Instruction file imported from moetez47/grmanproject (.cursor/rules/ocp-processing.mdc). Copyright stays with the author.

OpenShift Processing Patterns

Report Types

OCP has multiple report types, each with raw and daily table variants:

• pod_usage - CPU, memory usage per pod
• storage_usage - Persistent volume claims
• node_labels - Node metadata and labels
• namespace_labels - Namespace metadata
• gpu_usage - GPU utilization (NVIDIA)

Table Name Mappings

TRINO_LINE_ITEM_TABLE_MAP = {
    "pod_usage": "openshift_pod_usage_line_items",
    "gpu_usage": "openshift_gpu_usage_line_items",
    ...
}

TRINO_LINE_ITEM_TABLE_DAILY_MAP = {
    "pod_usage": "openshift_pod_usage_line_items_daily",
    "gpu_usage": "openshift_gpu_usage_line_items_daily",
    ...
}

Daily tables are identified by /daily in the S3 path.

GPU Processing

GPU data includes:

• gpu_uuid - Physical GPU identifier
• mig_instance_uuid - MIG (Multi-Instance GPU) partition ID
• mig_slice_count / parent_gpu_max_slices - For fractional GPU billing
• gpu_pod_uptime - Seconds of GPU utilization

MIG cost formula:

cost = (rate / days_in_month) * (uptime / 86400) * (slice_count / max_slices)

Cost Model Application

Cost models apply rates via SQL templates:

1. monthly_cost_*.sql - Calculate costs from rates
2. distribute_cost/*.sql - Distribute unallocated costs
3. UI summary tables aggregate for display

Cluster Identification

• cluster_id - Unique cluster identifier
• cluster_alias - Human-readable name (from Provider.name)
• source_uuid - Provider UUID linking to sources table