Multi-Database Backend Awareness
This section provides backend-specific operational awareness for multi-database deployments. DBA teams may also reference sizing tables herein.
Backend-Specific Sizing Reference (Service Implementation Team Awareness)
The Service Implementation Team does not size or tune infrastructure, but must know minimum baselines in order to identify when observed resource constraints warrant DBA escalation.
| Backend | CPU | RAM | Disk / Storage | Notes |
|---|---|---|---|---|
| Oracle Database 19c or later. Oracle 21c | 16 cores | 64 GB | 500 GB SSD | Scale to 128–256 GB RAM + 1 TB SSD for production. Add Oracle RAC nodes for HA. |
| SQL Server 2019 / 2022 | 16 cores | 72 GB | 750 GB | Enable Columnstore indexes + In-Memory OLTP. TempDB: multiple data files aligned to CPU cores. Sizing TBD — consult DBA. SQL Server Node to 1. |
| Airflow (all backends) | 8 cores | 32 GB | 500 GB | Scale worker count for parallel DAG execution. |
| Snowflake | N/A (cloud) | N/A (cloud) | N/A (cloud) | Start Medium warehouse for ETL, Small for analytics. Scale by concurrency. Auto-suspend at 60s idle. |
Key DBA escalation triggers per backend:
- Snowflake: Virtual Warehouse credit exhaustion, warehouse fails to auto-resume, COPY INTO stage access permission failures, repeated query result cache misses causing sustained slow analytics.
- SQL Server 2019/2022: Repeated T+1 SLA breaches, JDBC connection failures, dbt model execution errors on SQL Server.
- Oracle: Space issues, locking, Oracle RAC node failures, AWR/ASH alert thresholds exceeded - same as in previous releases.
Service Implementation Team Operational Awareness
The following new components have specific operational behaviors the Service Implementation Team must understand for monitoring and failure response. These complement the failure modes already documented in Monitoring, Logging, and Typical Failure Modes.
Audience Resolution Layer (RDV → Unica 360 transition):
- Driven by Audience_map metadata table and Campaign–Offer–Channel–Product bridge. Service Implementation Team must not modify these tables.
- If Campaign 360 or Flowchart 360 aggregations appear incorrect (wrong audience counts, missing customer IDs).
- Escalate to: Services team.
- Failure mode: missing or misconfigured Audience_map entries silently produce incorrect aggregations rather than hard errors. Check Metadata Architect before raising a DBA ticket.
Aggregate Layer (dbt incremental models in Unica 360):
- Replaces direct BDV snapshot reads for Customer 360 and Campaign 360. Both 360 layers now depend on this layer completing successfully.
- Supports rolling-window metrics (7d, 30d, 90d). If 360 metrics appear stale by more than one day, check the dbt incremental model run status in Airflow before escalating.
- No full-refresh fallback exists in Release 26.1. If an incremental run fails mid-way, capture the failing dbt model name and escalate to Services team.
Two supported database backends are included in addition to Oracle: Snowflake Cloud and Microsoft SQL Server (2019 and 2022). Service Implementation Team Operations teams must understand which backend their deployment uses, as monitoring steps, log locations, and escalation paths differ by backend.
Service Implementation Team-relevant operational differences per backend. For full installation and configuration details, refer to the HCL MaxAI Canonical Installation Guide.
Supported Backend Summary
| Backend | Service Implementation Team Log Location | Key Service Implementation Team Operational Notes | DBA Escalation Trigger |
|---|---|---|---|
| Oracle Database 19c, Oracle 21c | ETL Pipeline Logs - Airflow UI and PV logs. | Primary production backend. All runbooks in Chapters 2–3 default to Oracle unless stated otherwise. | Space issues, locking, resource limits, DDL DAG failures. |
| Snowflake Cloud | ETL Pipeline Logs - Airflow UI and PV logs. | Ingestion via COPY INTO from cloud storage stages. Auto-suspend/resume means a “warehouse not running” error may appear on the first DAG task — confirm warehouse resumes before escalating. | Warehouse credit exhaustion, stage access permission failures, COPY INTO repeated failures. |
| SQL Server 2019/ SQL Server 2022 | ETL Pipeline Logs - Airflow UI and PV logs. | Transformation logic is implemented via dbt models across all backends, including SQL Server. CDC delays may be higher at high volumes. If T+1 SLA is breached, escalate to DBA before rerunning. Airflow connects via Microsoft JDBC driver bundled in the HCL MaxAI CDM image. | Repeated T+1 SLA breaches, JDBC connection failures, T-SQL procedure errors. |
Oracle Setup DAGs — Backend-Specific Note
The three Oracle Setup DAGs described in Chapter 2 Runbook 6 (airflow_variable_sync, dddl_execution_dag_multidb, etl_date_control_update_dag) apply to Oracle /SQL Server /Snowflake deployments. For Oracle /SQL Server /Snowflake backends, equivalent setup procedures are documented in the HCL MaxAI CDM Installation Guide. The Service Implementation Team must confirm with the Services Team which setup DAG set is applicable for their deployment before executing any DDL automation.