%Y%m%d).sql

Backup Master Database

# Master database backup (critical!)
pg_dump -h localhost -U postgres -Fc -d vigia_master -f backup_master_$(date +%Y%m%d).dump

# With verbose output
pg_dump -h localhost -U postgres -Fc -v -d vigia_master -f backup_master_$(date +%Y%m%d).dump

Backup All Tenants (Script)

#!/bin/bash
# backup_all_tenants.sh

BACKUP_DIR="/var/backups/vigia"
DATE=$(date +%Y%m%d_%H%M%S)
HOST="localhost"
USER="postgres"

# Create backup directory
mkdir -p "$BACKUP_DIR/$DATE"

# Backup master database
echo "Backing up master database..."
pg_dump -h "$HOST" -U "$USER" -Fc -d vigia_master \
  -f "$BACKUP_DIR/$DATE/master.dump"

# Get list of tenant databases
TENANTS=$(psql -h "$HOST" -U "$USER" -d vigia_master -t -c \
  "SELECT db_name FROM clientes_saas WHERE is_active = true;")

# Backup each tenant
for tenant in $TENANTS; do
  echo "Backing up $tenant..."
  pg_dump -h "$HOST" -U "$USER" -Fc -d "$tenant" \
    -f "$BACKUP_DIR/$DATE/$tenant.dump"
done

# Create manifest
ls -lh "$BACKUP_DIR/$DATE/" > "$BACKUP_DIR/$DATE/manifest.txt"

echo "Backup completed: $BACKUP_DIR/$DATE"

pg_basebackup (Physical Backup)

Physical backups copy the entire PostgreSQL data directory.

# Full cluster backup
pg_basebackup -h localhost -U postgres -D /var/backups/postgres_base -Ft -z -P

# With WAL files for point-in-time recovery
pg_basebackup -h localhost -U postgres -D /var/backups/postgres_base \
  -Ft -z -P -X stream

Use Cases:

Full cluster recovery
Replication setup
Disaster recovery
Large databases (faster than pg_dump)

Automated Backup Scripts

Daily Backup with Rotation

#!/bin/bash
# /usr/local/bin/vigia_daily_backup.sh

set -e

BACKUP_DIR="/var/backups/vigia"
RETENTION_DAYS=90
DATE=$(date +%Y%m%d)
HOST="localhost"
USER="postgres"
LOG_FILE="/var/log/vigia_backup.log"

log() {
    echo "[$(date +'%Y-%m-%d %H:%M:%S')] $1" | tee -a "$LOG_FILE"
}

mkdir -p "$BACKUP_DIR"

log "Starting VIGIA backup..."

# Backup master
log "Backing up master database..."
pg_dump -h "$HOST" -U "$USER" -Fc -d vigia_master \
  -f "$BACKUP_DIR/master_$DATE.dump" 2>> "$LOG_FILE"

# Query active tenants from master
log "Querying active tenants..."
TENANTS=$(psql -h "$HOST" -U "$USER" -d vigia_master -t -c \
  "SELECT subdominio FROM clientes_saas WHERE is_active = true;" 2>> "$LOG_FILE")

# Backup each tenant
for tenant_sub in $TENANTS; do
    db_name="vigia_${tenant_sub}"
    log "Backing up tenant: $db_name"
    
    pg_dump -h "$HOST" -U "$USER" -Fc -d "$db_name" \
      -f "$BACKUP_DIR/${db_name}_$DATE.dump" 2>> "$LOG_FILE" || {
        log "ERROR: Failed to backup $db_name"
        continue
    }
done

# Remove old backups
log "Removing backups older than $RETENTION_DAYS days..."
find "$BACKUP_DIR" -name "*.dump" -mtime +"$RETENTION_DAYS" -delete

log "Backup completed successfully"

# Optional: Upload to S3/cloud storage
# aws s3 sync "$BACKUP_DIR" s3://my-vigia-backups/

Cron Configuration

# Edit crontab
sudo crontab -e

# Add daily backup at 2 AM
0 2 * * * /usr/local/bin/vigia_daily_backup.sh

# Hourly master backup
0 * * * * pg_dump -h localhost -U postgres -Fc vigia_master -f /var/backups/vigia/master_hourly_$(date +\%H).dump

Restore Procedures

Restore Single Tenant

From SQL Dump

# 1. Drop existing database (if needed)
psql -h localhost -U postgres -c "SELECT pg_terminate_backend(pid) FROM pg_stat_activity WHERE datname = 'vigia_empresa1';"
psql -h localhost -U postgres -c "DROP DATABASE IF EXISTS vigia_empresa1;"

# 2. Create fresh database
psql -h localhost -U postgres -c "CREATE DATABASE vigia_empresa1;"

# 3. Restore from SQL
psql -h localhost -U postgres -d vigia_empresa1 < backup_empresa1_20240303.sql

# Or from compressed
gunzip -c backup_empresa1_20240303.sql.gz | psql -h localhost -U postgres -d vigia_empresa1

From Custom Format

# Restore using pg_restore
pg_restore -h localhost -U postgres -d vigia_empresa1 -c backup_empresa1_20240303.dump

# Parallel restore (faster for large databases)
pg_restore -h localhost -U postgres -d vigia_empresa1 -j 4 backup_empresa1_20240303.dump

# Verbose output
pg_restore -h localhost -U postgres -d vigia_empresa1 -v backup_empresa1_20240303.dump

# Create database and restore
pg_restore -h localhost -U postgres -C -d postgres backup_empresa1_20240303.dump

Restore Master Database

CRITICAL: Restoring master database affects all tenant references.

# 1. Stop application
systemctl stop vigia-backend

# 2. Terminate connections
psql -h localhost -U postgres -c \
  "SELECT pg_terminate_backend(pid) FROM pg_stat_activity WHERE datname = 'vigia_master';"

# 3. Drop and recreate
psql -h localhost -U postgres -c "DROP DATABASE IF EXISTS vigia_master;"
psql -h localhost -U postgres -c "CREATE DATABASE vigia_master;"

# 4. Restore
pg_restore -h localhost -U postgres -d vigia_master backup_master_20240303.dump

# 5. Verify
psql -h localhost -U postgres -d vigia_master -c "SELECT count(*) FROM clientes_saas;"

# 6. Restart application
systemctl start vigia-backend

Disaster Recovery (Full System)

#!/bin/bash
# restore_full_system.sh

BACKUP_DATE="20240303"
BACKUP_DIR="/var/backups/vigia/$BACKUP_DATE"
HOST="localhost"
USER="postgres"

set -e

echo "Starting full system restore from $BACKUP_DATE..."

# Stop application
echo "Stopping VIGIA backend..."
systemctl stop vigia-backend

# Restore master
echo "Restoring master database..."
psql -h "$HOST" -U "$USER" -c "DROP DATABASE IF EXISTS vigia_master;"
psql -h "$HOST" -U "$USER" -c "CREATE DATABASE vigia_master;"
pg_restore -h "$HOST" -U "$USER" -d vigia_master "$BACKUP_DIR/master.dump"

# Get tenant list from restored master
TENANTS=$(psql -h "$HOST" -U "$USER" -d vigia_master -t -c \
  "SELECT db_name FROM clientes_saas;")

# Restore each tenant
for tenant in $TENANTS; do
    echo "Restoring $tenant..."
    
    # Drop if exists
    psql -h "$HOST" -U "$USER" -c "DROP DATABASE IF EXISTS $tenant;"
    
    # Restore
    if [ -f "$BACKUP_DIR/$tenant.dump" ]; then
        pg_restore -h "$HOST" -U "$USER" -C -d postgres "$BACKUP_DIR/$tenant.dump"
    else
        echo "WARNING: Backup not found for $tenant"
    fi
done

# Restart application
echo "Starting VIGIA backend..."
systemctl start vigia-backend

echo "Full system restore completed"

Data Export/Import

Export Tenant Data (JSON)

Create a Python script to export tenant data as JSON:

# export_tenant_data.py
import json
import sys
from sqlalchemy import create_engine, MetaData
from sqlalchemy.orm import sessionmaker

def export_tenant_data(db_url: str, output_file: str):
    engine = create_engine(db_url)
    Session = sessionmaker(bind=engine)
    session = Session()
    
    # Export key tables
    tables_to_export = [
        'users', 'icsr', 'products', 'clients', 
        'eventos_adversos', 'surveillance_alerts'
    ]
    
    export_data = {}
    
    for table_name in tables_to_export:
        print(f"Exporting {table_name}...")
        result = session.execute(f"SELECT * FROM {table_name}")
        rows = [dict(row) for row in result]
        export_data[table_name] = rows
    
    with open(output_file, 'w', encoding='utf-8') as f:
        json.dump(export_data, f, indent=2, default=str)
    
    print(f"Exported {sum(len(v) for v in export_data.values())} total records")
    session.close()

if __name__ == "__main__":
    db_url = sys.argv[1]
    output = sys.argv[2]
    export_tenant_data(db_url, output)

Usage:

python export_tenant_data.py \
  "postgresql://user:pass@localhost/vigia_empresa1" \
  export_empresa1.json

Export ICSR Cases (CSV)

-- export_icsr.sql
\COPY (
  SELECT 
    i.id,
    i.numero_caso,
    i.fecha_recepcion,
    i.estado,
    i.gravedad,
    i.pais_origen,
    p.nombre_generico,
    e.descripcion_evento
  FROM icsr i
  LEFT JOIN icsr_productos ip ON i.id = ip.icsr_id
  LEFT JOIN productos p ON ip.producto_id = p.id
  LEFT JOIN icsr_eventos ie ON i.id = ie.icsr_id
  LEFT JOIN eventos_adversos e ON ie.evento_id = e.id
  WHERE i.created_at >= '2024-01-01'
) TO '/tmp/icsr_export.csv' WITH CSV HEADER;

Import Data to New Tenant

After tenant provisioning, import historical data:

# import_tenant_data.py
import json
from sqlalchemy import create_engine, text
from sqlalchemy.orm import sessionmaker

def import_tenant_data(db_url: str, json_file: str):
    engine = create_engine(db_url)
    Session = sessionmaker(bind=engine)
    session = Session()
    
    with open(json_file, 'r', encoding='utf-8') as f:
        data = json.load(f)
    
    # Import in dependency order
    import_order = ['users', 'products', 'clients', 'icsr', 'eventos_adversos']
    
    for table_name in import_order:
        if table_name not in data:
            continue
            
        print(f"Importing {len(data[table_name])} rows to {table_name}...")
        
        for row in data[table_name]:
            columns = ', '.join(row.keys())
            placeholders = ', '.join(f":{k}" for k in row.keys())
            sql = f"INSERT INTO {table_name} ({columns}) VALUES ({placeholders})"
            
            try:
                session.execute(text(sql), row)
            except Exception as e:
                print(f"Error importing row: {e}")
                continue
        
        session.commit()
    
    print("Import completed")
    session.close()

if __name__ == "__main__":
    import sys
    db_url = sys.argv[1]
    json_file = sys.argv[2]
    import_tenant_data(db_url, json_file)

Tenant Deletion and Data Retention

Soft Delete (Deactivate)

Preserves data for potential reactivation (backend/app/routers/admin_clientes.py:124-128):

# Soft delete via API
curl -X DELETE http://api.vigia.com/admin/clientes/5

# Or directly in database
UPDATE clientes_saas SET is_active = false WHERE id = 5;

Data retained: All tenant data remains in database, tenant cannot authenticate.

Hard Delete (Permanent)

Completely removes tenant and database (backend/app/routers/admin_clientes.py:130-142):

# Via API (requires explicit ?hard=true)
curl -X DELETE "http://api.vigia.com/admin/clientes/5?hard=true"

Implementation:

Terminates all active connections to tenant database
Drops tenant database using DROP DATABASE
Deletes tenant record from clientes_saas
Logs deletion with tenant ID, subdomain, and database name

Recommendation: Always create backup before hard delete:

# Backup before deletion
pg_dump -h localhost -U postgres -Fc vigia_empresa1 \
  -f "archive/empresa1_deleted_$(date +%Y%m%d).dump"

# Then delete
curl -X DELETE "http://api.vigia.com/admin/clientes/5?hard=true"

Point-in-Time Recovery (PITR)

PostgreSQL supports recovering to any point in time using WAL archives.

Enable WAL Archiving

Edit postgresql.conf:

# Enable WAL archiving
wal_level = replica
archive_mode = on
archive_command = 'cp %p /var/lib/postgresql/wal_archive/%f'

# Retention
wal_keep_size = 1GB

Restart PostgreSQL:

systemctl restart postgresql

Create Base Backup

pg_basebackup -h localhost -U postgres \
  -D /var/backups/postgres_base_$(date +%Y%m%d) \
  -Ft -z -P -X stream

Recover to Point in Time

# 1. Stop PostgreSQL
systemctl stop postgresql

# 2. Restore base backup
cd /var/lib/postgresql/14/main
rm -rf *
tar -xzf /var/backups/postgres_base_20240303/base.tar.gz

# 3. Create recovery.conf
cat > recovery.conf <<EOF
restore_command = 'cp /var/lib/postgresql/wal_archive/%f %p'
recovery_target_time = '2024-03-03 14:30:00'
recovery_target_action = 'promote'
EOF

# 4. Start PostgreSQL (will recover to target time)
systemctl start postgresql

Cloud Backup Integration

AWS S3 Backup

#!/bin/bash
# Upload to S3 after backup

BACKUP_FILE="/var/backups/vigia/master_$(date +%Y%m%d).dump"
S3_BUCKET="s3://my-vigia-backups/"

# Create backup
pg_dump -h localhost -U postgres -Fc vigia_master -f "$BACKUP_FILE"

# Upload to S3
aws s3 cp "$BACKUP_FILE" "$S3_BUCKET"

# Set lifecycle policy (optional)
aws s3api put-object-tagging --bucket my-vigia-backups \
  --key "$(basename $BACKUP_FILE)" \
  --tagging 'TagSet=[{Key=Retention,Value=90days}]'

Google Cloud Storage

# Upload to GCS
gsutil cp /var/backups/vigia/master_20240303.dump \
  gs://my-vigia-backups/

# Set retention policy
gsutil lifecycle set retention.json gs://my-vigia-backups/

Monitoring and Alerts

Backup Verification Script

#!/bin/bash
# verify_backups.sh

BACKUP_DIR="/var/backups/vigia"
MAX_AGE_HOURS=26  # Alert if backup older than 26 hours

latest_master=$(find "$BACKUP_DIR" -name "master_*.dump" -mtime -1 | wc -l)

if [ "$latest_master" -eq 0 ]; then
    echo "ALERT: No master backup in last 24 hours"
    # Send notification
    curl -X POST https://api.slack.com/webhooks/... \
      -d '{"text": "VIGIA backup alert: No recent master backup"}'
    exit 1
fi

echo "Backup verification passed"

Cron for Verification

# Check backups every 6 hours
0 */6 * * * /usr/local/bin/verify_backups.sh

Best Practices

Backup Strategy

3-2-1 Rule: 3 copies, 2 different media, 1 offsite
Test restores regularly: Monthly restore drills
Automate backups: Use cron or systemd timers
Monitor backup jobs: Alert on failures
Document procedures: Keep runbooks updated

Security

Encrypt backups: Use gpg or cloud provider encryption
Secure credentials: Use .pgpass or environment variables
Limit access: Restrict backup file permissions (chmod 600)
Audit logs: Track who accesses backups

Performance

Parallel dumps: Use -j flag for large databases
Compression: Always compress backups (-Fc or gzip)
Off-peak backups: Schedule during low usage
Incremental backups: Use WAL archiving for large databases

Retention

Daily backups: Keep 90 days for active tenants
Weekly backups: Keep 1 year
Monthly backups: Keep indefinitely (compliance)
Archive deleted tenants: Keep 2 years after deletion

Disaster Recovery Plan

Recovery Time Objective (RTO)

Master Database: < 1 hour
Critical Tenants: < 4 hours
All Tenants: < 24 hours

Recovery Point Objective (RPO)

Master Database: < 1 hour (hourly backups)
Tenant Databases: < 24 hours (daily backups)
PITR: < 5 minutes (WAL archiving)

Emergency Contacts

# Document in /etc/vigia/emergency_contacts.yml
contacts:
  - role: Database Administrator
    name: John Doe
    phone: +51 999 999 999
    email: dba@empresa.com
  
  - role: System Administrator
    name: Jane Smith
    phone: +51 888 888 888
    email: sysadmin@empresa.com

Get Started

Core Features

User Guides

Modules

Regulatory

Administration

Documentation Index

​Overview

​Backup Strategy

​Database Hierarchy

​Backup Frequency Recommendations

​PostgreSQL Backup Methods

​pg_dump (Logical Backup)

​Backup Single Tenant

​Backup Master Database

​Backup All Tenants (Script)

​pg_basebackup (Physical Backup)

​Automated Backup Scripts

​Daily Backup with Rotation

​Cron Configuration

​Restore Procedures

​Restore Single Tenant

​From SQL Dump

​From Custom Format

​Restore Master Database

​Disaster Recovery (Full System)

​Data Export/Import

​Export Tenant Data (JSON)

​Export ICSR Cases (CSV)

​Import Data to New Tenant

​Tenant Deletion and Data Retention

​Soft Delete (Deactivate)

​Hard Delete (Permanent)

​Point-in-Time Recovery (PITR)

​Enable WAL Archiving

​Create Base Backup

​Recover to Point in Time

​Cloud Backup Integration

​AWS S3 Backup

​Google Cloud Storage

​Monitoring and Alerts

​Backup Verification Script

​Cron for Verification

​Best Practices

​Backup Strategy

​Security

​Performance

​Retention

​Disaster Recovery Plan

​Recovery Time Objective (RTO)

​Recovery Point Objective (RPO)

​Emergency Contacts

​Related Documentation

Build docs developers (and LLMs) love

Overview

Backup Strategy

Database Hierarchy

Backup Frequency Recommendations

PostgreSQL Backup Methods

pg_dump (Logical Backup)

Backup Single Tenant

Backup Master Database

Backup All Tenants (Script)

pg_basebackup (Physical Backup)

Automated Backup Scripts

Daily Backup with Rotation

Cron Configuration

Restore Procedures

Restore Single Tenant

From SQL Dump

From Custom Format

Restore Master Database

Disaster Recovery (Full System)

Data Export/Import

Export Tenant Data (JSON)

Export ICSR Cases (CSV)

Import Data to New Tenant

Tenant Deletion and Data Retention

Soft Delete (Deactivate)

Hard Delete (Permanent)

Point-in-Time Recovery (PITR)

Enable WAL Archiving

Create Base Backup

Recover to Point in Time

Cloud Backup Integration

AWS S3 Backup

Google Cloud Storage

Monitoring and Alerts

Backup Verification Script

Cron for Verification

Best Practices

Backup Strategy

Security

Performance

Retention

Disaster Recovery Plan

Recovery Time Objective (RTO)

Recovery Point Objective (RPO)

Emergency Contacts

Related Documentation