classify postgres slot allocation issue (#3320)

Author: heavycrystal

flow/alerting/classifier.go

@@ -157,6 +157,12 @@ var (
 		// These are unclassified and should not be exposed
 		Class: "OTHER", action: NotifyTelemetry,
 	}
+	// Postgres 16.9/17.5 etc. introduced a bug where certain workloads can cause logical replication to
+	// request a memory allocation of >1GB, which is not allowed by Postgres. Fixed already, but we need to handle this error
+	// https://github.com/postgres/postgres/commit/d87d07b7ad3b782cb74566cd771ecdb2823adf6a
+	ErrorPostgresSlotMemalloc = ErrorClass{
+		Class: "ERROR_POSTGRES_SLOT_MEMALLOC", action: NotifyUser,
+	}
 )
 
 func (e ErrorClass) String() string {
@@ -322,6 +328,10 @@ func GetErrorClass(ctx context.Context, err error) (ErrorClass, ErrorInfo) {
 				return ErrorNotifyConnectivity, pgErrorInfo
 			}
 
+			if strings.Contains(pgErr.Message, "invalid memory alloc request size") {
+				return ErrorPostgresSlotMemalloc, pgErrorInfo
+			}
+
 			// Fall through for other internal errors
 			return ErrorOther, pgErrorInfo
 

flow/alerting/classifier_test.go

@@ -134,6 +134,39 @@ func TestAuroraInternalWALErrorShouldBeRecoverable(t *testing.T) {
 	}, errInfo, "Unexpected error info")
 }
 
+func TestNeonProjectQuotaExceededErrorShouldBeConnectivity(t *testing.T) {
+	// Simulate a Neon project quota exceeded error
+	err := &pgconn.PgError{
+		Severity: "ERROR",
+		Code:     pgerrcode.InternalError,
+		Message:  "Your account or project has exceeded the compute time quota. Upgrade your plan to increase limits.",
+	}
+	errorClass, errInfo := GetErrorClass(t.Context(),
+		exceptions.NewPeerCreateError(fmt.Errorf("failed to create connection: failed to connect to `<user, host>: server error: `: %w", err)))
+	assert.Equal(t, ErrorNotifyConnectivity, errorClass, "Unexpected error class")
+	assert.Equal(t, ErrorInfo{
+		Source: ErrorSourcePostgres,
+		Code:   pgerrcode.InternalError,
+	}, errInfo, "Unexpected error info")
+}
+
+func TestPostgresMemoryAllocErrorShouldBeSlotMemalloc(t *testing.T) {
+	// Simulate a Postgres memory allocation error
+	err := &exceptions.PostgresWalError{
+		Msg: &pgproto3.ErrorResponse{
+			Severity: "ERROR",
+			Code:     pgerrcode.InternalError,
+			Message:  "invalid memory alloc request size 1073741824",
+		},
+	}
+	errorClass, errInfo := GetErrorClass(t.Context(), fmt.Errorf("error in WAL: %w", err))
+	assert.Equal(t, ErrorPostgresSlotMemalloc, errorClass, "Unexpected error class")
+	assert.Equal(t, ErrorInfo{
+		Source: ErrorSourcePostgres,
+		Code:   pgerrcode.InternalError,
+	}, errInfo, "Unexpected error info")
+}
+
 func TestClickHouseAccessEntityNotFoundErrorShouldBeRecoverable(t *testing.T) {
 	// Simulate a ClickHouse access entity not found error
 	for idx, msg := range []string{