proxy-pool/internal/adapters/redisactivity/health.go
youfak 19cf3e32fa
Some checks are pending
ci / proto (push) Waiting to run
ci / test (ubuntu-latest) (push) Waiting to run
ci / test (windows-latest) (push) Waiting to run
ci / race (push) Waiting to run
ci / integration (push) Waiting to run
feat: persist target health profiles
2026-07-31 18:04:05 +08:00

264 lines
9.8 KiB
Go

package redisactivity
import (
"context"
"crypto/sha256"
"encoding/hex"
"errors"
"time"
"proxy-pool/internal/domain/activitypool"
healthDomain "proxy-pool/internal/domain/health"
)
var _ activitypool.HealthStore = (*Adapter)(nil)
var _ activitypool.GlobalHealthStore = (*Adapter)(nil)
var _ activitypool.TargetHealthStore = (*Adapter)(nil)
type targetHealthRecord struct {
Version int `json:"version"`
Status string `json:"status"`
ConsecutiveFailures int `json:"consecutiveFailures"`
LastHealthTaskID string `json:"lastHealthTaskId,omitempty"`
LastHealthDigest string `json:"lastHealthDigest,omitempty"`
LastHealthObservedAtMS int64 `json:"lastHealthObservedAtMs,omitempty"`
LastSuccessAtMS int64 `json:"lastSuccessAtMs,omitempty"`
LatencyNS int64 `json:"latencyNs"`
}
func (a *Adapter) ApplyHealth(ctx context.Context, update activitypool.HealthUpdate) (activitypool.Entry, error) {
if ctx == nil {
return activitypool.Entry{}, activitypool.ErrInvalidHealthUpdate
}
if err := ctx.Err(); err != nil {
return activitypool.Entry{}, err
}
if a == nil || update.ProxyID == "" || update.CheckedAt.IsZero() || update.Latency < 0 ||
!validProxyState(string(update.NextState)) {
return activitypool.Entry{}, activitypool.ErrInvalidHealthUpdate
}
operationID, err := newOperationID()
if err != nil {
return activitypool.Entry{}, err
}
result, err := runScript(ctx, a.client, healthScript, []string{
a.keys.records, a.keys.unique, a.keys.idkeys, a.keys.expiry, a.keys.available,
a.keys.inventory, a.keys.stateInventory, a.keys.owners, a.keys.ownerExpiry,
a.keys.operation(operationID),
}, update.CheckedAt.UnixMilli(), string(update.NextState), int64(update.Latency),
a.options.CleanupLimit, operationTTLMillis(a.options.OperationTTL), update.ProxyID)
if err != nil {
return activitypool.Entry{}, err
}
var reply healthScriptReply
if err := decodeScriptResult(result, &reply); err != nil {
return activitypool.Entry{}, err
}
switch reply.Status {
case scriptNotFound:
return activitypool.Entry{}, activitypool.ErrActivityNotFound
case scriptStale:
return activitypool.Entry{}, activitypool.ErrStaleHealthUpdate
case scriptInvalid:
return activitypool.Entry{}, activitypool.ErrInvalidHealthUpdate
case scriptOK:
if reply.Record == "" {
return activitypool.Entry{}, invalidScriptReply("health reply omitted record")
}
record, err := decodeProxyRecord(reply.Record)
if err != nil {
return activitypool.Entry{}, invalidScriptReply("health reply contained an invalid record")
}
return proxyRecordEntry(record), nil
default:
return activitypool.Entry{}, invalidScriptReply("unexpected health status")
}
}
// ApplyGlobalObservation reduces one BASIC or EGRESS observation inside the
// same Redis Lua transaction that updates state indexes and inventory.
func (a *Adapter) ApplyGlobalObservation(
ctx context.Context,
command activitypool.GlobalHealthCommand,
) (activitypool.Entry, error) {
if ctx == nil {
return activitypool.Entry{}, activitypool.ErrInvalidHealthUpdate
}
if err := ctx.Err(); err != nil {
return activitypool.Entry{}, err
}
normalized, err := healthDomain.NormalizeObservation(command.Observation)
if err != nil {
return activitypool.Entry{}, err
}
if a == nil || normalized.Level == healthDomain.LevelTarget || command.MaxConsecutiveFailures <= 0 {
return activitypool.Entry{}, activitypool.ErrInvalidHealthUpdate
}
digest, err := healthDomain.ObservationDigest(normalized)
if err != nil {
return activitypool.Entry{}, err
}
operationID, err := newOperationID()
if err != nil {
return activitypool.Entry{}, err
}
success := "0"
if normalized.Success {
success = "1"
}
result, err := runScript(ctx, a.client, healthScript, []string{
a.keys.records, a.keys.unique, a.keys.idkeys, a.keys.expiry, a.keys.available,
a.keys.inventory, a.keys.stateInventory, a.keys.owners, a.keys.ownerExpiry,
a.keys.operation(operationID),
}, normalized.ObservedAt.UnixMilli(), "", int64(normalized.Latency),
a.options.CleanupLimit, operationTTLMillis(a.options.OperationTTL), normalized.ProxyID,
"global", success, command.MaxConsecutiveFailures, normalized.TaskID, hex.EncodeToString(digest[:]))
if err != nil {
return activitypool.Entry{}, err
}
var reply healthScriptReply
if err := decodeScriptResult(result, &reply); err != nil {
return activitypool.Entry{}, err
}
switch reply.Status {
case scriptNotFound:
return activitypool.Entry{}, activitypool.ErrActivityNotFound
case scriptStale:
return activitypool.Entry{}, healthDomain.ErrStaleObservation
case scriptConflict:
return activitypool.Entry{}, healthDomain.ErrConflictingObservation
case scriptInvalid:
return activitypool.Entry{}, activitypool.ErrInvalidHealthUpdate
case scriptOK:
if reply.Record == "" {
return activitypool.Entry{}, invalidScriptReply("global health reply omitted record")
}
record, err := decodeProxyRecord(reply.Record)
if err != nil {
return activitypool.Entry{}, invalidScriptReply("global health reply contained an invalid record")
}
return proxyRecordEntry(record), nil
default:
return activitypool.Entry{}, invalidScriptReply("unexpected global health status")
}
}
// ApplyTargetObservation reduces a TARGET check into a short-lived Profile
// record keyed by (proxy, routing, target URL). It intentionally does not
// update the proxy record, global health, or any proxy selection index.
func (a *Adapter) ApplyTargetObservation(
ctx context.Context,
command activitypool.TargetHealthCommand,
) (healthDomain.TargetState, error) {
if ctx == nil {
return healthDomain.TargetState{}, activitypool.ErrInvalidHealthUpdate
}
if err := ctx.Err(); err != nil {
return healthDomain.TargetState{}, err
}
normalized, err := healthDomain.NormalizeObservation(command.Observation)
if err != nil {
return healthDomain.TargetState{}, err
}
if a == nil || normalized.Level != healthDomain.LevelTarget || command.MaxConsecutiveFailures <= 0 {
return healthDomain.TargetState{}, activitypool.ErrInvalidHealthUpdate
}
profile, err := healthDomain.NormalizeTargetProfile(healthDomain.TargetProfile{
RoutingName: normalized.RoutingName,
TargetURL: normalized.TargetURL,
})
if err != nil {
return healthDomain.TargetState{}, err
}
digest, err := healthDomain.ObservationDigest(normalized)
if err != nil {
return healthDomain.TargetState{}, err
}
operationID, err := newOperationID()
if err != nil {
return healthDomain.TargetState{}, err
}
success := "0"
if normalized.Success {
success = "1"
}
result, err := runScript(ctx, a.client, targetHealthScript, []string{
a.keys.records, a.keys.targetHealth(normalized.ProxyID), a.keys.operation(operationID),
}, normalized.ObservedAt.UnixMilli(), int64(normalized.Latency), normalized.ProxyID,
digestToken(profile.Key()), success, command.MaxConsecutiveFailures, normalized.TaskID,
hex.EncodeToString(digest[:]), operationTTLMillis(a.options.OperationTTL))
if err != nil {
return healthDomain.TargetState{}, err
}
var reply targetHealthScriptReply
if err := decodeScriptResult(result, &reply); err != nil {
return healthDomain.TargetState{}, err
}
switch reply.Status {
case scriptNotFound:
return healthDomain.TargetState{}, activitypool.ErrActivityNotFound
case scriptStale:
return healthDomain.TargetState{}, healthDomain.ErrStaleObservation
case scriptConflict:
return healthDomain.TargetState{}, healthDomain.ErrConflictingObservation
case scriptInvalid:
return healthDomain.TargetState{}, activitypool.ErrInvalidHealthUpdate
case scriptOK:
if reply.Target == "" {
return healthDomain.TargetState{}, invalidScriptReply("target health reply omitted target state")
}
state, err := decodeTargetHealthState(reply.Target)
if err != nil {
return healthDomain.TargetState{}, invalidScriptReply("target health reply contained an invalid target state")
}
return state, nil
default:
return healthDomain.TargetState{}, invalidScriptReply("unexpected target health status")
}
}
func decodeTargetHealthState(payload string) (healthDomain.TargetState, error) {
var record targetHealthRecord
if err := decodeJSON(payload, &record); err != nil {
return healthDomain.TargetState{}, err
}
if record.Version != recordVersion || record.ConsecutiveFailures < 0 || record.LatencyNS < 0 {
return healthDomain.TargetState{}, ErrInvalidRecord
}
state := healthDomain.TargetState{
Status: healthDomain.TargetStatus(record.Status),
ConsecutiveFailures: record.ConsecutiveFailures,
LastTaskID: record.LastHealthTaskID,
Latency: time.Duration(record.LatencyNS),
}
switch state.Status {
case healthDomain.TargetUnknown, healthDomain.TargetAvailable, healthDomain.TargetSuspect, healthDomain.TargetUnhealthy:
default:
return healthDomain.TargetState{}, ErrInvalidRecord
}
if record.LastHealthObservedAtMS == 0 {
if state.Status != healthDomain.TargetUnknown || state.LastTaskID != "" || state.ConsecutiveFailures != 0 ||
record.LastHealthDigest != "" || record.LastSuccessAtMS != 0 || state.Latency != 0 {
return healthDomain.TargetState{}, ErrInvalidRecord
}
return state, nil
}
if record.LastHealthObservedAtMS < 0 || record.LastSuccessAtMS < 0 || state.LastTaskID == "" ||
len(record.LastHealthDigest) != sha256.Size*2 {
return healthDomain.TargetState{}, ErrInvalidRecord
}
digest, err := hex.DecodeString(record.LastHealthDigest)
if err != nil || len(digest) != sha256.Size {
return healthDomain.TargetState{}, errors.Join(ErrInvalidRecord, err)
}
copy(state.LastObservationDigest[:], digest)
state.LastObservedAt = time.UnixMilli(record.LastHealthObservedAtMS).UTC()
if record.LastSuccessAtMS > 0 {
state.LastSuccessAt = time.UnixMilli(record.LastSuccessAtMS).UTC()
if state.LastSuccessAt.After(state.LastObservedAt) {
return healthDomain.TargetState{}, ErrInvalidRecord
}
}
return state, nil
}