feat: support for spot reliability on node templates

Author: primoz-hrovat

castai/resource_eviction_config_test.go

@@ -160,7 +160,8 @@ func TestEvictionConfig_CreateContext(t *testing.T) {
           "matchLabels": {
             "key1": "value1"
           }
-        }
+        },
+		"replicasMin": null
       },
       "settings": {
         "aggressive": {
@@ -265,7 +266,8 @@ func TestEvictionConfig_UpdateContext(t *testing.T) {
           "matchLabels": {
             "key1": "value1"
           }
-        }
+        },
+        "replicasMin": null
       },
       "settings": {
         "aggressive": {

castai/resource_node_template.go

@@ -56,6 +56,8 @@ const (
 	FieldNodeTemplateShouldTaint                              = "should_taint"
 	FieldNodeTemplateSpot                                     = "spot"
 	FieldNodeTemplateSpotDiversityPriceIncreaseLimitPercent   = "spot_diversity_price_increase_limit_percent"
+	FieldNodeTemplateSpotReliabilityEnabled                   = "spot_reliability_enabled"
+	FieldNodeTemplateSpotReliabilityPriceIncreaseLimitPercent = "spot_reliability_price_increase_limit_percent"
 	FieldNodeTemplateSpotInterruptionPredictionsEnabled       = "spot_interruption_predictions_enabled"
 	FieldNodeTemplateSpotInterruptionPredictionsType          = "spot_interruption_predictions_type"
 	FieldNodeTemplateStorageOptimized                         = "storage_optimized"
@@ -235,6 +237,17 @@ func resourceNodeTemplate() *schema.Resource {
 							Optional:    true,
 							Description: "Allowed node configuration price increase when diversifying instance types. E.g. if the value is 10%, then the overall price of diversified instance types can be 10% higher than the price of the optimal configuration.",
 						},
+						FieldNodeTemplateSpotReliabilityEnabled: {
+							Type:        schema.TypeBool,
+							Default:     false,
+							Optional:    true,
+							Description: "Enable/disable spot reliability. When enabled, autoscaler will create instances with highest reliability score within price increase threshold.",
+						},
+						FieldNodeTemplateSpotReliabilityPriceIncreaseLimitPercent: {
+							Type:        schema.TypeInt,
+							Optional:    true,
+							Description: "Allowed node price increase when using spot reliability on ordering the instance types . E.g. if the value is 10%, then the overall price of instance types can be 10% higher than the price of the optimal configuration.",
+						},
 						FieldNodeTemplateSpotInterruptionPredictionsEnabled: {
 							Type:        schema.TypeBool,
 							Default:     false,
@@ -766,6 +779,12 @@ func flattenConstraints(c *sdk.NodetemplatesV1TemplateConstraints) ([]map[string
 	if c.SpotDiversityPriceIncreaseLimitPercent != nil {
 		out[FieldNodeTemplateSpotDiversityPriceIncreaseLimitPercent] = c.SpotDiversityPriceIncreaseLimitPercent
 	}
+	if c.EnableSpotReliability != nil {
+		out[FieldNodeTemplateSpotReliabilityEnabled] = c.EnableSpotReliability
+	}
+	if c.SpotReliabilityPriceIncreaseLimitPercent != nil {
+		out[FieldNodeTemplateSpotReliabilityPriceIncreaseLimitPercent] = c.SpotReliabilityPriceIncreaseLimitPercent
+	}
 	if c.SpotInterruptionPredictionsEnabled != nil {
 		out[FieldNodeTemplateSpotInterruptionPredictionsEnabled] = c.SpotInterruptionPredictionsEnabled
 	}
@@ -1347,6 +1366,12 @@ func toTemplateConstraints(obj map[string]any) *sdk.NodetemplatesV1TemplateConst
 	if v, ok := obj[FieldNodeTemplateSpotDiversityPriceIncreaseLimitPercent].(int); ok {
 		out.SpotDiversityPriceIncreaseLimitPercent = toPtr(int32(v))
 	}
+	if v, ok := obj[FieldNodeTemplateSpotReliabilityEnabled].(bool); ok {
+		out.EnableSpotReliability = toPtr(v)
+	}
+	if v, ok := obj[FieldNodeTemplateSpotReliabilityPriceIncreaseLimitPercent].(int); ok {
+		out.SpotReliabilityPriceIncreaseLimitPercent = toPtr(int32(v))
+	}
 	if v, ok := obj[FieldNodeTemplateSpotInterruptionPredictionsEnabled].(bool); ok {
 		out.SpotInterruptionPredictionsEnabled = toPtr(v)
 	}

castai/resource_node_template_test.go

@@ -54,6 +54,8 @@ func TestNodeTemplateResourceReadContext(t *testing.T) {
 				  "fallbackRestoreRateSeconds": 0,
 				  "enableSpotDiversity": false,
 				  "spotDiversityPriceIncreaseLimitPercent": 20,
+				  "enableSpotReliability": true,
+				  "spotReliabilityPriceIncreaseLimitPercent": 10,
 				  "spotInterruptionPredictionsEnabled": true,
 				  "spotInterruptionPredictionsType": "aws-rebalance-recommendations",
 				  "storageOptimized": true,
@@ -223,6 +225,8 @@ constraints.0.resource_limits.0.cpu_limit_enabled = true
 constraints.0.resource_limits.0.cpu_limit_max_cores = 20
 constraints.0.spot = false
 constraints.0.spot_diversity_price_increase_limit_percent = 20
+constraints.0.spot_reliability_enabled = true
+constraints.0.spot_reliability_price_increase_limit_percent = 10
 constraints.0.spot_interruption_predictions_enabled = true
 constraints.0.spot_interruption_predictions_type = aws-rebalance-recommendations
 constraints.0.storage_optimized = false
@@ -710,8 +714,10 @@ func TestAccResourceNodeTemplate_basic(t *testing.T) {
 					resource.TestCheckResourceAttr(resourceName, "constraints.0.azs.1", "eu-central-1b"),
 					resource.TestCheckResourceAttr(resourceName, "constraints.0.azs.2", "eu-central-1c"),
 					resource.TestCheckResourceAttr(resourceName, "is_default", "false"),
-					resource.TestCheckResourceAttr(resourceName, "constraints.0.enable_spot_diversity", "true"),
+					resource.TestCheckResourceAttr(resourceName, "constraints.0.enable_spot_diversity", "false"),
 					resource.TestCheckResourceAttr(resourceName, "constraints.0.spot_diversity_price_increase_limit_percent", "22"),
+					resource.TestCheckResourceAttr(resourceName, "constraints.0.enable_spot_reliability", "true"),
+					resource.TestCheckResourceAttr(resourceName, "constraints.0.spot_reliability_price_increase_limit_percent", "15"),
 					resource.TestCheckResourceAttr(resourceName, "constraints.0.spot_interruption_predictions_enabled", "true"),
 					resource.TestCheckResourceAttr(resourceName, "constraints.0.spot_interruption_predictions_type", "interruption-predictions"),
 					resource.TestCheckResourceAttr(resourceName, "constraints.0.custom_priority.#", "2"),
@@ -852,8 +858,10 @@ func testNodeTemplateUpdated(rName, clusterName string) string {
 				use_spot_fallbacks = true
 				spot = true
 				on_demand = true
-				enable_spot_diversity = true
+				enable_spot_diversity = false
 				spot_diversity_price_increase_limit_percent = 22
+				enable_spot_reliability = true
+				spot_reliability_price_increase_limit_percent = 15
 				spot_interruption_predictions_enabled = true
 				spot_interruption_predictions_type = "interruption-predictions"
 				fallback_restore_rate_seconds = 1800