status: Hide generic operator status in favor of more specific errors

When upgrading and a sync loop exits, we often see:

* Operator A not upgraded yet
* Operator B not upgraded yet
* Operator C not upgraded yet
* Some more specific error
* Operator D not upgraded yet

Since the not available messages carry little value and are expected
when we are processing parallel output, if we have one or more errors
during sync that are not the generic error, only report those. This
focuses the user's attention on the actual problem, rather than
distracting them with noise. In the ideal case, the user immediately
sees "router is totally broken" vs "a, b, c, d, e, router is broken, f".

Author: smarterclayton

pkg/cvo/sync_worker.go

@@ -10,9 +10,9 @@ import (
 	"sync"
 	"time"
 
-	"k8s.io/klog"
 	"github.com/prometheus/client_golang/prometheus"
 	"golang.org/x/time/rate"
+	"k8s.io/klog"
 
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/errors"
@@ -751,12 +751,44 @@ func summarizeTaskGraphErrors(errs []error) error {
 	if len(errs) == 1 {
 		return errs[0]
 	}
+	// hide the generic "not available yet" when there are more specific errors present
+	if filtered := filterErrors(errs, isClusterOperatorNotAvailable); len(filtered) > 0 {
+		return newMultipleError(filtered)
+	}
+	// if we're only waiting for operators, condense the error down to a singleton
 	if err := newClusterOperatorsNotAvailable(errs); err != nil {
 		return err
 	}
 	return newMultipleError(errs)
 }
 
+// filterErrors returns only the errors in errs which are false for all fns.
+func filterErrors(errs []error, fns ...func(err error) bool) []error {
+	var filtered []error
+	for _, err := range errs {
+		if errorMatches(err, fns...) {
+			continue
+		}
+		filtered = append(filtered, err)
+	}
+	return filtered
+}
+
+func errorMatches(err error, fns ...func(err error) bool) bool {
+	for _, fn := range fns {
+		if fn(err) {
+			return true
+		}
+	}
+	return false
+}
+
+// isClusterOperatorNotAvailable returns true if this is a ClusterOperatorNotAvailable error
+func isClusterOperatorNotAvailable(err error) bool {
+	uErr, ok := err.(*payload.UpdateError)
+	return ok && uErr != nil && uErr.Reason == "ClusterOperatorNotAvailable"
+}
+
 // newClusterOperatorsNotAvailable unifies multiple ClusterOperatorNotAvailable errors into
 // a single error. It returns nil if the provided errors are not of the same type.
 func newClusterOperatorsNotAvailable(errs []error) error {