Files
oc-discovery/docs/diagrams/03_indexer_heartbeat.puml
2026-03-11 16:28:15 +01:00

60 lines
4.2 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

@startuml indexer_heartbeat
title Heartbeat bidirectionnel node → indexeur (scoring 7 dimensions + challenges)
participant "Node A" as NodeA
participant "Node B" as NodeB
participant "IndexerService" as Indexer
note over NodeA,NodeB: SendHeartbeat goroutine — tick every 20s
== Tick Node A ==
NodeA -> Indexer: NewStream /opencloud/heartbeat/1.0\n(long-lived, réutilisé aux ticks suivants)
NodeA -> Indexer: stream.Encode(Heartbeat{\n name, PeerID_A, timestamp,\n indexersBinded: [addr1, addr2],\n need: maxPool - len(pool),\n challenges: [PeerID_A, PeerID_B], ← batch (tous les 1-10 HBs)\n challengeDID: "uuid-did-A", ← DHT challenge (tous les 5 batches)\n record: SignedPeerRecord_A ← expiry=now+2min\n})
Indexer -> Indexer: CheckHeartbeat(stream, maxNodes)\n→ len(Peers()) >= maxNodes → reject
Indexer -> Indexer: HandleHeartbeat → UptimeTracker.RecordHeartbeat()\n→ gap ≤ 2×interval : TotalOnline += gap
Indexer -> Indexer: Republish PeerRecord A to DHT\nDHT.PutValue("/node/"+DID_A, record_A)
== Réponse indexeur → node A ==
Indexer -> Indexer: BuildHeartbeatResponse(remotePeer=A, need, challenges, challengeDID)\n\nfillRate = connected_nodes / MaxNodesConn()\npeerCount = connected_nodes\nmaxNodes = MaxNodesConn()\nbornAt = time of indexer startup\n\nChallenges: pour chaque PeerID challengé\n found = PeerID dans StreamRecords[ProtocolHeartbeat]?\n lastSeen = HeartbeatStream.UptimeTracker.LastSeen\n\nDHT challenge:\n DHT.GetValue("/node/"+challengeDID, timeout=3s)\n → dhtFound + dhtPayload\n\nWitnesses: jusqu'à 3 AddrInfos de nœuds connectés\n (adresses connues dans Peerstore)\n\nSuggestions: jusqu'à `need` indexeurs depuis dhtCache\n (refresh asynchrone 2min, SelectByFillRate)\n\nSuggestMigrate: fillRate > 80%\n ET node dans offload.inBatch (batch ≤ 5, grace 3×HB)
Indexer --> NodeA: stream.Encode(HeartbeatResponse{\n fillRate, peerCount, maxNodes, bornAt,\n challenges, dhtFound, dhtPayload,\n witnesses, suggestions, suggestMigrate\n})
== Traitement score côté Node A ==
NodeA -> NodeA: score = ensureScore(Indexers, addr_indexer)\nscore.UptimeTracker.RecordHeartbeat()\n\nlatencyScore = max(0, 1 - RTT / (BaseRoundTrip × 10))\n\nBornAt stability:\n bornAt changed? → score.bornAtChanges++\n\nfillConsistency:\n expected = peerCount / maxNodes\n |expected - fillRate| < 10% → fillConsistent++\n\nChallenge PeerID (ground truth own PeerID):\n found=true AND lastSeen < 2×interval → challengeCorrect++\n\nDHT challenge:\n dhtFound=true → dhtSuccess++\n\nWitness query (async):\n go queryWitnesses(h, indexerID, bornAt, fillRate, witnesses, score)
NodeA -> NodeA: score.Score = ComputeNodeSideScore(latencyScore)\n\nScore = (\n 0.20 × uptimeRatio\n+ 0.20 × challengeAccuracy\n+ 0.15 × latencyScore\n+ 0.10 × fillScore ← 1 - fillRate\n+ 0.10 × fillConsistency\n+ 0.15 × witnessConsistency\n+ 0.10 × dhtSuccessRate\n) × 100 × bornAtPenalty\n\nbornAtPenalty = max(0, 1 - 0.30 × bornAtChanges)\nminScore = clamp(20 + 60 × (age.Hours/24), 20, 80)
alt score < minScore\n AND TotalOnline ≥ 2×interval\n AND !IsSeed\n AND len(pool) > 1
NodeA -> NodeA: evictPeer(dir, addr, id, proto)\n→ delete Addr + Score + Stream\ngo TriggerConsensus(h, voters, need)\n ou replenishIndexersFromDHT(h, need)
end
alt resp.SuggestMigrate == true AND nonSeedCount >= MinIndexer
alt IsSeed
NodeA -> NodeA: score.IsSeed = false\n(de-stickied — score eviction maintenant possible)
else !IsSeed
NodeA -> NodeA: evictPeer → migration acceptée
end
end
alt len(resp.Suggestions) > 0
NodeA -> NodeA: handleSuggestions(dir, indexerID, suggestions)\n→ inconnus ajoutés à Indexers Directory\n→ NudgeIt() si ajout effectif
end
== Tick Node B (concurrent) ==
NodeB -> Indexer: stream.Encode(Heartbeat{PeerID_B, ...})
Indexer -> Indexer: CheckHeartbeat → UptimeTracker → BuildHeartbeatResponse
Indexer --> NodeB: HeartbeatResponse{...}
== GC côté Indexeur ==
note over Indexer: GC ticker 30s — gc()\nnow.After(Expiry) où Expiry = lastHBTime + 2min\n→ AfterDelete(pid, name, did) hors lock\n→ publishNameEvent(NameIndexDelete, ...)\nFillRate recalculé automatiquement
@enduml