區塊鏈教程btcpool礦池源碼分析StratumServer模組解析

標籤：變更   api   receive   dmi   Nid   info   ble   mit   abr   

兄弟連區塊鏈教程btcpool礦池源碼分析StratumServer模組解析

核心機制總結

• 接收的job延遲超過60秒將丟棄
• 如果job中prevHash與本地job中prevHash不同，即為已產生新塊，job中isClean狀態將置為true
  ????* true即要求礦機立即切換job
• 三種情況下將向礦機下發新job：
  ???? 收到新高度的job
  ???? 過去一個job為新高度且為空白塊job，且最新job為非空塊job
  ????* 達到預定的時間間隔30秒
• 最近一次下發job的時間將寫入檔案（由file_last_notify_time指定）
• 本地job有效期間為300秒
• 每10秒拉取一次新使用者列表（由list_id_api_url指定），使用者寫入本地map中
• sserver最大可用SessionId數為16777214
• btcpool支援BtcAgent擴充協議和Stratum協議，使用magic_number（0x7F）區分
• 處理Stratum協議：
  ???? suggest_target與suggest_difficulty等價，用於設定初始挖礦難度，需在subscribe之前請求
  ???? 使用sessionID作為extraNonce1_以確保礦機任務不重複
  ???? authorize之前有15秒讀逾時，authorize之後有10分鐘讀逾時，10分鐘無提交將中斷連線
  ???? 初始難度為16384，或從suggest_difficulty指定，下次將一次調整到位保持10s提交share
  ????* 每個session會維護一個localJobs_隊列，隊列長度為10條
• share被拒絕的幾種情況：
  ???? JOB_NOT_FOUND，localJobs_隊列中該job已被擠出
  ???? DUPLICATESHARE，share已提交過，已提交的share會計入submitShares
  ???? JOB_NOT_FOUND，jobRepository_中job不存在，即job已到期（300秒到期時間）
  ???? JOB_NOT_FOUND，jobRepository_中job狀態為Stale，即job是舊的非新job
  ???? TIME_TOO_OLD，share中提交的nTime小於job規定的minTime
  ???? TIME_TOO_OLD，share中提交的nTime比目前時間大10分鐘
  ????* LOW_DIFFICULTY，share中提交的hash不滿足難度目標
• 處理BtcAgent擴充協議：
  ???? Agent下礦機預設難度也為16384
  ???? 使用Agent sessionID作為extraNonce2_前半部分，以確保Agent下礦機任務不重複
  ???? 礦池下發新任務時，如session為BtcAgent，將為Agent下所有礦機計算難度
  ???????? 如難度發生變更，將按難度不同，分別構造多條CMD_MINING_SET_DIFF指令一併下發處理

StratumServer命令使用

sserver -c sserver.cfg -l log_dir#-c指定sserver設定檔#-l指定日誌目錄

sserver.cfg設定檔

//是否使用testnettestnet = true;//kafka叢集kafka = {brokers = "1.1.1.1:9092,2.2.2.2:9092,3.3.3.3:9092";};//sserver配置sserver = {//IP和連接埠ip = "0.0.0.0";port = 3333;//server id，全域唯一，取值範圍[1, 255]id = 1;//最近一次挖礦通知時間寫入檔案，用於監控file_last_notify_time = "/work/xxx/sserver_lastnotifytime.txt";//如果啟用模擬器，所有share均被接受，用於測試enable_simulator = false;//如果啟用，所有share都將成塊並被提交，用於測試enable_submit_invalid_block = false;//兩次share提交的間隔時間share_avg_seconds = 10;};users = {//使用者列表apilist_id_api_url = "example.com/get_user_id_list";};

SOLVED_SHARE訊息

```c++
if (isSubmitInvalidBlock == true || bnBlockHash <= bnNetworkTarget) {
????//
????// build found block
????//
????FoundBlock foundBlock;
????foundBlock.jobId = share.jobId;
????foundBlock.workerId = share.workerHashId;
????foundBlock.userId = share.userId;
????foundBlock.height = sjob->height;
????memcpy(foundBlock.header80, (const uint8t *)&header, sizeof(CBlockHeader));
????snprintf(foundBlock.workerFullName, sizeof(foundBlock.workerFullName_),
????????"%s", workFullName.c_str());
????// send
????sendSolvedShare2Kafka(&foundBlock, coinbaseBin);

????// mark jobs as stale
????jobRepository_->markAllJobsAsStale();

????LOG(INFO) << ">>>> found a new block: " << blkHash.ToString()
????<< ", jobId: " << share.jobId << ", userId: " << share.userId
????<< ", by: " << workFullName << " <<<<";
}

## 計算挖礦難度```c++//建構函式//kMinDiff_為最小難度，static const uint64 kMinDiff_ = 64;//kMaxDiff_為最大難度，static const uint64 kMaxDiff_ = 4611686018427387904ull;//kDefaultDiff_為預設初始難度，static const uint64 kDefaultDiff_ = 16384;//kDiffWindow_為N個share的時間視窗，static const time_t kDiffWindow_ = 900;//kRecordSeconds_為1個share的時間，static const time_t kRecordSeconds_ = 10;//sharesNum_和shares_初始值均為90DiffController(const int32_t shareAvgSeconds) :????startTime_(0),????minDiff_(kMinDiff_), curDiff_(kDefaultDiff_), curHashRateLevel_(0),????sharesNum_(kDiffWindow_/kRecordSeconds_), /* every N seconds as a record */????shares_ (kDiffWindow_/kRecordSeconds_){????if (shareAvgSeconds >= 1 && shareAvgSeconds <= 60) {????????shareAvgSeconds_ = shareAvgSeconds;????} else {????????shareAvgSeconds_ = 8;????}}//代碼btcpool/src/StratumSession.h//計算挖礦難度//不低於最小難度64uint64 DiffController::calcCurDiff() {uint64 diff = _calcCurDiff();if (diff < minDiff_) {diff = minDiff_;}return diff;}uint64 DiffController::_calcCurDiff() {const time_t now = time(nullptr);const int64 k = now / kRecordSeconds_;const double sharesCount = (double)sharesNum_.sum(k);if (startTime_ == 0) { // first time, we set the start timestartTime_ = time(nullptr);}const double kRateHigh = 1.40;const double kRateLow = 0.40;//時間視窗（900秒）內預期的share數double expectedCount = round(kDiffWindow_ / (double)shareAvgSeconds_);//return now >= startTime_ + kDiffWindow_;if (isFullWindow(now)) { /* have a full window now */// big miner have big expected share count to make it looks more smooth.expectedCount *= minerCoefficient(now, k);}if (expectedCount > kDiffWindow_) {//最多1秒提交1個，預期share數最大為900expectedCount = kDiffWindow_; // one second per share is enough}// this is for very low hashrate miner, eg. USB miners// should received at least one share every 60 seconds//非完整時間視窗、且時間已超過60s、提交的share數小於（60秒1個）、當前難度大於或等於2倍最小難度//此時降低難度為之前1/2if (!isFullWindow(now) && now >= startTime_ + 60 &&sharesCount <= (int32_t)((now - startTime_)/60.0) &&curDiff_ >= minDiff_*2) {setCurDiff(curDiff_ / 2);sharesNum_.mapMultiply(2.0);return curDiff_;}// too fast//如果提交share數超過預期數的1.4倍時if (sharesCount > expectedCount * kRateHigh) {//如果share數大於預期share數，且當前難度<最大難度時，提升難度為原來2倍while (sharesNum_.sum(k) > expectedCount &&curDiff_ < kMaxDiff_) {setCurDiff(curDiff_ * 2);sharesNum_.mapDivide(2.0); //share數/2}return curDiff_;}// too slow//如果是完整時間視窗，且當前難度大於或等於2被最小難度if (isFullWindow(now) && curDiff_ >= minDiff_*2) {//如果share數低於預期數的0.4，且當前難度大於或等於2倍最小難度，降低難度為原來的1/2while (sharesNum_.sum(k) < expectedCount * kRateLow &&curDiff_ >= minDiff_*2) {setCurDiff(curDiff_ / 2);sharesNum_.mapMultiply(2.0); //share數乘2}assert(curDiff_ >= minDiff_);return curDiff_;}return curDiff_;}

sserver校正share的機制

//本地job列表localJobs_最多保留最近10條任務，如有新任務，將擠出1條老任務。如果share所對應的job未在本地列表中，將StratumError::JOB_NOT_FOUND//本地share列表submitShares_中，如果已有本條share，即重複，將StratumError::DUPLICATE_SHARE//校正share不通過//　　job列表exJobs_沒有找到job，exJobs_中job有300秒到期時間，到期將刪除，報StratumError::JOB_NOT_FOUND//　　share中nTime小於job的最小時間，過老，報StratumError::TIME_TOO_OLD//　　share中nTime超過job中的nTime 10分鐘，過新，報StratumError::TIME_TOO_NEW//　　區塊雜湊>job難度目標，不合格，報StratumError::LOW_DIFFICULTY

sserver下發新job的機制

1、如果收到新高度statum job，將立即下發新job

```c++
bool isClean = false;
if (latestPrevBlockHash != sjob->prevHash) {
????isClean = true;
????latestPrevBlockHash = sjob->prevHash;
????LOG(INFO) << "received new height statum job, height: " << sjob->height
????<< ", prevhash: " << sjob->prevHash.ToString();
}
shared_ptr<StratumJobEx> exJob = std::makeshared<StratumJobEx>(sjob, isClean);
{
????ScopeLock sl(lock);
if (isClean) {
????????// mark all jobs as stale, should do this before insert new job
????????for (auto it : exJobs) {
????????????it.second->markStale();
????????}
????}
????// insert new job
????exJobs[sjob->jobId_] = exJob;
}
if (isClean) {
????sendMiningNotify(exJob);
????return;
}

2、如果過去一個job為新高度且為空白塊job，並且最新job非空塊job，將儘快下發新job

if (isClean == false && exJobs.size() >= 2) {
????auto itr = exJobs.rbegin();
????shared_ptr<StratumJobEx> exJob1 = itr->second;
????itr++;
????shared_ptr<StratumJobEx> exJob2 = itr->second;

????if (exJob2->isClean == true &&
????????exJob2->sjob->merkleBranch.size() == 0 &&
????????exJob1->sjob->merkleBranch_.size() != 0) {
????????sendMiningNotify(exJob);
????}
}

3、每超過一定時間間隔（30秒），將下發新job

void JobRepository::checkAndSendMiningNotify() {
????// last job is ‘expried‘, send a new one
????if (exJobs.size() &&
????????lastJobSendTime + kMiningNotifyInterval_ <= time(nullptr))
????{
????????sharedptr<StratumJobEx> exJob = exJobs.rbegin()->second;
????????sendMiningNotify(exJob);
????}
}

JobRepository::JobRepository(const char kafkaBrokers,
????const string &fileLastNotifyTime,
????Server server):
running(true),
kafkaConsumer(kafkaBrokers, KAFKA_TOPIC_STRATUMJOB, 0/patition/),
server(server), fileLastNotifyTime(fileLastNotifyTime),
kMaxJobsLifeTime(300),
kMiningNotifyInterval(30), // TODO: make as config arg
lastJobSendTime(0)
{
????assert(kMiningNotifyInterval < kMaxJobsLifeTime);
}

區塊鏈教程btcpool礦池源碼分析StratumServer模組解析