Semantic Layer Deployment¶

Deploy semantic models and metrics to production data warehouses for BI tool consumption.

Deployment Overview¶

Seeknal deploys semantic models and metrics to StarRocks, creating optimized materialized views that BI tools can query directly.

Pre-Deployment Checklist¶

Before deploying, ensure:

Semantic models are tested and validated
Metrics are calculated correctly
Performance is acceptable
Access controls are configured
Documentation is complete

Deploying to StarRocks¶

Initial Deployment¶

# Deploy all semantic models and metrics
seeknal deploy-semantic-layer

# Deploy specific model
seeknal deploy-semantic-layer --model ecommerce_semantic_model

# Deploy with environment
seeknal deploy-semantic-layer --env production

What Gets Deployed¶

Base Tables: Raw data from sources
Semantic Models: Joined and transformed data
Metric Tables: Pre-calculated aggregations
Materialized Views: Optimized query performance

Deployment Configuration¶

StarRocks Connection¶

# config/prod.toml
[starrocks]
host = "prod-starrocks.example.com"
port = 9030
username = "seeknal_user"
password = "${STARROCKS_PASSWORD}"
database = "analytics_prod"

Materialization Strategy¶

semantic_model:
  name: ecommerce_semantic_model

  materialization:
    type: materialized_view
    refresh_strategy: manual  # or scheduled
    schedule: "0 2 * * *"  # 2 AM daily

Refresh Strategies¶

Manual Refresh¶

# Trigger manual refresh
seeknal refresh-semantic-layer --model ecommerce_semantic_model

Scheduled Refresh¶

materialization:
  type: materialized_view
  refresh_strategy: scheduled
  schedule: "0 2 * * *"  # Cron schedule

Incremental Refresh¶

materialization:
  type: incremental
  incremental_key: updated_at
  lookback_days: 7

Validation¶

Post-Deployment Validation¶

# Validate deployment
seeknal validate-semantic-layer --model ecommerce_semantic_model

Checks: - View creation success - Row counts match expectations - Query performance acceptable - Access controls working

Manual Validation¶

-- Test deployed semantic model
SELECT region, total_revenue
FROM ecommerce_semantic_model_mv
GROUP BY region;

Access Control¶

Granting Permissions¶

-- Grant read access to BI users
GRANT SELECT ON ecommerce_semantic_model_mv TO bi_user;
GRANT SELECT ON ecommerce_metrics_mv TO bi_user;

Row-Level Security¶

semantic_model:
  name: customer_metrics

  security:
    row_filter: "region = CURRENT_USER_REGION()"

Troubleshooting¶

Deployment Failures¶

Issue: Deployment fails with connection error.

Solution: Check StarRocks connection configuration.

Issue: View creation fails due to invalid SQL.

Solution: Test semantic model locally first.

Performance Issues¶

Issue: Queries are slow after deployment.

Solutions: - Check materialization strategy - Add appropriate indexes - Review query plans

Best Practices¶

Test locally before deploying
Use environments (dev → staging → prod)
Monitor performance after deployment
Version changes with Git
Document refresh schedules

Semantic Models - Defining models
Metrics - Creating metrics
Querying - Query patterns

Next: Learn about Querying or return to Semantic Layer Overview