weed/command: skip remote bucket delete when bucket is unmounted (#12418) (#12425)

* weed/command: skip remote bucket delete when bucket is unmounted

* remote_storage: count deleting an already-missing bucket as done

filer.remote.gateway can see the same bucket delete more than once, for
example once per filer when several filers peer their metadata. A copy
that still finds the mount mapping calls DeleteBucket again, and S3
(NoSuchBucket), Azure (ContainerNotFound, ContainerBeingDeleted) and GCS
(404) all failed it, which held the sync offset on the delete. A
missing bucket now counts as deleted, the way deleting a missing object
already does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer.remote.gateway: keep the mapping table writable after mount.mapping is deleted

Deleting mount.mapping replaced the gateway's mappings with a table whose
map was nil, so the next bucket create panicked writing its mount into
it. Reset to an empty map instead.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer.remote.gateway: count a delete of an absent object as done

filer.remote.sync already treats ErrRemoteObjectNotFound from DeleteFile
as deleted, since GCS reports a missing object that way where S3 and
Azure succeed. The gateway returned it, so a delete event seen twice, as
it is when several filers peer their metadata, held the sync offset on
GCS. Use the same deleteRemoteFile.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer.remote.gateway: drop cached remote clients after the new gateway tests

GetRemoteStorage caches clients by configuration, so a repeated run
(-count=2) got the first run's recorder back and saw no calls.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer.remote.gateway: keep the default storage across mount.mapping changes

The default storage for new buckets is derived at startup and kept only
in the gateway's copy of the mappings; mount.mapping never stores it.
Replacing or deleting that file dropped it, so a later bucket create
with no -createBucketAt skipped the remote bucket.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Chris Lu <chrislusf@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
authored and GitHub committed 2026-11-05 05:06:41 -08:00
1 parent c971252871
commit 80f32bc074
10 files changed
+235 -2

No files matched your search

+13 -2
View File
@@ -158,6 +158,11 @@ func (option *RemoteGatewayOptions) makeBucketedEventProcessor(filerSource *sour
client, remoteStorageMountLocation, err := option.findRemoteStorageClient(entry.Name)
if err != nil {
bucket := util.FullPath(option.bucketsDir).Child(entry.Name)
if _, isMounted := option.mappings.Mappings[string(bucket)]; !isMounted {
glog.V(0).Infof("remote bucket %s is not mounted, skipping delete", entry.Name)
return nil
}
return fmt.Errorf("findRemoteStorageClient %s: %v", entry.Name, err)
}
@@ -180,6 +185,9 @@ func (option *RemoteGatewayOptions) makeBucketedEventProcessor(filerSource *sour
if readErr != nil {
return fmt.Errorf("unmarshal mappings: %w", readErr)
}
if newMappings.PrimaryBucketStorageName == "" {
newMappings.PrimaryBucketStorageName = option.mappings.PrimaryBucketStorageName
}
option.mappings = newMappings
}
if strings.HasSuffix(message.NewEntry.Name, filer.REMOTE_STORAGE_CONF_SUFFIX) {
@@ -192,7 +200,10 @@ func (option *RemoteGatewayOptions) makeBucketedEventProcessor(filerSource *sour
} else if metadataEventRemovesFromDirectory(resp, filer.DirectoryEtcRemote) {
// deletion
if message.OldEntry.Name == filer.REMOTE_STORAGE_MOUNT_FILE {
option.mappings = &remote_pb.RemoteStorageMapping{}
option.mappings = &remote_pb.RemoteStorageMapping{
PrimaryBucketStorageName: option.mappings.PrimaryBucketStorageName,
Mappings: map[string]*remote_pb.RemoteStorageLocation{},
}
}
if strings.HasSuffix(message.OldEntry.Name, filer.REMOTE_STORAGE_CONF_SUFFIX) {
conf := &remote_pb.RemoteConf{}
@@ -331,7 +342,7 @@ func (option *RemoteGatewayOptions) makeBucketedEventProcessor(filerSource *sour
return client.RemoveDirectory(dest)
}
glog.V(0).Infof("delete %s", remote_storage.FormatLocation(dest))
return client.DeleteFile(dest)
return deleteRemoteFile(client, dest)
}
if message.OldEntry != nil && message.NewEntry != nil {
if resp.Directory == option.bucketsDir {
@@ -340,3 +340,145 @@ func TestGatewayMetadataOnlyDeleteKeepsTheRemoteObject(t *testing.T) {
t.Fatalf("deletes = %+v, want the ordinary delete", remote.deletes)
}
}
func TestGatewayUnmountedBucketDeleteIsSkipped(t *testing.T) {
remote := &recordingRemote{}
remote_storage.RemoteStorageClientMakers["gatewaytest"] = recordingRemoteMaker{client: remote}
t.Cleanup(func() { delete(remote_storage.RemoteStorageClientMakers, "gatewaytest") })
option := &RemoteGatewayOptions{
bucketsDir: "/buckets",
mappings: &remote_pb.RemoteStorageMapping{Mappings: map[string]*remote_pb.RemoteStorageLocation{}},
remoteConfs: map[string]*remote_pb.RemoteConf{"gatewaytest": {Name: "gatewaytest", Type: "gatewaytest"}},
}
process, err := option.makeBucketedEventProcessor(nil)
if err != nil {
t.Fatal(err)
}
deleteBucketEvent := &filer_pb.SubscribeMetadataResponse{
Directory: "/buckets",
EventNotification: &filer_pb.EventNotification{
OldEntry: &filer_pb.Entry{
Name: "unmounted-bucket",
IsDirectory: true,
},
},
}
// Deleting a bucket that is not mounted (for instance because it was already unmounted
// by an earlier copy of the delete event) should return nil and not error.
if err := process(deleteBucketEvent); err != nil {
t.Fatalf("expected nil for unmounted bucket delete, got: %v", err)
}
}
// TestGatewayCreatesABucketAfterTheMappingFileIsDeleted: deleting
// mount.mapping empties the mapping table, and a later bucket create still
// records its mount in it.
func TestGatewayCreatesABucketAfterTheMappingFileIsDeleted(t *testing.T) {
option, remote := gatewayTestOption("gatewaytest-mapping-reset")
t.Cleanup(func() {
delete(remote_storage.RemoteStorageClientMakers, "gatewaytest-mapping-reset")
remote_storage.ResetRemoteStorageClients()
})
createAt, noFilter, noSuffix := "gatewaytest-mapping-reset", "", false
option.createBucketAt, option.include, option.exclude, option.createBucketRandomSuffix = &createAt, &noFilter, &noFilter, &noSuffix
process, err := option.makeBucketedEventProcessor(nil)
if err != nil {
t.Fatal(err)
}
if err := process(&filer_pb.SubscribeMetadataResponse{
Directory: filer.DirectoryEtcRemote,
EventNotification: &filer_pb.EventNotification{OldEntry: &filer_pb.Entry{Name: filer.REMOTE_STORAGE_MOUNT_FILE}},
}); err != nil {
t.Fatal(err)
}
// saving the new mount needs a filer, so the create itself errors
_ = process(&filer_pb.SubscribeMetadataResponse{
Directory: "/buckets",
EventNotification: &filer_pb.EventNotification{
NewParentPath: "/buckets",
NewEntry: &filer_pb.Entry{Name: "fresh", IsDirectory: true},
},
})
if !slices.Equal(remote.bucketCreates, []string{"fresh"}) {
t.Errorf("bucket creates = %v, want [fresh]", remote.bucketCreates)
}
if loc := option.mappings.Mappings["/buckets/fresh"]; loc.GetBucket() != "fresh" {
t.Errorf("mapping for /buckets/fresh = %v, want bucket fresh", loc)
}
}
// TestGatewayDeleteOfAnAbsentObjectSucceeds: a repeated delete finds the
// object gone, which GCS reports as ErrRemoteObjectNotFound; the event has
// nothing left to do and must not hold the sync offset.
func TestGatewayDeleteOfAnAbsentObjectSucceeds(t *testing.T) {
option, remote := gatewayTestOption("gatewaytest-absent-delete")
t.Cleanup(func() {
delete(remote_storage.RemoteStorageClientMakers, "gatewaytest-absent-delete")
remote_storage.ResetRemoteStorageClients()
})
remote.deleteErr = remote_storage.ErrRemoteObjectNotFound
process, err := option.makeBucketedEventProcessor(nil)
if err != nil {
t.Fatal(err)
}
if err := process(&filer_pb.SubscribeMetadataResponse{
Directory: "/buckets/gold/dir",
EventNotification: &filer_pb.EventNotification{
OldEntry: &filer_pb.Entry{Name: "obj.bin", Attributes: &filer_pb.FuseAttributes{}},
},
}); err != nil {
t.Fatalf("delete of an absent object: %v", err)
}
if len(remote.deletes) != 1 || remote.deletes[0].Path != "/dir/obj.bin" {
t.Errorf("deletes = %+v, want /dir/obj.bin", remote.deletes)
}
}
// TestGatewayKeepsTheDefaultStorageAcrossMappingChanges: the default storage
// for new buckets is derived at startup and never saved in mount.mapping, so
// replacing or deleting that file must not drop it.
func TestGatewayKeepsTheDefaultStorageAcrossMappingChanges(t *testing.T) {
for _, tt := range []struct {
name string
event *filer_pb.EventNotification
}{
{"mapping updated", &filer_pb.EventNotification{NewParentPath: filer.DirectoryEtcRemote, NewEntry: &filer_pb.Entry{Name: filer.REMOTE_STORAGE_MOUNT_FILE}}},
{"mapping deleted", &filer_pb.EventNotification{OldEntry: &filer_pb.Entry{Name: filer.REMOTE_STORAGE_MOUNT_FILE}}},
} {
t.Run(tt.name, func(t *testing.T) {
option, remote := gatewayTestOption("gatewaytest-default-storage")
t.Cleanup(func() {
delete(remote_storage.RemoteStorageClientMakers, "gatewaytest-default-storage")
remote_storage.ResetRemoteStorageClients()
})
option.mappings.PrimaryBucketStorageName = "gatewaytest-default-storage"
createAt, noFilter, noSuffix := "", "", false
option.createBucketAt, option.include, option.exclude, option.createBucketRandomSuffix = &createAt, &noFilter, &noFilter, &noSuffix
process, err := option.makeBucketedEventProcessor(nil)
if err != nil {
t.Fatal(err)
}
if err := process(&filer_pb.SubscribeMetadataResponse{Directory: filer.DirectoryEtcRemote, EventNotification: tt.event}); err != nil {
t.Fatal(err)
}
// saving the new mount needs a filer, so the create itself errors
_ = process(&filer_pb.SubscribeMetadataResponse{
Directory: "/buckets",
EventNotification: &filer_pb.EventNotification{
NewParentPath: "/buckets",
NewEntry: &filer_pb.Entry{Name: "fresh", IsDirectory: true},
},
})
if !slices.Equal(remote.bucketCreates, []string{"fresh"}) {
t.Errorf("bucket creates = %v, want [fresh] in the default storage", remote.bucketCreates)
}
})
}
}
@@ -884,6 +884,7 @@ type recordingRemote struct {
// metadataUpdateResult is the RemoteEntry UpdateFileMetadata reports; nil
// means the update left the object untouched
metadataUpdateResult *filer_pb.RemoteEntry
bucketCreates []string
}
func (r *recordingRemote) WriteFile(loc *remote_pb.RemoteStorageLocation, entry *filer_pb.Entry, reader io.Reader) (*filer_pb.RemoteEntry, error) {
@@ -932,6 +933,11 @@ func (r *recordingRemote) DeleteFile(loc *remote_pb.RemoteStorageLocation) error
return r.deleteErr
}
func (r *recordingRemote) CreateBucket(name string) error {
r.bucketCreates = append(r.bucketCreates, name)
return nil
}
func (r *recordingRemote) UpdateFileMetadata(loc *remote_pb.RemoteStorageLocation, oldEntry, newEntry *filer_pb.Entry) (*filer_pb.RemoteEntry, error) {
r.metadataUpdates = append(r.metadataUpdates, loc)
return r.metadataUpdateResult, nil
@@ -645,6 +645,9 @@ func (az *azureRemoteStorageClient) DeleteBucket(name string) (err error) {
containerClient := az.client.ServiceClient().NewContainerClient(name)
_, err = containerClient.Delete(context.Background(), nil)
if err != nil {
if bloberror.HasCode(err, bloberror.ContainerNotFound, bloberror.ContainerBeingDeleted) {
return nil
}
return fmt.Errorf("delete bucket %s: %w", name, err)
}
return
@@ -483,3 +483,26 @@ func TestAzureReadNotFoundClassification(t *testing.T) {
})
}
}
func TestAzureDeleteBucketOfAMissingContainerSucceeds(t *testing.T) {
tests := []struct {
name string
statusCode int
code string
wantErr bool
}{
{"missing container", http.StatusNotFound, "ContainerNotFound", false},
{"container being deleted", http.StatusConflict, "ContainerBeingDeleted", false},
{"auth failure", http.StatusForbidden, "AuthenticationFailed", true},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
conf := &remote_pb.RemoteConf{Name: "test", AzureAccountName: "testaccount", AzureAccountKey: "dGVzdGtleQ=="}
rs, err := MakeWithHTTPClient(conf, &http.Client{Transport: &azureErrorRoundTripper{statusCode: tt.statusCode, code: tt.code}})
require.NoError(t, err)
err = rs.DeleteBucket("container")
require.Equal(t, tt.wantErr, err != nil, "DeleteBucket: %v", err)
})
}
}
@@ -488,6 +488,10 @@ func (gcs *gcsRemoteStorageClient) CreateBucket(name string) (err error) {
func (gcs *gcsRemoteStorageClient) DeleteBucket(name string) (err error) {
err = gcs.client.Bucket(name).Delete(context.Background())
if err != nil {
var gerr *googleapi.Error
if errors.Is(err, storage.ErrBucketNotExist) || (errors.As(err, &gerr) && gerr.Code == http.StatusNotFound) {
return nil
}
return fmt.Errorf("delete bucket %s: %v", name, err)
}
return
@@ -107,3 +107,22 @@ func TestGCSReadNotFoundClassification(t *testing.T) {
})
}
}
func TestGCSDeleteBucketOfAMissingBucketSucceeds(t *testing.T) {
tests := []struct {
name string
status int
wantErr bool
}{
{"missing bucket", http.StatusNotFound, false},
{"bucket not empty", http.StatusConflict, true},
{"access denied", http.StatusForbidden, true},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
client := (&fakeGCS{readStatus: tt.status}).client(t)
err := client.DeleteBucket("bucket")
require.Equal(t, tt.wantErr, err != nil, "DeleteBucket: %v", err)
})
}
}
+1
View File
@@ -119,6 +119,7 @@ type RemoteStorageClient interface {
DeleteFile(loc *remote_pb.RemoteStorageLocation) (err error)
ListBuckets() ([]*Bucket, error)
CreateBucket(name string) (err error)
// DeleteBucket succeeds when the bucket is already gone.
DeleteBucket(name string) (err error)
}
@@ -1001,6 +1001,10 @@ func (s *s3RemoteStorageClient) DeleteBucket(name string) (err error) {
Bucket: aws.String(name),
})
if err != nil {
var aerr awserr.Error
if errors.As(err, &aerr) && aerr.Code() == s3.ErrCodeNoSuchBucket {
return nil
}
return fmt.Errorf("delete bucket %s: %v", name, err)
}
return
@@ -582,3 +582,23 @@ func TestS3MakeResolvesEnvCredentialsWhenAnonymousDefault(t *testing.T) {
require.Equal(t, "env-access", value.AccessKeyID)
require.Equal(t, "env-secret", value.SecretAccessKey)
}
func TestS3DeleteBucketOfAMissingBucketSucceeds(t *testing.T) {
tests := []struct {
name string
statusCode int
code string
wantErr bool
}{
{"missing bucket", http.StatusNotFound, "NoSuchBucket", false},
{"bucket not empty", http.StatusConflict, "BucketNotEmpty", true},
{"access denied", http.StatusForbidden, "AccessDenied", true},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
client := newS3ClientWithTransport(t, &errorRoundTripper{statusCode: tt.statusCode, code: tt.code}, false)
err := client.DeleteBucket("bucket")
require.Equal(t, tt.wantErr, err != nil, "DeleteBucket: %v", err)
})
}
}