fix(cluster): fix scaling logic and dedup metrics

Author: MasterPtato

packages/core/services/cluster/src/ops/datacenter/topology_get/mod.rs

@@ -118,8 +118,7 @@ pub async fn cluster_datacenter_topology_get(
 		FROM db_cluster.servers
 		WHERE
 			datacenter_id = ANY($1) AND
-			cloud_destroy_ts IS NULL AND
-			taint_ts IS NULL
+			cloud_destroy_ts IS NULL
 		",
 		&input.datacenter_ids,
 	)

packages/core/services/cluster/src/ops/datacenter/topology_get/pegboard.rs

@@ -37,23 +37,23 @@ pub async fn pegboard_client_usage_get(ctx: &OperationCtx, input: &Input) -> Glo
 			r#"
 			label_replace(
 				sum by (client_id) (
-					last_over_time(
-						rivet_pegboard_client_cpu_allocated{{
-							client_id=~"({client_ids})",
-						}}
-						[15m:15s]
+					rivet_pegboard_client_cpu_allocated{{client_id=~"({client_ids})"}}
+					# This "and on" deduplicates metrics by their inherent timestamp, choosing the latest data
+					and on(client_id, instance) (
+						timestamp(rivet_pegboard_client_cpu_allocated{{client_id=~"({client_ids})"}}) ==
+						scalar(max(timestamp(rivet_pegboard_client_cpu_allocated{{client_id=~"({client_ids})"}})))
 					)
 				),
 				"metric", "cpu", "", ""
 			)
 			OR
 			label_replace(
 				sum by (client_id) (
-					last_over_time(
-						rivet_pegboard_client_memory_allocated{{
-							client_id=~"({client_ids})",
-						}}
-						[15m:15s]
+					rivet_pegboard_client_memory_allocated{{client_id=~"({client_ids})"}}
+					# This "and on" deduplicates metrics by their inherent timestamp, choosing the latest data
+					and on(client_id, instance) (
+						timestamp(rivet_pegboard_client_memory_allocated{{client_id=~"({client_ids})"}}) ==
+						scalar(max(timestamp(rivet_pegboard_client_memory_allocated{{client_id=~"({client_ids})"}})))
 					)
 				),
 				"metric", "mem", "", ""