tabletmanager/grpctmclient: use vtrpcpb/gRPC error codes

changelog/23.0/23.0.0/summary.md

@@ -20,6 +20,7 @@
         - [CLI Flags](#flags-vttablet)
         - [Managed MySQL configuration defaults to caching-sha2-password](#mysql-caching-sha2-password) 
         - [MySQL timezone environment propagation](#mysql-timezone-env)
+        - [gRPC `tabletmanager` client error changes](#grpctmclient-err-changes)
     - **[Docker](#docker)**
 
 ## <a id="major-changes"/>Major Changes</a>
@@ -123,6 +124,10 @@ Fixed a bug where environment variables like `TZ` were not propagated from mysql
 As a result, timezone settings from the environment were previously ignored. Now mysqld correctly inherits environment variables.
 ⚠️ Deployments that relied on the old behavior and explicitly set a non-UTC timezone may see changes in how DATETIME values are interpreted. To preserve compatibility, set `TZ=UTC` explicitly in MySQL pods.
 
+#### <a id="grpctmclient-err-changes"/>gRPC `tabletmanager` client error changes</a>
+
+The `vttablet` gRPC `tabletmanager` client now returns errors wrapped by the internal `go/vt/vterrors` package. External automation relying on google-gRPC error codes should now use `vterrors.Code(err)` to inspect the code of an error, which returns `vtrpcpb.Code`s defined in [the `proto/vtrpc.proto` protobuf](https://github.com/vitessio/vitess/blob/main/proto/vtrpc.proto#L60).
+
 ### <a id="docker"/>Docker</a>
 
 [Bullseye went EOL 1 year ago](https://www.debian.org/releases/), so starting from v23, we will no longer build or publish images based on debian:bullseye.

go/mysql/sqlerror/sql_error.go

@@ -93,6 +93,67 @@ func (se *SQLError) SQLState() string {
 	return se.State
 }
 
+// VtRpcErrorCode returns the vtrpcpb.Code for the error.
+func (se *SQLError) VtRpcErrorCode() vtrpcpb.Code {
+	switch se.Number() {
+	case ERNotSupportedYet:
+		return vtrpcpb.Code_UNIMPLEMENTED
+	case ERDiskFull, EROutOfMemory, EROutOfSortMemory, ERConCount, EROutOfResources, ERRecordFileFull, ERHostIsBlocked,
+		ERCantCreateThread, ERTooManyDelayedThreads, ERNetPacketTooLarge, ERTooManyUserConnections, ERLockTableFull, ERUserLimitReached:
+		return vtrpcpb.Code_RESOURCE_EXHAUSTED
+	case ERLockWaitTimeout:
+		return vtrpcpb.Code_DEADLINE_EXCEEDED
+	case CRServerGone, ERServerShutdown, ERServerIsntAvailable, CRConnectionError, CRConnHostError:
+		return vtrpcpb.Code_UNAVAILABLE
+	case ERFormNotFound, ERKeyNotFound, ERBadFieldError, ERNoSuchThread, ERUnknownTable, ERCantFindUDF, ERNonExistingGrant,
+		ERNoSuchTable, ERNonExistingTableGrant, ERKeyDoesNotExist:
+		return vtrpcpb.Code_NOT_FOUND
+	case ERDBAccessDenied, ERAccessDeniedError, ERKillDenied, ERNoPermissionToCreateUsers:
+		return vtrpcpb.Code_PERMISSION_DENIED
+	case ERNoDb, ERNoSuchIndex, ERCantDropFieldOrKey, ERTableNotLockedForWrite, ERTableNotLocked, ERTooBigSelect, ERNotAllowedCommand,
+		ERTooLongString, ERDelayedInsertTableLocked, ERDupUnique, ERRequiresPrimaryKey, ERCantDoThisDuringAnTransaction, ERReadOnlyTransaction,
+		ERCannotAddForeign, ERNoReferencedRow, ERRowIsReferenced, ERCantUpdateWithReadLock, ERNoDefault, EROperandColumns,
+		ERSubqueryNo1Row, ERNonUpdateableTable, ERFeatureDisabled, ERDuplicatedValueInType, ERRowIsReferenced2,
+		ErNoReferencedRow2, ERWarnDataOutOfRange, ERInnodbIndexCorrupt:
+		return vtrpcpb.Code_FAILED_PRECONDITION
+	case EROptionPreventsStatement:
+		return vtrpcpb.Code_CLUSTER_EVENT
+	case ERTableExists, ERDupEntry, ERFileExists, ERUDFExists:
+		return vtrpcpb.Code_ALREADY_EXISTS
+	case ERGotSignal, ERForcingClose, ERAbortingConnection, ERLockDeadlock:
+		// For ERLockDeadlock, a deadlock rolls back the transaction.
+		return vtrpcpb.Code_ABORTED
+	case ERUnknownComError, ERBadNullError, ERBadDb, ERBadTable, ERNonUniq, ERWrongFieldWithGroup, ERWrongGroupField,
+		ERWrongSumSelect, ERWrongValueCount, ERTooLongIdent, ERDupFieldName, ERDupKeyName, ERWrongFieldSpec, ERParseError,
+		EREmptyQuery, ERNonUniqTable, ERInvalidDefault, ERMultiplePriKey, ERTooManyKeys, ERTooManyKeyParts, ERTooLongKey,
+		ERKeyColumnDoesNotExist, ERBlobUsedAsKey, ERTooBigFieldLength, ERWrongAutoKey, ERWrongFieldTerminators, ERBlobsAndNoTerminated,
+		ERTextFileNotReadable, ERWrongSubKey, ERCantRemoveAllFields, ERUpdateTableUsed, ERNoTablesUsed, ERTooBigSet,
+		ERBlobCantHaveDefault, ERWrongDbName, ERWrongTableName, ERUnknownProcedure, ERWrongParamCountToProcedure,
+		ERWrongParametersToProcedure, ERFieldSpecifiedTwice, ERInvalidGroupFuncUse, ERTableMustHaveColumns, ERUnknownCharacterSet,
+		ERTooManyTables, ERTooManyFields, ERTooBigRowSize, ERWrongOuterJoin, ERNullColumnInIndex, ERFunctionNotDefined,
+		ERWrongValueCountOnRow, ERInvalidUseOfNull, ERRegexpError, ERMixOfGroupFuncAndFields, ERIllegalGrantForTable, ERSyntaxError,
+		ERWrongColumnName, ERWrongKeyColumn, ERBlobKeyWithoutLength, ERPrimaryCantHaveNull, ERTooManyRows, ERUnknownSystemVariable,
+		ERSetConstantsOnly, ERWrongArguments, ERWrongUsage, ERWrongNumberOfColumnsInSelect, ERDupArgument, ERLocalVariable,
+		ERGlobalVariable, ERWrongValueForVar, ERWrongTypeForVar, ERVarCantBeRead, ERCantUseOptionHere, ERIncorrectGlobalLocalVar,
+		ERWrongFKDef, ERKeyRefDoNotMatchTableRef, ERCyclicReference, ERCollationCharsetMismatch, ERCantAggregate2Collations,
+		ERCantAggregate3Collations, ERCantAggregateNCollations, ERVariableIsNotStruct, ERUnknownCollation, ERWrongNameForIndex,
+		ERWrongNameForCatalog, ERBadFTColumn, ERTruncatedWrongValue, ERTooMuchAutoTimestampCols, ERInvalidOnUpdate, ERUnknownTimeZone,
+		ERInvalidCharacterString, ERIllegalReference, ERDerivedMustHaveAlias, ERTableNameNotAllowedHere, ERDataTooLong, ERDataOutOfRange,
+		ERTruncatedWrongValueForField, ERIllegalValueForType, ERWrongValue, ERWrongParamcountToNativeFct:
+		return vtrpcpb.Code_INVALID_ARGUMENT
+	case ERSpecifiedAccessDenied:
+		if strings.Contains(se.Message, "failover in progress") {
+			return vtrpcpb.Code_FAILED_PRECONDITION
+		}
+		return vtrpcpb.Code_PERMISSION_DENIED
+	case CRServerLost:
+		// Query was killed.
+		return vtrpcpb.Code_CANCELED
+	default:
+		return vterrors.Code(se)
+	}
+}
+
 var errExtract = regexp.MustCompile(`\(errno ([0-9]*)\) \(sqlstate ([0-9a-zA-Z]{5})\)`)
 
 // NewSQLErrorFromError returns a *SQLError from the provided error.

go/test/endtoend/tabletmanager/tablet_test.go

@@ -21,12 +21,17 @@ import (
 	"fmt"
 	"strconv"
 	"testing"
+	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
 	"vitess.io/vitess/go/test/endtoend/cluster"
 	"vitess.io/vitess/go/vt/log"
+	"vitess.io/vitess/go/vt/vterrors"
+	tmc "vitess.io/vitess/go/vt/vttablet/grpctmclient"
+
+	vtrpcpb "vitess.io/vitess/go/vt/proto/vtrpc"
 )
 
 // TestEnsureDB tests that vttablet creates the db as needed
@@ -64,6 +69,38 @@ func TestEnsureDB(t *testing.T) {
 	killTablets(tablet)
 }
 
+// TestGRPCErrorCode_UNAVAILABLE tests that vttablet returns correct gRPC codes,
+// in this case codes.Unavailable/vtrpcpb.Code_UNAVAILABLE when mysqld is down.
+func TestGRPCErrorCode_UNAVAILABLE(t *testing.T) {
+	// Create new tablet
+	tablet := clusterInstance.NewVttabletInstance("replica", 0, "")
+	defer killTablets(tablet)
+	mysqlctlProcess, err := cluster.MysqlCtlProcessInstance(tablet.TabletUID, tablet.MySQLPort, clusterInstance.TmpDirectory)
+	require.NoError(t, err)
+
+	tablet.MysqlctlProcess = *mysqlctlProcess
+	err = tablet.MysqlctlProcess.Start()
+	require.NoError(t, err)
+
+	log.Info(fmt.Sprintf("Started vttablet %v", tablet))
+	// Start vttablet process as replica. It won't be able to serve because there's no db.
+	err = clusterInstance.StartVttablet(tablet, false, "SERVING", false, cell, "dbtest", hostname, "0")
+	require.NoError(t, err)
+
+	// kill the mysql process
+	err = tablet.MysqlctlProcess.Stop()
+	require.NoError(t, err)
+
+	// confirm we get vtrpcpb.Code_UNAVAILABLE when calling FullStatus,
+	// because this will try and fail to connect to mysql
+	ctx, cancel := context.WithTimeout(context.Background(), time.Second*10)
+	defer cancel()
+	tmClient := tmc.NewClient()
+	vttablet := getTablet(tablet.GrpcPort)
+	_, err = tmClient.FullStatus(ctx, vttablet)
+	assert.Equal(t, vtrpcpb.Code_UNAVAILABLE, vterrors.Code(err))
+}
+
 // TestResetReplicationParameters tests that the RPC ResetReplicationParameters works as intended.
 func TestResetReplicationParameters(t *testing.T) {