[COST-5923] OCP virtual machine distributed cost SQL (#5517)

* [COST-5678] sql to insert distributed costs for vms * add comments to sql file * dont group by cost-model-rate-type * calculate over the month, not last day --------- Co-authored-by: Luke Couzens <[email protected]>
project-koku · Feb 26, 2025 · 96dccda · 96dccda
1 parent 97adbb0
commit 96dccda
Showing 1 changed file with 90 additions and 1 deletion.
diff --git a/koku/masu/database/sql/openshift/reporting_ocp_vm_summary_p.sql b/koku/masu/database/sql/openshift/reporting_ocp_vm_summary_p.sql
@@ -6,7 +6,7 @@ WHERE usage_start >= {{start_date}}::date
 
 WITH second_to_last_day AS (
     SELECT DISTINCT usage_start::date AS day
-    FROM reporting_ocpusagelineitem_daily_summary
+    FROM {{schema | sqlsafe}}.reporting_ocpusagelineitem_daily_summary
     WHERE usage_start >= {{start_date}}::date
         AND usage_start <= {{end_date}}::date
         AND source_uuid = {{source_uuid}}
@@ -94,3 +94,92 @@ WHERE usage_start >= {{start_date}}::date
         + COALESCE(infrastructure_markup_cost, 0)) != 0
 GROUP BY cluster_alias, cluster_id, namespace, latest.node_name, vm_name, cost_model_rate_type, latest.labels
 ;
+
+
+WITH cte_distribution_type AS (
+    -- get the distribution type from the cost model associated with this source
+    SELECT
+        {{source_uuid}}::uuid as source_uuid,
+        coalesce(max(cm.distribution), 'cpu')  as dt -- coalesce(max()) to return 'cpu' if a cost model does not exist for source
+    FROM {{schema | sqlsafe}}.cost_model_map as cmm
+    JOIN {{schema | sqlsafe}}.cost_model as cm
+        ON cmm.cost_model_id = cm.uuid
+    WHERE cmm.provider_uuid = {{source_uuid}}
+),
+cte_node_usage AS (
+    SELECT
+        namespace,
+        node,
+        usage_start,
+        sum(pod_effective_usage_cpu_core_hours) as cpu_usage,
+        sum(pod_effective_usage_memory_gigabyte_hours) as mem_usage
+    FROM {{schema | sqlsafe}}.reporting_ocpusagelineitem_daily_summary
+    WHERE usage_start >= {{start_date}}::date
+        AND usage_start <= {{end_date}}::date
+        AND source_uuid = {{source_uuid}}
+    GROUP BY namespace, node, usage_start
+),
+cte_latest_resources AS (
+    SELECT
+        pod_labels->>'vm_kubevirt_io_name' AS vm_name,
+        CASE WHEN cte_distribution_type.dt = 'cpu'
+            THEN sum(pod_effective_usage_cpu_core_hours) / sum(cte_node_usage.cpu_usage)
+            ELSE sum(pod_effective_usage_memory_gigabyte_hours) / sum(cte_node_usage.mem_usage)
+        END as ratio,
+        cte_node_usage.node as node_name,
+        cte_node_usage.namespace as namespace_name,
+        pod_labels as labels,
+        cte_node_usage.usage_start
+    FROM {{schema | sqlsafe}}.reporting_ocpusagelineitem_daily_summary as ocp
+    JOIN cte_distribution_type
+        ON cte_distribution_type.source_uuid = ocp.source_uuid
+    JOIN cte_node_usage
+        ON cte_node_usage.namespace = ocp.namespace
+        AND cte_node_usage.node = ocp.node
+        AND cte_node_usage.usage_start = ocp.usage_start
+    WHERE pod_labels ? 'vm_kubevirt_io_name'
+        AND pod_effective_usage_cpu_core_hours != 0
+        AND pod_effective_usage_memory_gigabyte_hours != 0
+    GROUP BY vm_name, cte_distribution_type.dt, cte_node_usage.usage_start, cte_node_usage.node, cte_node_usage.namespace, pod_labels
+)
+INSERT INTO {{schema | sqlsafe}}.reporting_ocp_vm_summary_p (
+    id,
+    cluster_alias,
+    cluster_id,
+    namespace,
+    node,
+    vm_name,
+    cost_model_rate_type,
+    distributed_cost,
+    pod_labels,
+    raw_currency,
+    resource_ids,
+    usage_start,
+    usage_end,
+    source_uuid
+)
+SELECT
+    uuid_generate_v4() as id,
+    cluster_alias,
+    cluster_id,
+    namespace,
+    node,
+    latest.vm_name as vm_name,
+    cost_model_rate_type,
+    sum(distributed_cost * latest.ratio) as distributed_cost, -- the only cost inserted in this statement
+    pod_labels,
+    max(raw_currency) as raw_currency,
+    array_agg(DISTINCT resource_id) as resource_ids,
+    min(latest.usage_start) as usage_start,
+    max(latest.usage_start) as usage_end,
+    {{source_uuid}}::uuid as source_uuid
+FROM {{schema | sqlsafe}}.reporting_ocpusagelineitem_daily_summary as ocp
+JOIN cte_latest_resources as latest
+    ON latest.node_name = ocp.node
+    AND latest.namespace_name = ocp.namespace
+    AND latest.usage_start = ocp.usage_start
+WHERE source_uuid = {{source_uuid}}
+    AND distributed_cost != 0
+GROUP BY cluster_alias, cluster_id, namespace, node, vm_name, pod_labels, cost_model_rate_type
+
+;