ImagePull failure is now reported on the UI (flyteorg#77)

go/tasks/pluginmachinery/flytek8s/pod_helper.go

@@ -138,6 +138,8 @@ func DemystifyPending(status v1.PodStatus) (pluginsCore.PhaseInfo, error) {
 							// There are a variety of reasons that can cause a pod to be in this waiting state.
 							// Waiting state may be legitimate when the container is being downloaded, started or init containers are running
 							reason := containerStatus.State.Waiting.Reason
+							finalReason := fmt.Sprintf("%s|%s", c.Reason, reason)
+							finalMessage := fmt.Sprintf("%s|%s", c.Message, containerStatus.State.Waiting.Message)
 							switch reason {
 							case "ErrImagePull", "ContainerCreating", "PodInitializing":
 								// But, there are only two "reasons" when a pod is successfully being created and hence it is in
@@ -149,12 +151,12 @@ func DemystifyPending(status v1.PodStatus) (pluginsCore.PhaseInfo, error) {
 								// ErrImagePull -> Transitionary phase to ImagePullBackOff
 								// ContainerCreating -> Image is being downloaded
 								// PodInitializing -> Init containers are running
-								return pluginsCore.PhaseInfoInitializing(c.LastTransitionTime.Time, pluginsCore.DefaultPhaseVersion, fmt.Sprintf("%s:%s", c.Reason, c.Message)), nil
+								return pluginsCore.PhaseInfoInitializing(c.LastTransitionTime.Time, pluginsCore.DefaultPhaseVersion, fmt.Sprintf("[%s]: %s", finalReason, finalMessage)), nil
 
 							case "CreateContainerError":
 								// This happens if for instance the command to the container is incorrect, ie doesn't run
 								t := c.LastTransitionTime.Time
-								return pluginsCore.PhaseInfoFailure(c.Reason, c.Message, &pluginsCore.TaskInfo{
+								return pluginsCore.PhaseInfoFailure(finalReason, finalMessage, &pluginsCore.TaskInfo{
 									OccurredAt: &t,
 								}), nil
 
@@ -168,7 +170,7 @@ func DemystifyPending(status v1.PodStatus) (pluginsCore.PhaseInfo, error) {
 								// So be default if the container is not waiting with the PodInitializing/ContainerCreating
 								// reasons, then we will assume a failure reason, and fail instantly
 								t := c.LastTransitionTime.Time
-								return pluginsCore.PhaseInfoSystemRetryableFailure(c.Reason, c.Message, &pluginsCore.TaskInfo{
+								return pluginsCore.PhaseInfoSystemRetryableFailure(finalReason, finalMessage, &pluginsCore.TaskInfo{
 									OccurredAt: &t,
 								}), nil
 							}

go/tasks/pluginmachinery/flytek8s/pod_helper_test.go

@@ -2,6 +2,9 @@ package flytek8s
 
 import (
 	"context"
+	"encoding/json"
+	"io/ioutil"
+	"path/filepath"
 	"testing"
 
 	config1 "github.com/lyft/flytestdlib/config"
@@ -443,3 +446,39 @@ func TestConvertPodFailureToError(t *testing.T) {
 		assert.Equal(t, code, "OOMKilled")
 	})
 }
+
+func TestDemystifyPending_testcases(t *testing.T) {
+
+	tests := []struct {
+		name     string
+		filename string
+		isErr    bool
+		errCode  string
+		message  string
+	}{
+		{"ImagePullBackOff", "imagepull-failurepod.json", false, "ContainersNotReady|ImagePullBackOff", "containers with unready status: [fdf98e4ed2b524dc3bf7-get-flyte-id-task-0]|Back-off pulling image \"image\""},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			testFile := filepath.Join("testdata", tt.filename)
+			data, err := ioutil.ReadFile(testFile)
+			assert.NoError(t, err, "failed to read file %s", testFile)
+			pod := &v1.Pod{}
+			if assert.NoError(t, json.Unmarshal(data, pod), "failed to unmarshal json in %s. Expected of type v1.Pod", testFile) {
+				p, err := DemystifyPending(pod.Status)
+				if tt.isErr {
+					assert.Error(t, err, "Error expected from method")
+				} else {
+					assert.NoError(t, err, "Error not expected")
+					assert.NotNil(t, p)
+					assert.Equal(t, p.Phase(), pluginsCore.PhaseRetryableFailure)
+					if assert.NotNil(t, p.Err()) {
+						assert.Equal(t, p.Err().Code, tt.errCode)
+						assert.Equal(t, p.Err().Message, tt.message)
+					}
+				}
+			}
+		})
+	}
+}