关于dubbo:聊聊dubbogo的DefaultHealthChecker

序

本文次要钻研一下 dubbo-go 的 DefaultHealthChecker

DefaultHealthChecker

dubbo-go-v1.4.2/cluster/router/healthcheck/default_health_check.go

func init() {extension.SethealthChecker(constant.DEFAULT_HEALTH_CHECKER, NewDefaultHealthChecker)
}

// DefaultHealthChecker is the default implementation of HealthChecker, which determines the health status of
// the invoker based on the number of successive bad request and the current active request.
type DefaultHealthChecker struct {
    // the limit of outstanding request
    outStandingRequestConutLimit int32
    // the threshold of successive-failure-request
    requestSuccessiveFailureThreshold int32
    // value of circuit-tripped timeout factor
    circuitTrippedTimeoutFactor int32
}

• DefaultHealthChecker 定义了 outStandingRequestConutLimit、requestSuccessiveFailureThreshold、circuitTrippedTimeoutFactor 属性

NewDefaultHealthChecker

dubbo-go-v1.4.2/cluster/router/healthcheck/default_health_check.go

// NewDefaultHealthChecker constructs a new DefaultHealthChecker based on the url
func NewDefaultHealthChecker(url *common.URL) router.HealthChecker {
    return &DefaultHealthChecker{outStandingRequestConutLimit:      int32(url.GetParamInt(constant.OUTSTANDING_REQUEST_COUNT_LIMIT_KEY, math.MaxInt32)),
        requestSuccessiveFailureThreshold: int32(url.GetParamInt(constant.SUCCESSIVE_FAILED_REQUEST_THRESHOLD_KEY, constant.DEFAULT_SUCCESSIVE_FAILED_REQUEST_MAX_DIFF)),
        circuitTrippedTimeoutFactor:       int32(url.GetParamInt(constant.CIRCUIT_TRIPPED_TIMEOUT_FACTOR_KEY, constant.DEFAULT_CIRCUIT_TRIPPED_TIMEOUT_FACTOR)),
    }
}

• NewDefaultHealthChecker 实例化 DefaultHealthChecker

IsHealthy

dubbo-go-v1.4.2/cluster/router/healthcheck/default_health_check.go

// IsHealthy evaluates the healthy state on the given Invoker based on the number of successive bad request
// and the current active request
func (c *DefaultHealthChecker) IsHealthy(invoker protocol.Invoker) bool {urlStatus := protocol.GetURLStatus(invoker.GetUrl())
    if c.isCircuitBreakerTripped(urlStatus) || urlStatus.GetActive() > c.GetOutStandingRequestConutLimit() {logger.Debugf("Invoker [%s] is currently in circuitbreaker tripped state", invoker.GetUrl().Key())
        return false
    }
    return true
}

• IsHealthy 办法通过 protocol.GetURLStatus(invoker.GetUrl())获取 urlStatus，之后依据 c.isCircuitBreakerTripped(urlStatus)或者 urlStatus.GetActive() > c.GetOutStandingRequestConutLimit() 判断是否处于 circuitbreaker tripped 状态

isCircuitBreakerTripped

dubbo-go-v1.4.2/cluster/router/healthcheck/default_health_check.go

// isCircuitBreakerTripped determine whether the invoker is in the tripped state by the number of successive bad request
func (c *DefaultHealthChecker) isCircuitBreakerTripped(status *protocol.RPCStatus) bool {circuitBreakerTimeout := c.getCircuitBreakerTimeout(status)
    currentTime := protocol.CurrentTimeMillis()
    if circuitBreakerTimeout <= 0 {return false}
    return circuitBreakerTimeout > currentTime
}

• isCircuitBreakerTripped 办法通过 c.getCircuitBreakerTimeout(status)获取 circuitBreakerTimeout，若 circuitBreakerTimeout 小于等于 0 返回 false，否则判断 circuitBreakerTimeout 是否大于 currentTime

getCircuitBreakerTimeout

dubbo-go-v1.4.2/cluster/router/healthcheck/default_health_check.go

// getCircuitBreakerTimeout get the timestamp recovered from tripped state, the unit is millisecond
func (c *DefaultHealthChecker) getCircuitBreakerTimeout(status *protocol.RPCStatus) int64 {sleepWindow := c.getCircuitBreakerSleepWindowTime(status)
    if sleepWindow <= 0 {return 0}
    return status.GetLastRequestFailedTimestamp() + sleepWindow}

• getCircuitBreakerTimeout 办法先获取 sleepWindow，若 sleepWindow 小于等于 0 则返回 0，否则返回 status.GetLastRequestFailedTimestamp() + sleepWindow

getCircuitBreakerSleepWindowTime

dubbo-go-v1.4.2/cluster/router/healthcheck/default_health_check.go

// getCircuitBreakerSleepWindowTime get the sleep window time of invoker, the unit is millisecond
func (c *DefaultHealthChecker) getCircuitBreakerSleepWindowTime(status *protocol.RPCStatus) int64 {successiveFailureCount := status.GetSuccessiveRequestFailureCount()
    diff := successiveFailureCount - c.GetRequestSuccessiveFailureThreshold()
    if diff < 0 {return 0} else if diff > constant.DEFAULT_SUCCESSIVE_FAILED_REQUEST_MAX_DIFF {diff = constant.DEFAULT_SUCCESSIVE_FAILED_REQUEST_MAX_DIFF}
    sleepWindow := (1 << diff) * c.GetCircuitTrippedTimeoutFactor()
    if sleepWindow > constant.MAX_CIRCUIT_TRIPPED_TIMEOUT_IN_MS {sleepWindow = constant.MAX_CIRCUIT_TRIPPED_TIMEOUT_IN_MS}
    return int64(sleepWindow)
}

• getCircuitBreakerSleepWindowTime 办法通过比照 status.GetSuccessiveRequestFailureCount()及 c.GetRequestSuccessiveFailureThreshold()计算 diff，进而依据 c.GetCircuitTrippedTimeoutFactor()计算 sleepWindow

小结

DefaultHealthChecker 定义了 outStandingRequestConutLimit、requestSuccessiveFailureThreshold、circuitTrippedTimeoutFactor 属性；IsHealthy 办法通过 protocol.GetURLStatus(invoker.GetUrl())获取 urlStatus，之后依据 c.isCircuitBreakerTripped(urlStatus)或者 urlStatus.GetActive() > c.GetOutStandingRequestConutLimit() 判断是否处于 circuitbreaker tripped 状态

doc

• default_health_check