Skip to content

Commit

Permalink
fix
Browse files Browse the repository at this point in the history
Signed-off-by: nolouch <[email protected]>
  • Loading branch information
nolouch committed Jul 4, 2024
1 parent 96d0fc4 commit d072fdd
Show file tree
Hide file tree
Showing 5 changed files with 26 additions and 52 deletions.
22 changes: 4 additions & 18 deletions client/resource_group/controller/config.go
Original file line number Diff line number Diff line change
Expand Up @@ -53,15 +53,12 @@ const (
defaultTargetPeriod = 5 * time.Second
// defaultMaxWaitDuration is the max duration to wait for the token before throwing error.
defaultMaxWaitDuration = 30 * time.Second
<<<<<<< HEAD
=======
// defaultLTBTokenRPCMaxDelay is the upper bound of backoff delay for local token bucket RPC.
defaultLTBTokenRPCMaxDelay = 1 * time.Second
// defaultWaitRetryTimes is the times to retry when waiting for the token.
defaultWaitRetryTimes = 20
// defaultWaitRetryInterval is the interval to retry when waiting for the token.
defaultWaitRetryInterval = 50 * time.Millisecond
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352))
)

const (
Expand All @@ -76,12 +73,7 @@ const (

// Because the resource manager has not been deployed in microservice mode,
// do not enable this function.
<<<<<<< HEAD
defaultDegradedModeWaitDuration = 0
=======
defaultDegradedModeWaitDuration = time.Duration(0)
defaultAvgBatchProportion = 0.7
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352))
)

// TokenRPCParams is the parameters for local bucket RPC.
Expand All @@ -107,12 +99,9 @@ type BaseConfig struct {
// LTBMaxWaitDuration is the max wait time duration for local token bucket.
LTBMaxWaitDuration Duration `toml:"ltb-max-wait-duration" json:"ltb-max-wait-duration"`

<<<<<<< HEAD
=======
// LTBTokenRPCMaxDelay is the upper bound of backoff delay for local token bucket RPC.
LTBTokenRPCMaxDelay Duration `toml:"ltb-token-rpc-max-delay" json:"ltb-token-rpc-max-delay"`

>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352))
// RequestUnit is the configuration determines the coefficients of the RRU and WRU cost.
// This configuration should be modified carefully.
RequestUnit RequestUnitConfig `toml:"request-unit" json:"request-unit"`
Expand Down Expand Up @@ -145,12 +134,6 @@ func (c *Config) Adjust() {
// DefaultConfig returns the default resource manager controller configuration.
func DefaultConfig() *Config {
return &Config{
<<<<<<< HEAD
DegradedModeWaitDuration: NewDuration(defaultDegradedModeWaitDuration),
LTBMaxWaitDuration: NewDuration(defaultMaxWaitDuration),
RequestUnit: DefaultRequestUnitConfig(),
EnableControllerTraceLog: false,
=======
BaseConfig: BaseConfig{
DegradedModeWaitDuration: NewDuration(defaultDegradedModeWaitDuration),
RequestUnit: DefaultRequestUnitConfig(),
Expand All @@ -164,7 +147,6 @@ func DefaultConfig() *Config {
WaitRetryTimes: defaultWaitRetryTimes,
},
},
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352))
}
}

Expand Down Expand Up @@ -212,6 +194,8 @@ type RUConfig struct {

// some config for client
LTBMaxWaitDuration time.Duration
WaitRetryInterval time.Duration
WaitRetryTimes int
DegradedModeWaitDuration time.Duration
}

Expand All @@ -231,6 +215,8 @@ func GenerateRUConfig(config *Config) *RUConfig {
WriteBytesCost: RequestUnit(config.RequestUnit.WriteCostPerByte),
CPUMsCost: RequestUnit(config.RequestUnit.CPUMsCost),
LTBMaxWaitDuration: config.LTBMaxWaitDuration.Duration,
WaitRetryInterval: config.WaitRetryInterval.Duration,
WaitRetryTimes: config.WaitRetryTimes,
DegradedModeWaitDuration: config.DegradedModeWaitDuration.Duration,
}
}
31 changes: 17 additions & 14 deletions client/resource_group/controller/controller.go
Original file line number Diff line number Diff line change
Expand Up @@ -37,8 +37,6 @@ import (

const (
controllerConfigPath = "resource_group/controller"
maxRetry = 10
retryInterval = 50 * time.Millisecond
maxNotificationChanLen = 200
needTokensAmplification = 1.1
trickleReserveDuration = 1250 * time.Millisecond
Expand Down Expand Up @@ -100,6 +98,20 @@ func WithMaxWaitDuration(d time.Duration) ResourceControlCreateOption {
}
}

// WithWaitRetryInterval is the option to set the retry interval when waiting for the token.
func WithWaitRetryInterval(d time.Duration) ResourceControlCreateOption {
return func(controller *ResourceGroupsController) {
controller.ruConfig.WaitRetryInterval = d
}
}

// WithWaitRetryTimes is the option to set the times to retry when waiting for the token.
func WithWaitRetryTimes(times int) ResourceControlCreateOption {
return func(controller *ResourceGroupsController) {
controller.ruConfig.WaitRetryTimes = times
}
}

var _ ResourceGroupKVInterceptor = (*ResourceGroupsController)(nil)

// ResourceGroupsController implements ResourceGroupKVInterceptor.
Expand Down Expand Up @@ -177,23 +189,14 @@ func loadServerConfig(ctx context.Context, provider ResourceGroupProvider) (*Con
if err != nil {
return nil, err
}
<<<<<<< HEAD
if len(resp.Kvs) == 0 {
=======
config := DefaultConfig()
defer config.Adjust()
kvs := resp.GetKvs()
if len(kvs) == 0 {
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352))
log.Warn("[resource group controller] server does not save config, load config failed")
return config, nil
}
<<<<<<< HEAD
config := &Config{}
err = json.Unmarshal(resp.Kvs[0].GetValue(), config)
=======
err = json.Unmarshal(kvs[0].GetValue(), config)
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352))
if err != nil {
return nil, err
}
Expand Down Expand Up @@ -372,7 +375,7 @@ func (c *ResourceGroupsController) Start(ctx context.Context) {
}
for _, item := range resp {
cfgRevision = item.Kv.ModRevision
config := &Config{}
config := DefaultConfig()
if err := json.Unmarshal(item.Kv.Value, config); err != nil {
continue
}
Expand Down Expand Up @@ -1195,7 +1198,7 @@ func (gc *groupCostController) onRequestWait(
var i int
var d time.Duration
retryLoop:
for i = 0; i < maxRetry; i++ {
for i = 0; i < gc.mainCfg.WaitRetryTimes; i++ {
switch gc.mode {
case rmpb.GroupMode_RawMode:
res := make([]*Reservation, 0, len(requestResourceLimitTypeList))
Expand All @@ -1219,7 +1222,7 @@ func (gc *groupCostController) onRequestWait(
}
}
gc.metrics.requestRetryCounter.Inc()
time.Sleep(retryInterval)
time.Sleep(gc.mainCfg.WaitRetryInterval)
}
if err != nil {
if errs.ErrClientResourceGroupThrottled.Equal(err) {
Expand Down
9 changes: 1 addition & 8 deletions pkg/mcs/resource_manager/server/config.go
Original file line number Diff line number Diff line change
Expand Up @@ -108,12 +108,6 @@ func (rmc *ControllerConfig) Adjust(meta *configutil.ConfigMetaData) {
if rmc == nil {
return
}
<<<<<<< HEAD:pkg/mcs/resource_manager/server/config.go
rmc.RequestUnit.Adjust()

configutil.AdjustDuration(&rmc.DegradedModeWaitDuration, defaultDegradedModeWaitDuration)
configutil.AdjustDuration(&rmc.LTBMaxWaitDuration, defaultMaxWaitDuration)
=======
rmc.RequestUnit.Adjust(meta.Child("request-unit"))
if !meta.IsDefined("degraded-mode-wait-duration") {
configutil.AdjustDuration(&rmc.DegradedModeWaitDuration, defaultDegradedModeWaitDuration)
Expand All @@ -124,7 +118,6 @@ func (rmc *ControllerConfig) Adjust(meta *configutil.ConfigMetaData) {
if !meta.IsDefined("ltb-token-rpc-max-delay") {
configutil.AdjustDuration(&rmc.LTBTokenRPCMaxDelay, defaultLTBTokenRPCMaxDelay)
}
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352)):pkg/mcs/resourcemanager/server/config.go
failpoint.Inject("enableDegradedMode", func() {
configutil.AdjustDuration(&rmc.DegradedModeWaitDuration, time.Second)
})
Expand All @@ -149,7 +142,7 @@ type RequestUnitConfig struct {
}

// Adjust adjusts the configuration and initializes it with the default value if necessary.
func (ruc *RequestUnitConfig) Adjust() {
func (ruc *RequestUnitConfig) Adjust(_ *configutil.ConfigMetaData) {
if ruc == nil {
return
}
Expand Down
5 changes: 0 additions & 5 deletions pkg/mcs/resource_manager/server/config_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -43,14 +43,9 @@ read-cpu-ms-cost = 5.0
err = cfg.Adjust(&meta, false)
re.NoError(err)

<<<<<<< HEAD:pkg/mcs/resource_manager/server/config_test.go
re.Equal(cfg.Controller.DegradedModeWaitDuration.Duration, time.Second*2)
re.Equal(cfg.Controller.LTBMaxWaitDuration.Duration, time.Second*60)
=======
re.Equal(2*time.Second, cfg.Controller.DegradedModeWaitDuration.Duration)
re.Equal(60*time.Second, cfg.Controller.LTBMaxWaitDuration.Duration)
re.Equal(500*time.Millisecond, cfg.Controller.LTBTokenRPCMaxDelay.Duration)
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352)):pkg/mcs/resourcemanager/server/config_test.go
re.LessOrEqual(math.Abs(cfg.Controller.RequestUnit.CPUMsCost-5), 1e-7)
re.LessOrEqual(math.Abs(cfg.Controller.RequestUnit.WriteCostPerByte-4), 1e-7)
re.LessOrEqual(math.Abs(cfg.Controller.RequestUnit.WriteBaseCost-3), 1e-7)
Expand Down
11 changes: 4 additions & 7 deletions tests/integrations/mcs/resource_manager/resource_manager_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,7 @@ import (
"github.com/tikv/pd/client/resource_group/controller"
"github.com/tikv/pd/pkg/mcs/resource_manager/server"
"github.com/tikv/pd/pkg/utils/testutil"
"github.com/tikv/pd/pkg/utils/typeutil"
"github.com/tikv/pd/tests"
"go.uber.org/goleak"

Expand Down Expand Up @@ -1238,10 +1239,6 @@ func (suite *resourceManagerClientTestSuite) TestResourceGroupControllerConfigCh
tokenRPCMaxDelay := 2 * time.Second
readBaseCost := 1.5
defaultCfg := controller.DefaultConfig()
<<<<<<< HEAD:tests/integrations/mcs/resource_manager/resource_manager_test.go
// failpoint enableDegradedMode will setup and set it be 1s.
defaultCfg.DegradedModeWaitDuration.Duration = time.Second
=======
expectCfg := server.ControllerConfig{
// failpoint enableDegradedMode will setup and set it be 1s.
DegradedModeWaitDuration: typeutil.NewDuration(time.Second),
Expand All @@ -1250,13 +1247,13 @@ func (suite *resourceManagerClientTestSuite) TestResourceGroupControllerConfigCh
RequestUnit: server.RequestUnitConfig(defaultCfg.RequestUnit),
EnableControllerTraceLog: defaultCfg.EnableControllerTraceLog,
}
>>>>>>> 6b25787af (resource_control: allow configuration of the maximum retry time for the local bucket (#8352)):tests/integrations/mcs/resourcemanager/resource_manager_test.go
expectRUCfg := controller.GenerateRUConfig(defaultCfg)
expectRUCfg.DegradedModeWaitDuration = time.Second
// initial config verification
respString := sendRequest("GET", getAddr()+configURL, nil)
defaultString, err := json.Marshal(defaultCfg)
expectStr, err := json.Marshal(expectCfg)
re.NoError(err)
re.JSONEq(string(respString), string(defaultString))
re.JSONEq(string(respString), string(expectStr))
re.EqualValues(expectRUCfg, c1.GetConfig())

testCases := []struct {
Expand Down

0 comments on commit d072fdd

Please sign in to comment.