Merge pull request #4166 from kubernetes-sigs/fix/softerror-should-no…

…t-fatal

fix: provide possibility to have a soft error mode

controller/controller.go

@@ -18,6 +18,7 @@ package controller
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"sync"
 	"time"
@@ -331,7 +332,11 @@ func (c *Controller) Run(ctx context.Context) {
 	for {
 		if c.ShouldRunOnce(time.Now()) {
 			if err := c.RunOnce(ctx); err != nil {
-				log.Fatal(err)
+				if errors.Is(err, provider.SoftError) {
+					log.Errorf("Failed to do run once: %v", err)
+				} else {
+					log.Fatalf("Failed to do run once: %v", err)
+				}
 			}
 		}
 		select {

provider/aws/aws.go

@@ -27,7 +27,6 @@ import (
 	"github.com/aws/aws-sdk-go/aws"
 	"github.com/aws/aws-sdk-go/aws/request"
 	"github.com/aws/aws-sdk-go/service/route53"
-	"github.com/pkg/errors"
 	log "github.com/sirupsen/logrus"
 
 	"sigs.k8s.io/external-dns/endpoint"
@@ -325,10 +324,10 @@ func (p *AWSProvider) Zones(ctx context.Context) (map[string]*route53.HostedZone
 
 	err := p.client.ListHostedZonesPagesWithContext(ctx, &route53.ListHostedZonesInput{}, f)
 	if err != nil {
-		return nil, errors.Wrap(err, "failed to list hosted zones")
+		return nil, provider.NewSoftError(fmt.Errorf("failed to list hosted zones: %w", err))
 	}
 	if tagErr != nil {
-		return nil, errors.Wrap(tagErr, "failed to list zones tags")
+		return nil, provider.NewSoftError(fmt.Errorf("failed to list zones tags: %w", tagErr))
 	}
 
 	for _, zone := range zones {
@@ -353,7 +352,7 @@ func wildcardUnescape(s string) string {
 func (p *AWSProvider) Records(ctx context.Context) (endpoints []*endpoint.Endpoint, _ error) {
 	zones, err := p.Zones(ctx)
 	if err != nil {
-		return nil, errors.Wrap(err, "records retrieval failed")
+		return nil, provider.NewSoftError(fmt.Errorf("records retrieval failed: %w", err))
 	}
 
 	return p.records(ctx, zones)
@@ -445,7 +444,7 @@ func (p *AWSProvider) records(ctx context.Context, zones map[string]*route53.Hos
 		}
 
 		if err := p.client.ListResourceRecordSetsPagesWithContext(ctx, params, f); err != nil {
-			return nil, errors.Wrapf(err, "failed to list resource records sets for zone %s", *z.Id)
+			return nil, provider.NewSoftError(fmt.Errorf("failed to list resource records sets for zone %s: %w", *z.Id, err))
 		}
 	}
 
@@ -530,7 +529,7 @@ func (p *AWSProvider) GetDomainFilter() endpoint.DomainFilter {
 func (p *AWSProvider) ApplyChanges(ctx context.Context, changes *plan.Changes) error {
 	zones, err := p.Zones(ctx)
 	if err != nil {
-		return errors.Wrap(err, "failed to list zones, not applying changes")
+		return provider.NewSoftError(fmt.Errorf("failed to list zones, not applying changes: %w", err))
 	}
 
 	updateChanges := p.createUpdateChanges(changes.UpdateNew, changes.UpdateOld)
@@ -632,7 +631,7 @@ func (p *AWSProvider) submitChanges(ctx context.Context, changes Route53Changes,
 	}
 
 	if len(failedZones) > 0 {
-		return errors.Errorf("failed to submit all changes for the following zones: %v", failedZones)
+		return provider.NewSoftError(fmt.Errorf("failed to submit all changes for the following zones: %v", failedZones))
 	}
 
 	return nil
@@ -847,7 +846,7 @@ func (p *AWSProvider) tagsForZone(ctx context.Context, zoneID string) (map[strin
 		ResourceId:   aws.String(zoneID),
 	})
 	if err != nil {
-		return nil, errors.Wrapf(err, "failed to list tags for zone %s", zoneID)
+		return nil, provider.NewSoftError(fmt.Errorf("failed to list tags for zone %s: %w", zoneID, err))
 	}
 	tagMap := map[string]string{}
 	for _, tag := range response.ResourceTagSet.Tags {

provider/provider.go

@@ -18,13 +18,24 @@ package provider
 
 import (
 	"context"
+	"errors"
 	"net"
 	"strings"
 
 	"sigs.k8s.io/external-dns/endpoint"
 	"sigs.k8s.io/external-dns/plan"
 )
 
+// SoftError is an error, that provider will only log as error instead
+// of fatal. It is meant for error propagation from providers to tell
+// that this is a transient error.
+var SoftError error = errors.New("soft error")
+
+// NewSoftError creates a SoftError from the given error
+func NewSoftError(err error) error {
+	return errors.Join(SoftError, err)
+}
+
 // Provider defines the interface DNS providers should implement.
 type Provider interface {
 	Records(ctx context.Context) ([]*endpoint.Endpoint, error)