s3tables, iceberg: log client refusals at V(1), not ERROR (#11896)

* s3tables, iceberg: log client refusals at V(1), not ERROR

Requests the S3 Tables and Iceberg REST APIs correctly refuse with a 4xx
(access denied, not authorized, no such namespace, already exists, not
empty, a commit conflict) were logged with glog.Errorf, so a client probing
for what it may do, or retrying an optimistic commit, filled the ERROR log
and hid real faults from anyone alerting on it.

- s3tables: HandleRequest wraps the ResponseWriter in a statusRecorder (with
  Unwrap for http.ResponseController) and logs a handler error by the status
  it answered: 4xx at V(1) as "<op> refused (<status>)", 5xx or no status at
  ERROR as before.
- s3tables: the "handleCreateNamespace called" trace moves from ERROR to V(3).
- iceberg: writeManagerError's mapping is split into classifyManagerError,
  shared with a new logManagerError that logs 4xx-mapped errors at V(1) and
  everything else at ERROR; used by CreateNamespace, CommitTable and
  CommitTransaction.
- storage: loading a remote-tier volume logged
  "readSuperBlock remote volume N: <nil>" at ERROR even when the read
  succeeded; it is now logged only when there is an error.

Tests: TestRequestRefusedWith4xxIsNotLoggedAsError, TestLogRequestErrorByStatus,
TestStatusRecorderKeepsTheFirstStatus, TestStatusRecorderUnwrapsForResponseController,
TestNoTraceLoggedAtErrorLevel (fails on master on the CreateNamespace trace),
TestLogManagerErrorRefusalsAreNotErrors, TestRemoteSuperBlockLogOnlyOnError.

* s3tables: do not log a malformed CreateNamespace body at ERROR

handleCreateNamespace logged a body it could not decode at ERROR before
answering 400; the dispatcher already logs that refusal once at V(1).

Also drop statusRecorder.Flush: it made the recorder an http.Flusher even
over a writer that cannot flush, so http.NewResponseController reported
success instead of ErrNotSupported. Unwrap already lets the controller
reach the wrapped writer.

* iceberg: require exactly one log line per classified manager error

The test accepted a refusal that produced no log, or both logs. Require
one V(1) refusal and no ERROR for a 4xx, and the reverse for a fault.

---------

Co-authored-by: Javier de la Torre <jatorre@lakehousebox.com>

weed/s3api/iceberg/handlers_commit.go

@@ -397,7 +397,7 @@ func (s *Server) handleUpdateTable(w http.ResponseWriter, r *http.Request) {
 		if cleanupErr := s.deleteMetadataFile(r.Context(), metadataBucket, metadataPath, metadataFileName); cleanupErr != nil {
 			glog.V(1).Infof("Iceberg: failed to cleanup metadata file %s after update failure: %v", newMetadataLocation, cleanupErr)
 		}
-		glog.Errorf("Iceberg: CommitTable UpdateTable error: %v", err)
+		logManagerError("CommitTable UpdateTable", err)
 		writeManagerError(w, err)
 		return
 	}

weed/s3api/iceberg/handlers_namespace.go

@@ -140,7 +140,7 @@ func (s *Server) handleCreateNamespace(w http.ResponseWriter, r *http.Request) {
 			writeError(w, http.StatusConflict, "AlreadyExistsException", err.Error())
 			return
 		}
-		glog.Errorf("Iceberg: CreateNamespace error: %v", err)
+		logManagerError("CreateNamespace", err)
 		writeManagerError(w, err)
 		return
 	}

weed/s3api/iceberg/handlers_transaction.go

@@ -109,7 +109,7 @@ func (s *Server) handleCommitTransaction(w http.ResponseWriter, r *http.Request)
 				writeError(w, http.StatusConflict, "CommitFailedException", "Version token mismatch")
 				return
 			}
-			glog.Errorf("Iceberg: CommitTransaction UpdateTable error: %v", err)
+			logManagerError("CommitTransaction UpdateTable", err)
 			writeManagerError(w, err)
 			return
 		}

weed/s3api/iceberg/log_manager_error_test.go

@@ -0,0 +1,36 @@
+package iceberg
+
+import (
+	"errors"
+	"fmt"
+	"testing"
+
+	"github.com/seaweedfs/seaweedfs/weed/s3api/s3tables"
+)
+
+func TestLogManagerErrorRefusalsAreNotErrors(t *testing.T) {
+	var refusals, faults []string
+	oldR, oldF := logManagerRefusal, logManagerFault
+	logManagerRefusal = func(f string, a ...interface{}) { refusals = append(refusals, fmt.Sprintf(f, a...)) }
+	logManagerFault = func(f string, a ...interface{}) { faults = append(faults, fmt.Sprintf(f, a...)) }
+	defer func() { logManagerRefusal, logManagerFault = oldR, oldF }()
+	for _, c := range []struct {
+		err   error
+		fault bool
+	}{
+		{&s3tables.S3TablesError{Type: s3tables.ErrCodeAccessDenied, Message: "not authorized to create namespace in this bucket"}, false},
+		{&s3tables.S3TablesError{Type: s3tables.ErrCodeNoSuchNamespace, Message: "no such namespace"}, false},
+		{&s3tables.S3TablesError{Type: s3tables.ErrCodeNamespaceAlreadyExists, Message: "exists"}, false},
+		{errors.New("filer: connection refused"), true}, // positive control: a real fault stays at ERROR
+	} {
+		refusals, faults = nil, nil
+		logManagerError("CreateNamespace", c.err)
+		wantFaults, wantRefusals := 0, 1
+		if c.fault {
+			wantFaults, wantRefusals = 1, 0
+		}
+		if len(faults) != wantFaults || len(refusals) != wantRefusals {
+			t.Errorf("%v: %d ERROR and %d refusal log(s), want %d and %d (refusals %v, faults %v)", c.err, len(faults), len(refusals), wantFaults, wantRefusals, refusals, faults)
+		}
+	}
+}

weed/s3api/iceberg/utils.go

@@ -158,7 +158,9 @@ func nameValidationError(err error) bool {
 // errors (400) and must carry the reason -- a variant field without
 // format-version 3 otherwise reads as a bare 500 with "variant is not supported
 // until v3" only in the log. Anything else is a server fault (500).
-func writeManagerError(w http.ResponseWriter, err error) {
+// classifyManagerError maps a storage-layer error to the HTTP status, the
+// Iceberg error type and the message the catalog answers with.
+func classifyManagerError(err error) (int, string, string) {
 	switch {
 	case nameValidationError(err),
 		errors.Is(err, iceberg.ErrInvalidSchema),
@@ -166,8 +168,7 @@ func writeManagerError(w http.ResponseWriter, err error) {
 		errors.Is(err, iceberg.ErrInvalidTypeString),
 		errors.Is(err, iceberg.ErrInvalidTransform),
 		errors.Is(err, iceberg.ErrInvalidArgument):
-		writeError(w, http.StatusBadRequest, "BadRequestException", err.Error())
-		return
+		return http.StatusBadRequest, "BadRequestException", err.Error()
 	}
 	// Storage-layer failures are mostly the client's, not the server's. Reporting
 	// a missing namespace or a name conflict as a 500 makes the catalog look
@@ -179,31 +180,49 @@ func writeManagerError(w http.ResponseWriter, err error) {
 			// The storage-layer message names the resolved bucket, which for a
 			// client that sent no warehouse at all is the default one it never
 			// asked for, so say how to select a real table bucket.
-			writeError(w, http.StatusNotFound, "NoSuchNamespaceException",
-				fmt.Sprintf("%s: each table bucket is a separate catalog, select one with warehouse=s3://<table-bucket>/ or /v1/<table-bucket>/", tableErr.Message))
+			return http.StatusNotFound, "NoSuchNamespaceException",
+				fmt.Sprintf("%s: each table bucket is a separate catalog, select one with warehouse=s3://<table-bucket>/ or /v1/<table-bucket>/", tableErr.Message)
 		case s3tables.ErrCodeNoSuchNamespace:
-			writeError(w, http.StatusNotFound, "NoSuchNamespaceException", tableErr.Message)
+			return http.StatusNotFound, "NoSuchNamespaceException", tableErr.Message
 		case s3tables.ErrCodeNoSuchTable:
-			writeError(w, http.StatusNotFound, "NoSuchTableException", tableErr.Message)
+			return http.StatusNotFound, "NoSuchTableException", tableErr.Message
 		case s3tables.ErrCodeNoSuchView:
-			writeError(w, http.StatusNotFound, "NoSuchViewException", tableErr.Message)
+			return http.StatusNotFound, "NoSuchViewException", tableErr.Message
 		case s3tables.ErrCodeNamespaceAlreadyExists, s3tables.ErrCodeBucketAlreadyExists,
 			s3tables.ErrCodeTableAlreadyExists, s3tables.ErrCodeViewAlreadyExists:
-			writeError(w, http.StatusConflict, "AlreadyExistsException", tableErr.Message)
+			return http.StatusConflict, "AlreadyExistsException", tableErr.Message
 		case s3tables.ErrCodeNamespaceNotEmpty, s3tables.ErrCodeBucketNotEmpty:
-			writeError(w, http.StatusConflict, "AlreadyExistsException", tableErr.Message)
+			return http.StatusConflict, "AlreadyExistsException", tableErr.Message
 		case s3tables.ErrCodeConflict:
-			writeError(w, http.StatusConflict, "CommitFailedException", tableErr.Message)
+			return http.StatusConflict, "CommitFailedException", tableErr.Message
 		case s3tables.ErrCodeAccessDenied:
-			writeError(w, http.StatusForbidden, "ForbiddenException", tableErr.Message)
+			return http.StatusForbidden, "ForbiddenException", tableErr.Message
 		case s3tables.ErrCodeInvalidRequest, s3tables.ErrCodeInvalidIcebergLayout:
-			writeError(w, http.StatusBadRequest, "BadRequestException", tableErr.Message)
-		default:
-			writeError(w, http.StatusInternalServerError, "InternalServerError", err.Error())
+			return http.StatusBadRequest, "BadRequestException", tableErr.Message
 		}
+	}
+	return http.StatusInternalServerError, "InternalServerError", err.Error()
+}
+
+func writeManagerError(w http.ResponseWriter, err error) {
+	status, typ, msg := classifyManagerError(err)
+	writeError(w, status, typ, msg)
+}
+
+// A storage-layer error the catalog answers with a 4xx (access denied, not
+// found, already exists, not empty, a conflict) is the API refusing a request
+// as intended, not a server fault; it is logged at V(1). A 5xx stays at ERROR.
+var (
+	logManagerRefusal = func(format string, args ...interface{}) { glog.V(1).Infof(format, args...) }
+	logManagerFault   = glog.Errorf
+)
+
+func logManagerError(op string, err error) {
+	if status, _, _ := classifyManagerError(err); status >= 400 && status < 500 {
+		logManagerRefusal("Iceberg: %s refused (%d): %v", op, status, err)
 		return
 	}
-	writeError(w, http.StatusInternalServerError, "InternalServerError", err.Error())
+	logManagerFault("Iceberg: %s error: %v", op, err)
 }
 
 // resolveWarehouseBucket maps a client-supplied warehouse value to a table