feat(httpapi): harden control plane and speed up prefix refresh

Очередь job получила lease и fencing, телеметрия и прогресс deploy сохраняются, ревизия пишет manifest входов. Обновление префиксов не переписывает неизменённый снимок, большие наборы идут через COPY, а устаревшая ревизия не попадает в deploy. Бандл ноды кэшируется, prune и постановка job больше не держат горячий путь.

Co-authored-by: Cursor <cursoragent@cursor.com>

internal/store/telemetry.go

@@ -0,0 +1,113 @@
+package store
+
+import (
+	"context"
+	"crypto/sha256"
+	"encoding/hex"
+	"errors"
+	"strconv"
+	"strings"
+	"time"
+
+	"evobgp/internal/birdfmt"
+)
+
+// ErrTelemetryLeaseLost means a collector result belongs to an expired or replaced lease.
+var ErrTelemetryLeaseLost = errors.New("store: telemetry lease lost")
+
+// TelemetryLocalTargetID identifies the synthetic local BIRD target of a tenant.
+func TelemetryLocalTargetID(tenantID, authority string) string {
+	return "cp-local/" + tenantID + "/" + authority
+}
+
+// TelemetryObservation is the most recent attempt and last successful reading of one source.
+type TelemetryObservation struct {
+	ObservedAt    time.Time `json:"observed_at"`
+	ReceivedAt    time.Time `json:"received_at"`
+	LastSuccessAt time.Time `json:"last_success_at"`
+	Source        string    `json:"source"`
+	Error         string    `json:"error,omitempty"`
+}
+
+// TelemetryBGP is one BIRD protocols measurement.
+type TelemetryBGP struct {
+	TelemetryObservation
+	Sessions []birdfmt.BGPSession `json:"sessions,omitempty"`
+}
+
+// TelemetryHealth is one legacy agent health measurement.
+type TelemetryHealth struct {
+	TelemetryObservation
+	OK                    bool   `json:"ok"`
+	LastSyncAt            string `json:"last_sync_at,omitempty"`
+	LastAppliedRevisionID string `json:"last_applied_revision_id,omitempty"`
+}
+
+// TelemetrySnapshot contains independent BGP and agent health readings for one target.
+type TelemetrySnapshot struct {
+	TenantID, TargetID, SpeakerID, Kind string
+	LocalAuthority                      string
+	ConfigFingerprint                   string
+	BGP                                 TelemetryBGP
+	Health                              TelemetryHealth
+	NextPollAt                          time.Time
+	FailureCount                        int
+	DiscoveryCursor                     int
+}
+
+// TelemetrySpeakerFingerprint binds a snapshot to the configuration it observed.
+func TelemetrySpeakerFingerprint(sp *Speaker) string {
+	if sp == nil {
+		return ""
+	}
+	if strings.EqualFold(strings.TrimSpace(sp.Role), "master") {
+		sum := sha256.Sum256([]byte("master\x00" + sp.Endpoint))
+		return hex.EncodeToString(sum[:])
+	}
+	meta := ParseSpeakerMeta(sp.MetaJSON)
+	sum := sha256.Sum256([]byte(sp.Role + "\x00" + sp.Endpoint + "\x00" + meta.AgentDomain + "\x00" + meta.AgentSecret + "\x00" + strconv.Itoa(meta.AgentPort)))
+	return hex.EncodeToString(sum[:])
+}
+
+// TelemetryClaim fences a bounded collector attempt. Speaker is the current config.
+type TelemetryClaim struct {
+	TelemetrySnapshot
+	Speaker      *Speaker
+	OwnerID      string
+	FencingToken int64
+}
+
+// TelemetryResult is the atomic pair of results of one collector attempt.
+type TelemetryResult struct {
+	BGP    TelemetryBGP
+	Health TelemetryHealth
+}
+
+// TelemetryDiscoveryBatch rotates over long protocol lists without starving later entries.
+func TelemetryDiscoveryBatch(sessions []birdfmt.BGPSession, cursor, limit int) ([]birdfmt.BGPSession, int) {
+	if len(sessions) == 0 || limit <= 0 {
+		return nil, 0
+	}
+	if cursor < 0 || cursor >= len(sessions) {
+		cursor = 0
+	}
+	out := make([]birdfmt.BGPSession, 0, limit)
+	for visited := 0; visited < len(sessions); visited++ {
+		idx := (cursor + visited) % len(sessions)
+		if birdfmt.IsDynamicDiscoverySession(sessions[idx].Name) {
+			out = append(out, sessions[idx])
+			if len(out) == limit {
+				return out, (idx + 1) % len(sessions)
+			}
+		}
+	}
+	return out, 0
+}
+
+// TelemetryStore coordinates polling and read-only snapshots across API replicas.
+type TelemetryStore interface {
+	ClaimTelemetry(context.Context, string, int, time.Duration, string) ([]TelemetryClaim, error)
+	CompleteTelemetry(context.Context, TelemetryClaim, TelemetryResult, time.Time) error
+	ListTelemetry(context.Context, string) ([]TelemetrySnapshot, error)
+	RequestTelemetryRefresh(context.Context, string, time.Duration) (bool, error)
+}