Remove clear-previous-error and use enable-cr-debug-metadata instead

Signed-off-by: Thomas Newton <[email protected]>
Tom-Newton · Jan 16, 2024 · 6c5650c · 6c5650c
1 parent 216e1a3
commit 6c5650c
Show file tree

Hide file tree

Showing 5 changed files with 8 additions and 27 deletions.
diff --git a/flytepropeller/pkg/controller/config/config.go b/flytepropeller/pkg/controller/config/config.go
@@ -99,7 +99,6 @@ var (
 			DefaultMaxAttempts:             1,
 			IgnoreRetryCause:               false,
 			EnableCRDebugMetadata:          false,
-			ClearPreviousError:             false,
 		},
 		MaxStreakLength: 8, // Turbo mode is enabled by default
 		ProfilerPort: config.Port{
@@ -215,8 +214,7 @@ type NodeConfig struct {
 	InterruptibleFailureThreshold  int32            `json:"interruptible-failure-threshold" pflag:"1,number of failures for a node to be still considered interruptible. Negative numbers are treated as complementary (ex. -1 means last attempt is non-interruptible).'"`
 	DefaultMaxAttempts             int32            `json:"default-max-attempts" pflag:"3,Default maximum number of attempts for a node"`
 	IgnoreRetryCause               bool             `json:"ignore-retry-cause" pflag:",Ignore retry cause and count all attempts toward a node's max attempts"`
-	EnableCRDebugMetadata          bool             `json:"enable-cr-debug-metadata" pflag:",Collapse node on any terminal state, not just successful terminations. This is useful to reduce the size of workflow state in etcd."`
-	ClearPreviousError             bool             `json:"clear-previous-error" pflag:",When using the fail after executable nodes complete mode many nodes in the workflow can fail. This option causes the previous error to be cleared every time a new error occurs. This helps reduce the size of workflow state in etcd."`
+	EnableCRDebugMetadata          bool             `json:"enable-cr-debug-metadata" pflag:",By default node state gets cleared after flytepropeller will no longer need it. This is useful to reduce the size of workflow state in etcd. Consider enabling this to keep this state for debugging purposes."`
 }
 
 // DefaultDeadlines contains default values for timeouts

diff --git a/flytepropeller/pkg/controller/config/config_flags.go b/flytepropeller/pkg/controller/config/config_flags.go
diff --git a/flytepropeller/pkg/controller/config/config_flags_test.go b/flytepropeller/pkg/controller/config/config_flags_test.go
diff --git a/flytepropeller/pkg/controller/nodes/executor.go b/flytepropeller/pkg/controller/nodes/executor.go
@@ -481,7 +481,6 @@ type nodeExecutor struct {
 	catalog                         catalog.Client
 	clusterID                       string
 	enableCRDebugMetadata           bool
-	clearPreviousError              bool
 	defaultActiveDeadline           time.Duration
 	defaultDataSandbox              storage.DataReference
 	defaultExecutionDeadline        time.Duration
@@ -870,7 +869,7 @@ func (c *nodeExecutor) execute(ctx context.Context, h interfaces.NodeHandler, nC
 }
 
 func (c *nodeExecutor) Clear(executableNodeStatus v1alpha1.ExecutableNodeStatus) {
-	if c.clearPreviousError {
+	if !c.enableCRDebugMetadata {
 		executableNodeStatus.ClearExecutionError()
 	}
 }
@@ -1449,7 +1448,6 @@ func NewExecutor(ctx context.Context, nodeConfig config.NodeConfig, store *stora
 		catalog:                         catalogClient,
 		clusterID:                       clusterID,
 		enableCRDebugMetadata:           nodeConfig.EnableCRDebugMetadata,
-		clearPreviousError:              nodeConfig.ClearPreviousError,
 		defaultActiveDeadline:           nodeConfig.DefaultDeadlines.DefaultNodeActiveDeadline.Duration,
 		defaultDataSandbox:              defaultRawOutputPrefix,
 		defaultExecutionDeadline:        nodeConfig.DefaultDeadlines.DefaultNodeExecutionDeadline.Duration,

diff --git a/flytepropeller/pkg/controller/workflow/executor_test.go b/flytepropeller/pkg/controller/workflow/executor_test.go
@@ -540,15 +540,15 @@ func TestWorkflowExecutor_HandleFlyteWorkflow_Failing(t *testing.T) {
 	tests := []struct {
 		name                         string
 		onFailurePolicy              v1alpha1.WorkflowOnFailurePolicy
-		clearPreviousError           bool
+		enableCRDebugMetadata        bool
 		expectedRoundsToFail         int
 		expectedNodesWithErrorsCount int
 		expectedFailedNodesCount     int
 	}{
 		{"failImidiately", v1alpha1.WorkflowOnFailurePolicy(core.WorkflowMetadata_FAIL_IMMEDIATELY), false, 6, 1, 1},
-		{"failImidiately clearPreviousError", v1alpha1.WorkflowOnFailurePolicy(core.WorkflowMetadata_FAIL_IMMEDIATELY), true, 6, 1, 1},
-		{"failAfterExecutableNodesComplete", v1alpha1.WorkflowOnFailurePolicy(core.WorkflowMetadata_FAIL_AFTER_EXECUTABLE_NODES_COMPLETE), false, 12, 2, 2},
-		{"failAfterExecutableNodesComplete clearPreviousError", v1alpha1.WorkflowOnFailurePolicy(core.WorkflowMetadata_FAIL_AFTER_EXECUTABLE_NODES_COMPLETE), true, 12, 1, 2},
+		{"failImidiately enableCRDebugMetadata", v1alpha1.WorkflowOnFailurePolicy(core.WorkflowMetadata_FAIL_IMMEDIATELY), true, 6, 1, 1},
+		{"failAfterExecutableNodesComplete", v1alpha1.WorkflowOnFailurePolicy(core.WorkflowMetadata_FAIL_AFTER_EXECUTABLE_NODES_COMPLETE), false, 12, 1, 2},
+		{"failAfterExecutableNodesComplete enableCRDebugMetadata", v1alpha1.WorkflowOnFailurePolicy(core.WorkflowMetadata_FAIL_AFTER_EXECUTABLE_NODES_COMPLETE), true, 12, 2, 2},
 	}
 
 	wJSON, err := yamlutils.ReadYamlFileAsJSON("testdata/benchmark_wf.yaml")
@@ -557,7 +557,7 @@ func TestWorkflowExecutor_HandleFlyteWorkflow_Failing(t *testing.T) {
 
 		t.Run(test.name, func(t *testing.T) {
 			nodeConfig := config.GetConfig().NodeConfig
-			nodeConfig.ClearPreviousError = test.clearPreviousError
+			nodeConfig.EnableCRDebugMetadata = test.enableCRDebugMetadata
 			nodeExec, err := nodes.NewExecutor(ctx, nodeConfig, store, enqueueWorkflow, eventSink, adminClient, adminClient,
 				maxOutputSize, "s3://bucket", fakeKubeClient, catalogClient, recoveryClient, eventConfig, testClusterID, signalClient, handlerFactory, promutils.NewTestScope())
 			assert.NoError(t, err)