# Use of this feature is not recommended, if it can be avoided.
ForwardSlashNameSubstitution: ""
+ # Include "folder objects" in S3 ListObjects responses.
+ S3FolderObjects: true
+
# Managed collection properties. At creation time, if the client didn't
# provide the listed keys, they will be automatically populated following
# one of the following behaviors:
// exists.
var whitelist = map[string]bool{
// | sort -t'"' -k2,2
- "ClusterID": true,
"API": true,
"API.AsyncPermissionsUpdateInterval": false,
"API.DisabledAPIs": false,
+ "API.KeepServiceRequestTimeout": false,
"API.MaxConcurrentRequests": false,
"API.MaxIndexDatabaseRead": false,
"API.MaxItemsPerResponse": true,
"API.MaxRequestSize": true,
"API.RailsSessionSecretToken": false,
"API.RequestTimeout": true,
- "API.WebsocketClientEventQueue": false,
"API.SendTimeout": true,
+ "API.WebsocketClientEventQueue": false,
"API.WebsocketServerEventQueue": false,
- "API.KeepServiceRequestTimeout": false,
"AuditLogs": false,
"AuditLogs.MaxAge": false,
"AuditLogs.MaxDeleteBatch": false,
"AuditLogs.UnloggedAttributes": false,
+ "ClusterID": true,
"Collections": true,
+ "Collections.BalanceCollectionBatch": false,
+ "Collections.BalanceCollectionBuffers": false,
+ "Collections.BalancePeriod": false,
+ "Collections.BalanceTimeout": false,
+ "Collections.BlobDeleteConcurrency": false,
+ "Collections.BlobMissingReport": false,
+ "Collections.BlobReplicateConcurrency": false,
"Collections.BlobSigning": true,
"Collections.BlobSigningKey": false,
"Collections.BlobSigningTTL": true,
"Collections.BlobTrash": false,
- "Collections.BlobTrashLifetime": false,
- "Collections.BlobTrashConcurrency": false,
"Collections.BlobTrashCheckInterval": false,
- "Collections.BlobDeleteConcurrency": false,
- "Collections.BlobReplicateConcurrency": false,
+ "Collections.BlobTrashConcurrency": false,
+ "Collections.BlobTrashLifetime": false,
"Collections.CollectionVersioning": false,
"Collections.DefaultReplication": true,
"Collections.DefaultTrashLifetime": true,
"Collections.ManagedProperties.*": true,
"Collections.ManagedProperties.*.*": true,
"Collections.PreserveVersionIfIdle": true,
+ "Collections.S3FolderObjects": true,
"Collections.TrashSweepInterval": false,
"Collections.TrustAllContent": false,
"Collections.WebDAVCache": false,
- "Collections.BalanceCollectionBatch": false,
- "Collections.BalancePeriod": false,
- "Collections.BalanceTimeout": false,
- "Collections.BlobMissingReport": false,
- "Collections.BalanceCollectionBuffers": false,
"Containers": true,
"Containers.CloudVMs": false,
- "Containers.CrunchRunCommand": false,
"Containers.CrunchRunArgumentsList": false,
+ "Containers.CrunchRunCommand": false,
"Containers.DefaultKeepCacheRAM": true,
"Containers.DispatchPrivateKey": false,
"Containers.JobsAPI": true,
"Login.OpenIDConnect": true,
"Login.OpenIDConnect.ClientID": false,
"Login.OpenIDConnect.ClientSecret": false,
- "Login.OpenIDConnect.Enable": true,
- "Login.OpenIDConnect.Issuer": false,
"Login.OpenIDConnect.EmailClaim": false,
"Login.OpenIDConnect.EmailVerifiedClaim": false,
+ "Login.OpenIDConnect.Enable": true,
+ "Login.OpenIDConnect.Issuer": false,
"Login.OpenIDConnect.UsernameClaim": false,
"Login.PAM": true,
"Login.PAM.DefaultEmailDomain": false,
"Login.PAM.Enable": true,
"Login.PAM.Service": false,
+ "Login.RemoteTokenRefresh": true,
"Login.SSO": true,
"Login.SSO.Enable": true,
"Login.SSO.ProviderAppID": false,
"Login.SSO.ProviderAppSecret": false,
- "Login.RemoteTokenRefresh": true,
"Mail": true,
+ "Mail.EmailFrom": false,
+ "Mail.IssueReporterEmailFrom": false,
+ "Mail.IssueReporterEmailTo": false,
"Mail.MailchimpAPIKey": false,
"Mail.MailchimpListID": false,
"Mail.SendUserSetupNotificationEmail": false,
- "Mail.IssueReporterEmailFrom": false,
- "Mail.IssueReporterEmailTo": false,
"Mail.SupportEmailAddress": true,
- "Mail.EmailFrom": false,
"ManagementToken": false,
"PostgreSQL": false,
"RemoteClusters": true,
"SystemRootToken": false,
"TLS": false,
"Users": true,
- "Users.AnonymousUserToken": true,
"Users.AdminNotifierEmailFrom": false,
+ "Users.AnonymousUserToken": true,
"Users.AutoAdminFirstUser": false,
"Users.AutoAdminUserWithEmail": false,
"Users.AutoSetupNewUsers": false,
"Workbench.EnableGettingStartedPopup": true,
"Workbench.EnablePublicProjectsPage": true,
"Workbench.FileViewersConfigURL": true,
+ "Workbench.InactivePageHTML": true,
"Workbench.LogViewerMaxBytes": true,
"Workbench.MultiSiteSearch": true,
"Workbench.ProfilingEnabled": true,
"Workbench.ShowUserAgreementInline": true,
"Workbench.ShowUserNotifications": true,
"Workbench.SiteName": true,
+ "Workbench.SSHHelpHostSuffix": true,
+ "Workbench.SSHHelpPageHTML": true,
"Workbench.Theme": true,
"Workbench.UserProfileFormFields": true,
"Workbench.UserProfileFormFields.*": true,
"Workbench.UserProfileFormMessage": true,
"Workbench.VocabularyURL": true,
"Workbench.WelcomePageHTML": true,
- "Workbench.InactivePageHTML": true,
- "Workbench.SSHHelpPageHTML": true,
- "Workbench.SSHHelpHostSuffix": true,
}
func redactUnsafe(m map[string]interface{}, mPrefix, lookupPrefix string) error {
# Use of this feature is not recommended, if it can be avoided.
ForwardSlashNameSubstitution: ""
+ # Include "folder objects" in S3 ListObjects responses.
+ S3FolderObjects: true
+
# Managed collection properties. At creation time, if the client didn't
# provide the listed keys, they will be automatically populated following
# one of the following behaviors:
TrashSweepInterval Duration
TrustAllContent bool
ForwardSlashNameSubstitution string
+ S3FolderObjects bool
BlobMissingReport string
BalancePeriod Duration
"git.arvados.org/arvados.git/sdk/go/arvados"
"github.com/coreos/go-systemd/daemon"
"github.com/ghodss/yaml"
+ "github.com/sirupsen/logrus"
log "github.com/sirupsen/logrus"
)
os.Setenv("ARVADOS_API_HOST", cfg.cluster.Services.Controller.ExternalURL.Host)
srv := &server{Config: cfg}
- if err := srv.Start(); err != nil {
+ if err := srv.Start(logrus.StandardLogger()); err != nil {
log.Fatal(err)
}
if _, err := daemon.SdNotify(false, "READY=1"); err != nil {
}
return true
}
+ if err == nil && fi.IsDir() && objectNameGiven && strings.HasSuffix(fspath, "/") && h.Config.cluster.Collections.S3FolderObjects {
+ w.Header().Set("Content-Type", "application/x-directory")
+ w.WriteHeader(http.StatusOK)
+ return true
+ }
if os.IsNotExist(err) ||
(err != nil && err.Error() == "not a directory") ||
(fi != nil && fi.IsDir()) {
http.FileServer(fs).ServeHTTP(w, &r)
return true
case r.Method == "PUT":
- if strings.HasSuffix(r.URL.Path, "/") {
- http.Error(w, "invalid object name (trailing '/' char)", http.StatusBadRequest)
+ if !objectNameGiven {
+ http.Error(w, "missing object name in PUT request", http.StatusBadRequest)
return true
}
fspath := "by_id" + r.URL.Path
- _, err = fs.Stat(fspath)
+ var objectIsDir bool
+ if strings.HasSuffix(fspath, "/") {
+ if !h.Config.cluster.Collections.S3FolderObjects {
+ http.Error(w, "invalid object name: trailing slash", http.StatusBadRequest)
+ return true
+ }
+ n, err := r.Body.Read(make([]byte, 1))
+ if err != nil && err != io.EOF {
+ http.Error(w, fmt.Sprintf("error reading request body: %s", err), http.StatusInternalServerError)
+ return true
+ } else if n > 0 {
+ http.Error(w, "cannot write a non-empty file with a trailing '/' char", http.StatusBadRequest)
+ return true
+ }
+ // Given PUT "foo/bar/", we'll use "foo/bar/."
+ // in the "ensure parents exist" block below,
+ // and then we'll be done.
+ fspath += "."
+ objectIsDir = true
+ }
+ fi, err := fs.Stat(fspath)
if err != nil && err.Error() == "not a directory" {
// requested foo/bar, but foo is a file
http.Error(w, "object name conflicts with existing object", http.StatusBadRequest)
return true
}
- f, err := fs.OpenFile(fspath, os.O_WRONLY|os.O_TRUNC|os.O_CREATE, 0644)
- if os.IsNotExist(err) {
- // create missing intermediate directories, then try again
- for i, c := range fspath {
- if i > 0 && c == '/' {
- dir := fspath[:i]
- if strings.HasSuffix(dir, "/") {
- err = errors.New("invalid object name (consecutive '/' chars)")
- http.Error(w, err.Error(), http.StatusBadRequest)
- return true
- }
- err := fs.Mkdir(dir, 0755)
- if err != nil && err != os.ErrExist {
- err = fmt.Errorf("mkdir %q failed: %w", dir, err)
- http.Error(w, err.Error(), http.StatusInternalServerError)
- return true
- }
- }
- }
- f, err = fs.OpenFile(fspath, os.O_WRONLY|os.O_TRUNC|os.O_CREATE, 0644)
- }
- if err != nil {
- err = fmt.Errorf("open %q failed: %w", r.URL.Path, err)
- http.Error(w, err.Error(), http.StatusBadRequest)
+ if strings.HasSuffix(r.URL.Path, "/") && err == nil && !fi.IsDir() {
+ // requested foo/bar/, but foo/bar is a file
+ http.Error(w, "object name conflicts with existing object", http.StatusBadRequest)
return true
}
- defer f.Close()
- _, err = io.Copy(f, r.Body)
- if err != nil {
- err = fmt.Errorf("write to %q failed: %w", r.URL.Path, err)
- http.Error(w, err.Error(), http.StatusBadGateway)
- return true
+ // create missing parent/intermediate directories, if any
+ for i, c := range fspath {
+ if i > 0 && c == '/' {
+ dir := fspath[:i]
+ if strings.HasSuffix(dir, "/") {
+ err = errors.New("invalid object name (consecutive '/' chars)")
+ http.Error(w, err.Error(), http.StatusBadRequest)
+ return true
+ }
+ err := fs.Mkdir(dir, 0755)
+ if err != nil && err != os.ErrExist {
+ err = fmt.Errorf("mkdir %q failed: %w", dir, err)
+ http.Error(w, err.Error(), http.StatusInternalServerError)
+ return true
+ }
+ }
}
- err = f.Close()
- if err != nil {
- err = fmt.Errorf("write to %q failed: close: %w", r.URL.Path, err)
- http.Error(w, err.Error(), http.StatusBadGateway)
- return true
+ if !objectIsDir {
+ f, err := fs.OpenFile(fspath, os.O_WRONLY|os.O_TRUNC|os.O_CREATE, 0644)
+ if os.IsNotExist(err) {
+ f, err = fs.OpenFile(fspath, os.O_WRONLY|os.O_TRUNC|os.O_CREATE, 0644)
+ }
+ if err != nil {
+ err = fmt.Errorf("open %q failed: %w", r.URL.Path, err)
+ http.Error(w, err.Error(), http.StatusBadRequest)
+ return true
+ }
+ defer f.Close()
+ _, err = io.Copy(f, r.Body)
+ if err != nil {
+ err = fmt.Errorf("write to %q failed: %w", r.URL.Path, err)
+ http.Error(w, err.Error(), http.StatusBadGateway)
+ return true
+ }
+ err = f.Close()
+ if err != nil {
+ err = fmt.Errorf("write to %q failed: close: %w", r.URL.Path, err)
+ http.Error(w, err.Error(), http.StatusBadGateway)
+ return true
+ }
}
err = fs.Sync()
if err != nil {
}
}
-func walkFS(fs arvados.CustomFileSystem, path string, ignoreNotFound bool, fn func(path string, fi os.FileInfo) error) error {
+// Call fn on the given path (directory) and its contents, in
+// lexicographic order.
+//
+// If isRoot==true and path is not a directory, return nil.
+//
+// If fn returns filepath.SkipDir when called on a directory, don't
+// descend into that directory.
+func walkFS(fs arvados.CustomFileSystem, path string, isRoot bool, fn func(path string, fi os.FileInfo) error) error {
+ if isRoot {
+ fi, err := fs.Stat(path)
+ if os.IsNotExist(err) || (err == nil && !fi.IsDir()) {
+ return nil
+ } else if err != nil {
+ return err
+ }
+ err = fn(path, fi)
+ if err == filepath.SkipDir {
+ return nil
+ } else if err != nil {
+ return err
+ }
+ }
f, err := fs.Open(path)
- if os.IsNotExist(err) && ignoreNotFound {
+ if os.IsNotExist(err) && isRoot {
return nil
} else if err != nil {
return fmt.Errorf("open %q: %w", path, err)
}
commonPrefixes := map[string]bool{}
err := walkFS(fs, strings.TrimSuffix(bucketdir+"/"+walkpath, "/"), true, func(path string, fi os.FileInfo) error {
+ if path == bucketdir {
+ return nil
+ }
path = path[len(bucketdir)+1:]
+ filesize := fi.Size()
+ if fi.IsDir() {
+ path += "/"
+ filesize = 0
+ }
if len(path) <= len(params.prefix) {
if path > params.prefix[:len(path)] {
// with prefix "foobar", walking "fooz" means we're done
// with prefix "foobar", walking "foobag" is pointless
return filepath.SkipDir
}
- if fi.IsDir() && !strings.HasPrefix(params.prefix+"/", path+"/") {
+ if fi.IsDir() && !strings.HasPrefix(params.prefix+"/", path) {
// with prefix "foo/bar", walking "fo"
// is pointless (but walking "foo" or
// "foo/bar" is necessary)
if path < params.marker || path < params.prefix {
return nil
}
- if fi.IsDir() {
+ if fi.IsDir() && !h.Config.cluster.Collections.S3FolderObjects {
+ // Note we don't add anything to
+ // commonPrefixes here even if delimiter is
+ // "/". We descend into the directory, and
+ // return a commonPrefix only if we end up
+ // finding a regular file inside it.
return nil
}
if params.delimiter != "" {
// with prefix "foobar" and delimiter
// "z", when we hit "foobar/baz", we
// add "/baz" to commonPrefixes and
- // stop descending (note that even if
- // delimiter is "/" we don't add
- // anything to commonPrefixes when
- // seeing a dir: we wait until we see
- // a file, so we don't incorrectly
- // return results for empty dirs)
+ // stop descending.
commonPrefixes[path[:len(params.prefix)+idx+1]] = true
return filepath.SkipDir
}
resp.Contents = append(resp.Contents, s3.Key{
Key: path,
LastModified: fi.ModTime().UTC().Format("2006-01-02T15:04:05.999") + "Z",
- Size: fi.Size(),
+ Size: filesize,
})
return nil
})
}
func (s *IntegrationSuite) testS3PutObjectSuccess(c *check.C, bucket *s3.Bucket, prefix string) {
for _, trial := range []struct {
- path string
- size int
+ path string
+ size int
+ contentType string
}{
{
- path: "newfile",
- size: 128000000,
+ path: "newfile",
+ size: 128000000,
+ contentType: "application/octet-stream",
+ }, {
+ path: "newdir/newfile",
+ size: 1 << 26,
+ contentType: "application/octet-stream",
}, {
- path: "newdir/newfile",
- size: 1 << 26,
+ path: "newdir1/newdir2/newfile",
+ size: 0,
+ contentType: "application/octet-stream",
}, {
- path: "newdir1/newdir2/newfile",
- size: 0,
+ path: "newdir1/newdir2/newdir3/",
+ size: 0,
+ contentType: "application/x-directory",
},
} {
c.Logf("=== %v", trial)
buf := make([]byte, trial.size)
rand.Read(buf)
- err = bucket.PutReader(objname, bytes.NewReader(buf), int64(len(buf)), "application/octet-stream", s3.Private, s3.Options{})
+ err = bucket.PutReader(objname, bytes.NewReader(buf), int64(len(buf)), trial.contentType, s3.Private, s3.Options{})
c.Check(err, check.IsNil)
rdr, err := bucket.GetReader(objname)
- if !c.Check(err, check.IsNil) {
+ if strings.HasSuffix(trial.path, "/") && !s.testServer.Config.cluster.Collections.S3FolderObjects {
+ c.Check(err, check.NotNil)
+ continue
+ } else if !c.Check(err, check.IsNil) {
continue
}
buf2, err := ioutil.ReadAll(rdr)
s.testS3PutObjectFailure(c, stage.projbucket, stage.coll.Name+"/")
}
func (s *IntegrationSuite) testS3PutObjectFailure(c *check.C, bucket *s3.Bucket, prefix string) {
+ s.testServer.Config.cluster.Collections.S3FolderObjects = false
var wg sync.WaitGroup
for _, trial := range []struct {
path string
stage := s.s3setup(c)
defer stage.teardown(c)
- filesPerDir := 1001
- stage.writeBigDirs(c, 2, filesPerDir)
- s.testS3List(c, stage.collbucket, "", 4000, 2+filesPerDir*2)
- s.testS3List(c, stage.collbucket, "", 131, 2+filesPerDir*2)
- s.testS3List(c, stage.collbucket, "dir0/", 71, filesPerDir)
+ var markers int
+ for markers, s.testServer.Config.cluster.Collections.S3FolderObjects = range []bool{false, true} {
+ dirs := 2
+ filesPerDir := 1001
+ stage.writeBigDirs(c, dirs, filesPerDir)
+ // Total # objects is:
+ // 2 file entries from s3setup (emptyfile and sailboat.txt)
+ // +1 fake "directory" marker from s3setup (emptydir) (if enabled)
+ // +dirs fake "directory" marker from writeBigDirs (dir0/, dir1/) (if enabled)
+ // +filesPerDir*dirs file entries from writeBigDirs (dir0/file0.txt, etc.)
+ s.testS3List(c, stage.collbucket, "", 4000, markers+2+(filesPerDir+markers)*dirs)
+ s.testS3List(c, stage.collbucket, "", 131, markers+2+(filesPerDir+markers)*dirs)
+ s.testS3List(c, stage.collbucket, "dir0/", 71, filesPerDir+markers)
+ }
}
func (s *IntegrationSuite) testS3List(c *check.C, bucket *s3.Bucket, prefix string, pageSize, expectFiles int) {
+ c.Logf("testS3List: prefix=%q pageSize=%d S3FolderObjects=%v", prefix, pageSize, s.testServer.Config.cluster.Collections.S3FolderObjects)
expectPageSize := pageSize
if expectPageSize > 1000 {
expectPageSize = 1000
}
func (s *IntegrationSuite) TestS3CollectionListRollup(c *check.C) {
+ for _, s.testServer.Config.cluster.Collections.S3FolderObjects = range []bool{false, true} {
+ s.testS3CollectionListRollup(c)
+ }
+}
+
+func (s *IntegrationSuite) testS3CollectionListRollup(c *check.C) {
stage := s.s3setup(c)
defer stage.teardown(c)
break
}
}
- c.Check(allfiles, check.HasLen, dirs*filesPerDir+3)
+ markers := 0
+ if s.testServer.Config.cluster.Collections.S3FolderObjects {
+ markers = 1
+ }
+ c.Check(allfiles, check.HasLen, dirs*(filesPerDir+markers)+3+markers)
+
+ gotDirMarker := map[string]bool{}
+ for _, name := range allfiles {
+ isDirMarker := strings.HasSuffix(name, "/")
+ if markers == 0 {
+ c.Check(isDirMarker, check.Equals, false, check.Commentf("name %q", name))
+ } else if isDirMarker {
+ gotDirMarker[name] = true
+ } else if i := strings.LastIndex(name, "/"); i >= 0 {
+ c.Check(gotDirMarker[name[:i+1]], check.Equals, true, check.Commentf("name %q", name))
+ gotDirMarker[name[:i+1]] = true // skip redundant complaints about this dir marker
+ }
+ }
for _, trial := range []struct {
prefix string
delimiter string
marker string
}{
+ {"", "", ""},
{"di", "/", ""},
{"di", "r", ""},
{"di", "n", ""},
{"dir0", "/", ""},
+ {"dir0/", "/", ""},
+ {"dir0/f", "/", ""},
+ {"dir0", "", ""},
+ {"dir0/", "", ""},
+ {"dir0/f", "", ""},
{"dir0", "/", "dir0/file14.txt"}, // no commonprefixes
{"", "", "dir0/file14.txt"}, // middle page, skip walking dir1
{"", "", "dir1/file14.txt"}, // middle page, skip walking dir0
{"dir2", "/", ""}, // prefix "dir2" does not exist
{"", "/", ""},
} {
- c.Logf("\n\n=== trial %+v", trial)
+ c.Logf("\n\n=== trial %+v markers=%d", trial, markers)
maxKeys := 20
resp, err := stage.collbucket.List(trial.prefix, trial.delimiter, trial.marker, maxKeys)
for _, prefix := range resp.CommonPrefixes {
gotPrefixes = append(gotPrefixes, prefix)
}
- c.Check(gotKeys, check.DeepEquals, expectKeys)
- c.Check(gotPrefixes, check.DeepEquals, expectPrefixes)
- c.Check(resp.NextMarker, check.Equals, expectNextMarker)
- c.Check(resp.IsTruncated, check.Equals, expectTruncated)
+ commentf := check.Commentf("trial %+v markers=%d", trial, markers)
+ c.Check(gotKeys, check.DeepEquals, expectKeys, commentf)
+ c.Check(gotPrefixes, check.DeepEquals, expectPrefixes, commentf)
+ c.Check(resp.NextMarker, check.Equals, expectNextMarker, commentf)
+ c.Check(resp.IsTruncated, check.Equals, expectTruncated, commentf)
c.Logf("=== trial %+v keys %q prefixes %q nextMarker %q", trial, gotKeys, gotPrefixes, resp.NextMarker)
}
}
Config *Config
}
-func (srv *server) Start() error {
+func (srv *server) Start(logger *logrus.Logger) error {
h := &handler{Config: srv.Config}
reg := prometheus.NewRegistry()
h.Config.Cache.registry = reg
- ctx := ctxlog.Context(context.Background(), logrus.StandardLogger())
- mh := httpserver.Instrument(reg, nil, httpserver.HandlerWithContext(ctx, httpserver.AddRequestIDs(httpserver.LogRequests(h))))
+ ctx := ctxlog.Context(context.Background(), logger)
+ mh := httpserver.Instrument(reg, logger, httpserver.HandlerWithContext(ctx, httpserver.AddRequestIDs(httpserver.LogRequests(h))))
h.MetricsAPI = mh.ServeAPI(h.Config.cluster.ManagementToken, http.NotFoundHandler())
srv.Handler = mh
var listen arvados.URL
cfg.cluster.ManagementToken = arvadostest.ManagementToken
cfg.cluster.Users.AnonymousUserToken = arvadostest.AnonymousToken
s.testServer = &server{Config: cfg}
- err = s.testServer.Start()
+ err = s.testServer.Start(ctxlog.TestLogger(c))
c.Assert(err, check.Equals, nil)
}
projects / arvados.git / commitdiff