foundationdb/fdbserver/commitproxy/CommitProxyServer.cpp

/*
 * CommitProxyServer.cpp
 *
 * This source file is part of the FoundationDB open source project
 *
 * Copyright 2013-2026 Apple Inc. and the FoundationDB project authors
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include <algorithm>
#include <string_view>
#include <tuple>
#include <variant>

#include "fdbclient/AccumulativeChecksum.h"
#include "fdbclient/Atomic.h"
#include "fdbclient/BackupAgent.h"
#include "fdbclient/BuildIdempotencyIdMutations.h"
#include "fdbclient/CommitTransaction.h"
#include "fdbclient/DatabaseContext.h"
#include "fdbclient/FDBTypes.h"
#include "fdbclient/IdempotencyId.h"
#include "fdbclient/Knobs.h"
#include "fdbclient/CommitProxyInterface.h"
#include "fdbclient/NativeAPI.actor.h"
#include "fdbclient/SystemData.h"
#include "fdbclient/Tracing.h"
#include "fdbclient/TransactionLineage.h"
#include "fdbrpc/sim_validation.h"
#include "fdbserver/core/AccumulativeChecksumUtil.h"
#include "fdbserver/logsystem/ApplyMetadataMutation.h"
#include "fdbserver/core/ConflictBatch.h"
#include "fdbserver/core/DataDistributorInterface.h"
#include "fdbserver/kvstore/IKeyValueStore.h"
#include "fdbserver/core/Knobs.h"
#include "fdbserver/logsystem/LogSystem.h"
#include "fdbserver/kvstore/FDBExecHelper.h"
#include "fdbserver/logsystem/LogSystemFactory.h"
#include "fdbserver/logsystem/LogSystemDiskQueueAdapter.h"
#include "fdbserver/core/MasterInterface.h"
#include "fdbserver/core/MutationTracking.h"
#include "ProxyCommitData.h"
#include "fdbserver/core/RatekeeperInterface.h"
#include "fdbserver/core/RecoveryState.h"
#include "fdbserver/core/ServerDBInfo.h"
#include "fdbserver/core/WaitFailure.h"
#include "fdbserver/commitproxy/CommitProxyServer.h"
#include "fdbserver/core/WorkerInterface.h"
#include "flow/ActorCollection.h"
#include "flow/CodeProbe.h"
#include "flow/CoroUtils.h"
#include "flow/EncryptUtils.h"
#include "flow/Error.h"
#include "flow/IRandom.h"
#include "flow/Knobs.h"
#include "flow/Trace.h"
#include "flow/UnitTest.h"
#include "flow/network.h"

using WriteMutationRefVar = std::variant<MutationRef, VectorRef<MutationRef>>;

struct ResolutionRequestBuilder {
	const ProxyCommitData* self;

	// One request per resolver.
	std::vector<ResolveTransactionBatchRequest> requests;

	// Txn i to resolvers that have i'th data sent
	std::vector<std::vector<int>> transactionResolverMap;
	std::vector<CommitTransactionRef*> outTr;

	// Used to report conflicting keys, the format is
	// [CommitTransactionRef_Index][Resolver_Index][Read_Conflict_Range_Index_on_Resolver]
	// -> read_conflict_range's original index in the commitTransactionRef
	std::vector<std::vector<std::vector<int>>> txReadConflictRangeIndexMap;

	ResolutionRequestBuilder(ProxyCommitData* self,
	                         Version version,
	                         Version prevVersion,
	                         Version lastReceivedVersion,
	                         Version lastShardMove,
	                         Span& parentSpan)
	  : self(self), requests(self->resolvers.size()) {
		for (auto& req : requests) {
			req.spanContext = parentSpan.context;
			req.prevVersion = prevVersion;
			req.version = version;
			req.lastReceivedVersion = lastReceivedVersion;
			req.lastShardMove = lastShardMove;
		}
	}

	CommitTransactionRef& getOutTransaction(int resolver, Version read_snapshot) {
		CommitTransactionRef*& out = outTr[resolver];
		if (!out) {
			ResolveTransactionBatchRequest& request = requests[resolver];
			request.transactions.resize(request.arena, request.transactions.size() + 1);
			out = &request.transactions.back();
			out->read_snapshot = read_snapshot;
		}
		return *out;
	}

	std::vector<int> getResolversForRange(const KeyRangeRef& range, Optional<Version> readSnapshot) const {
		std::vector<int> resolvers;
		resolvers.reserve(self->resolvers.size());
		std::vector<unsigned char> seen(self->resolvers.size(), 0);
		for (auto& intersectingRange : self->keyResolvers.intersectingRanges(range)) {
			auto& versionResolvers = intersectingRange.value();
			if (readSnapshot.present()) {
				for (int i = versionResolvers.size() - 1; i >= 0; --i) {
					const int resolver = versionResolvers[i].second;
					if (!seen[resolver]) {
						seen[resolver] = 1;
						resolvers.push_back(resolver);
					}
					if (versionResolvers[i].first < readSnapshot.get()) {
						break;
					}
				}
			} else if (!versionResolvers.empty()) {
				const int resolver = versionResolvers.back().second;
				if (!seen[resolver]) {
					seen[resolver] = 1;
					resolvers.push_back(resolver);
				}
			}
		}

		if (SERVER_KNOBS->PROXY_USE_RESOLVER_PRIVATE_MUTATIONS && systemKeys.intersects(range)) {
			resolvers.clear();
			for (int resolver = 0; resolver < self->resolvers.size(); ++resolver) {
				resolvers.push_back(resolver);
			}
		}

		ASSERT(!resolvers.empty());
		return resolvers;
	}

	// Returns a read conflict index map: [resolver_index][read_conflict_range_index_on_the_resolver]
	// -> read_conflict_range's original index
	std::vector<std::vector<int>> addReadConflictRanges(CommitTransactionRef& trIn) {
		std::vector<std::vector<int>> rCRIndexMap(requests.size());
		for (int idx = 0; idx < trIn.read_conflict_ranges.size(); ++idx) {
			const auto& r = trIn.read_conflict_ranges[idx];
			for (int resolver : getResolversForRange(r, trIn.read_snapshot)) {
				getOutTransaction(resolver, trIn.read_snapshot)
				    .read_conflict_ranges.push_back(requests[resolver].arena, r);
				rCRIndexMap[resolver].push_back(idx);
			}
		}
		return rCRIndexMap;
	}

	void addWriteConflictRanges(CommitTransactionRef& trIn) {
		for (auto& r : trIn.write_conflict_ranges) {
			for (int resolver : getResolversForRange(r, Optional<Version>())) {
				getOutTransaction(resolver, trIn.read_snapshot)
				    .write_conflict_ranges.push_back(requests[resolver].arena, r);
			}
		}
	}

	void addTransaction(CommitTransactionRequest& trRequest, Version ver, int transactionNumberInBatch) {
		auto& trIn = trRequest.transaction;
		// SOMEDAY: There are a couple of unnecessary O( # resolvers ) steps here
		outTr.assign(requests.size(), nullptr);
		ASSERT(transactionNumberInBatch >= 0 && transactionNumberInBatch < 32768);

		bool isTXNStateTransaction = false;
		for (auto& m : trIn.mutations) {
			DEBUG_MUTATION("AddTr", ver, m, self->dbgid).detail("Idx", transactionNumberInBatch);
			if (m.type == MutationRef::SetVersionstampedKey) {
				transformVersionstampMutation(m, &MutationRef::param1, requests[0].version, transactionNumberInBatch);
				trIn.write_conflict_ranges.push_back(requests[0].arena, singleKeyRange(m.param1, requests[0].arena));
			} else if (m.type == MutationRef::SetVersionstampedValue) {
				transformVersionstampMutation(m, &MutationRef::param2, requests[0].version, transactionNumberInBatch);
			}
			if (isMetadataMutation(m)) {
				isTXNStateTransaction = true;
				auto& tr = getOutTransaction(0, trIn.read_snapshot);
				tr.mutations.push_back(requests[0].arena, m);
				tr.lock_aware = trRequest.isLockAware();
			}
		}
		if (isTXNStateTransaction && !trRequest.isLockAware()) {
			// This mitigates https://github.com/apple/foundationdb/issues/3647. Since this transaction is not lock
			// aware, if this transaction got a read version then \xff/dbLocked must not have been set at this
			// transaction's read snapshot. If that changes by commit time, then it won't commit on any proxy because of
			// a conflict. A client could set a read version manually so this isn't totally bulletproof.
			trIn.read_conflict_ranges.push_back(trRequest.arena, KeyRangeRef(databaseLockedKey, databaseLockedKeyEnd));
		}

		std::vector<std::vector<int>> rCRIndexMap = addReadConflictRanges(trIn);
		txReadConflictRangeIndexMap.push_back(std::move(rCRIndexMap));

		addWriteConflictRanges(trIn);

		if (isTXNStateTransaction) {
			for (int r = 0; r < requests.size(); r++) {
				int transactionNumberInRequest =
				    &getOutTransaction(r, trIn.read_snapshot) - requests[r].transactions.begin();
				requests[r].txnStateTransactions.push_back(requests[r].arena, transactionNumberInRequest);
			}
			// Note only Resolver 0 got the correct spanContext, which means
			// the reply from Resolver 0 has the right one back.
			auto& tr = getOutTransaction(0, trIn.read_snapshot);
			tr.spanContext = trRequest.spanContext;
		}

		std::vector<int> resolversUsed;
		for (int r = 0; r < outTr.size(); r++) {
			if (outTr[r]) {
				resolversUsed.push_back(r);
				outTr[r]->report_conflicting_keys = trIn.report_conflicting_keys;
			}
		}
		transactionResolverMap.emplace_back(std::move(resolversUsed));
	}
};

Future<Void> commitBatcher(ProxyCommitData* commitData,
                           PromiseStream<std::pair<std::vector<CommitTransactionRequest>, int>> out,
                           FutureStream<CommitTransactionRequest> in,
                           int desiredBytes,
                           int64_t memBytesLimit) {
	co_await delayJittered(commitData->commitBatchInterval, TaskPriority::ProxyCommitBatcher);

	double lastBatch = 0;

	while (true) {
		Future<Void> timeout;
		std::vector<CommitTransactionRequest> batch;
		int batchBytes = 0;
		auto flushBatch = [&](ProxyStats::CommitBatchFlushReason reason) {
			commitData->stats.recordCommitBatchFlush(reason);
			out.send({ std::move(batch), batchBytes });
			lastBatch = now();
		};
		// TODO: Enable this assertion (currently failing with gcc)
		// static_assert(std::is_nothrow_move_constructible_v<CommitTransactionRequest>);

		if (SERVER_KNOBS->MAX_COMMIT_BATCH_INTERVAL <= 0) {
			timeout = Never();
		} else {
			timeout = delayJittered(SERVER_KNOBS->MAX_COMMIT_BATCH_INTERVAL, TaskPriority::ProxyCommitBatcher);
		}

		while (!timeout.isReady() &&
		       !(batch.size() == SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_COUNT_MAX || batchBytes >= desiredBytes)) {
			auto res = co_await race(in, timeout, commitData->triggerCommit.onChange());
			if (res.index() == 0) {
				CommitTransactionRequest req = std::get<0>(std::move(res));

				// WARNING: this code is run at a high priority, so it needs to do as little work as possible
				int bytes = getBytes(req);

				// Drop requests if memory is under severe pressure
				if (commitData->commitBatchesMemBytesCount + bytes > memBytesLimit) {
					++commitData->stats.txnCommitErrors;
					req.reply.sendError(commit_proxy_memory_limit_exceeded());
					TraceEvent(SevWarnAlways, "ProxyCommitBatchMemoryThresholdExceeded")
					    .suppressFor(60)
					    .detail("MemBytesCount", commitData->commitBatchesMemBytesCount)
					    .detail("MemLimit", memBytesLimit);
					continue;
				}
				commitData->stats.transactionSizeDist->sample(bytes);

				if (bytes > FLOW_KNOBS->PACKET_WARNING) {
					TraceEvent(SevWarn, "LargeTransaction")
					    .suppressFor(1.0)
					    .detail("Size", bytes)
					    .detail("Client", req.reply.getEndpoint().getPrimaryAddress());
				}

				++commitData->stats.txnCommitIn;
				commitData->stats.uniqueClients.insert(req.reply.getEndpoint().getPrimaryAddress());

				if (req.debugID.present()) {
					g_traceBatch.addEvent("CommitDebug", req.debugID.get().first(), "CommitProxyServer.batcher");
				}

				if (batch.empty()) {
					if (now() - lastBatch > commitData->commitBatchInterval) {
						timeout = delayJittered(SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_INTERVAL_FROM_IDLE,
						                        TaskPriority::ProxyCommitBatcher);
					} else {
						timeout = delayJittered(commitData->commitBatchInterval - (now() - lastBatch),
						                        TaskPriority::ProxyCommitBatcher);
					}
				}

				if ((batchBytes + bytes > CLIENT_KNOBS->TRANSACTION_SIZE_LIMIT || req.firstInBatch()) &&
				    !batch.empty()) {
					auto reason = batchBytes + bytes > CLIENT_KNOBS->TRANSACTION_SIZE_LIMIT
					                  ? ProxyStats::CommitBatchFlushReason::TRANSACTION_SIZE_LIMIT
					                  : ProxyStats::CommitBatchFlushReason::FIRST_IN_BATCH;
					commitData->triggerCommit.set(false);
					flushBatch(reason);
					timeout = delayJittered(commitData->commitBatchInterval, TaskPriority::ProxyCommitBatcher);
					batch.clear();
					batchBytes = 0;
				}

				batch.push_back(req);
				batchBytes += bytes;
				commitData->commitBatchesMemBytesCount += bytes;
			} else if (res.index() == 1) {
			} else if (res.index() == 2) {
				ASSERT(commitData->triggerCommit.get());
				double commitTime = lastBatch + SERVER_KNOBS->COMMIT_TRIGGER_DELAY;
				if (now() > commitTime) {
					break;
				}

				timeout = timeout || delayJittered(commitTime - now(), TaskPriority::ProxyCommitBatcher);
			} else {
				UNREACHABLE();
			}
		}
		auto reason = batch.size() == SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_COUNT_MAX
		                  ? ProxyStats::CommitBatchFlushReason::COUNT_LIMIT
		              : batchBytes >= desiredBytes ? ProxyStats::CommitBatchFlushReason::BYTE_LIMIT
		                                           : ProxyStats::CommitBatchFlushReason::TIMEOUT;
		commitData->triggerCommit.set(false);
		flushBatch(reason);
	}
}

void createWhitelistBinPathVec(const std::string& binPath, std::vector<Standalone<StringRef>>& binPathVec) {
	TraceEvent(SevDebug, "BinPathConverter").detail("Input", binPath);
	StringRef input(binPath);
	while (!input.empty()) {
		StringRef token = input.eat(","_sr);
		if (!token.empty()) {
			const uint8_t* ptr = token.begin();
			while (ptr != token.end() && *ptr == ' ') {
				ptr++;
			}
			if (ptr != token.end()) {
				Standalone<StringRef> newElement(token.substr(ptr - token.begin()));
				TraceEvent(SevDebug, "BinPathItem").detail("Element", newElement);
				binPathVec.push_back(newElement);
			}
		}
	}
	return;
}

bool isWhitelisted(const std::vector<Standalone<StringRef>>& binPathVec, StringRef binPath) {
	TraceEvent("BinPath").detail("Value", binPath);
	for (const auto& item : binPathVec) {
		TraceEvent("Element").detail("Value", item);
	}
	return std::find(binPathVec.begin(), binPathVec.end(), binPath) != binPathVec.end();
}

Future<Void> addBackupMutations(ProxyCommitData* self,
                                const std::map<Key, MutationListRef>* logRangeMutations,
                                LogPushData* toCommit,
                                Version commitVersion,
                                double* computeDuration,
                                double* computeStart) {
	const int32_t version = commitVersion / CLIENT_KNOBS->LOG_RANGE_BLOCK_SIZE;
	int yieldBytes = 0;

	toCommit->addTransactionInfo(SpanContext());

	// Serialize the log range mutations within the map
	for (auto logRangeMutation = logRangeMutations->cbegin(); logRangeMutation != logRangeMutations->cend();
	     ++logRangeMutation) {
		// FIXME: this is re-implementing the serialize function of MutationListRef in order to have a yield
		// this is 0x0FDB00A200090001
		BinaryWriter valueWriter(IncludeVersion(ProtocolVersion::withBackupMutations()));
		valueWriter << logRangeMutation->second.totalSize(); // this is int32 by default

		MutationListRef::Blob* blobIter = logRangeMutation->second.blob_begin;
		while (blobIter) {
			if (yieldBytes > SERVER_KNOBS->DESIRED_TOTAL_BYTES) {
				yieldBytes = 0;
				if (g_network->check_yield(TaskPriority::ProxyCommitYield1)) {
					*computeDuration += g_network->timer_monotonic() - *computeStart;
					co_await delay(0, TaskPriority::ProxyCommitYield1);
					*computeStart = g_network->timer_monotonic();
				}
			}
			valueWriter.serializeBytes(blobIter->data);
			yieldBytes += blobIter->data.size();
			blobIter = blobIter->next;
		}

		Key val = valueWriter.toValue();

		BinaryWriter wr(Unversioned()); // backupName/hash/commitVersion/part, so wr is param1

		// Serialize the log destination
		wr.serializeBytes(logRangeMutation->first);

		// Write the log keys and version information
		wr << (uint8_t)hashlittle(&version, sizeof(version), 0);
		wr << bigEndian64(commitVersion);

		uint32_t* partBuffer = nullptr;

		for (int part = 0; part * CLIENT_KNOBS->MUTATION_BLOCK_SIZE < val.size(); part++) {

			MutationRef backupMutation;
			backupMutation.type = MutationRef::SetValue;
			// Assign the second parameter as the part
			// Define the mutation type and and location
			backupMutation.param2 = getBackupValue(val, part);
			Key key = getBackupKey(wr, &partBuffer, part); // holds the memory for backupMutation
			backupMutation.param1 = key;
			ASSERT(backupMutation.param1.startsWith(
			    logRangeMutation->first)); // We are writing into the configured destination

			auto& tags = self->tagsForKey(backupMutation.param1);
			toCommit->addTags(tags);

			if (self->acsBuilder != nullptr) {
				updateMutationWithAcsAndAddMutationToAcsBuilder(
				    self->acsBuilder,
				    backupMutation,
				    tags,
				    getCommitProxyAccumulativeChecksumIndex(self->commitProxyIndex),
				    self->epoch,
				    commitVersion,
				    self->dbgid);
			}

			toCommit->writeTypedMessage(backupMutation);

			//			if (DEBUG_MUTATION("BackupProxyCommit", commitVersion, backupMutation)) {
			//				TraceEvent("BackupProxyCommitTo", self->dbgid).detail("To",
			// describe(tags)).detail("BackupMutation", backupMutation.toString())
			// .detail("BackupMutationSize", val.size()).detail("Version", commitVersion).detail("DestPath",
			// logRangeMutation.first) 					.detail("PartIndex", part).detail("PartIndexEndian",
			// bigEndian32(part)).detail("PartData", backupMutation.param1);
			//			}
		}
	}
}

static Future<Void> releaseResolvingAfterImpl(ProxyCommitData* self,
                                              Future<Void> releaseDelay,
                                              int64_t localBatchNumber) {
	co_await releaseDelay;
	ASSERT(self->latestLocalCommitBatchResolving.get() == localBatchNumber - 1);
	self->latestLocalCommitBatchResolving.set(localBatchNumber);
	co_return;
}

Future<Void> releaseResolvingAfter(ProxyCommitData* self, Future<Void> releaseDelay, int64_t localBatchNumber) {
	if (releaseDelay.isReady()) {
		releaseDelay.get();
		ASSERT(self->latestLocalCommitBatchResolving.get() == localBatchNumber - 1);
		self->latestLocalCommitBatchResolving.set(localBatchNumber);
		return Void();
	}

	return releaseResolvingAfterImpl(self, releaseDelay, localBatchNumber);
}

static Future<ResolveTransactionBatchReply> trackResolutionMetrics(Reference<Histogram> dist,
                                                                   Future<ResolveTransactionBatchReply> in) {
	double startTime = g_network->timer_monotonic();
	ResolveTransactionBatchReply reply = co_await in;
	dist->sampleSeconds(g_network->timer_monotonic() - startTime);
	co_return reply;
}

namespace CommitBatch {

constexpr const std::string_view UNSET = std::string_view();
constexpr const std::string_view INITIALIZE = "initialize"sv;
constexpr const std::string_view PRE_RESOLUTION = "preResolution"sv;
constexpr const std::string_view RESOLUTION = "resolution"sv;
constexpr const std::string_view POST_RESOLUTION = "postResolution"sv;
constexpr const std::string_view TRANSACTION_LOGGING = "transactionLogging"sv;
constexpr const std::string_view REPLY = "reply"sv;
constexpr const std::string_view COMPLETE = "complete"sv;

struct CommitBatchContext {
	using StoreCommit_t = std::vector<std::pair<Future<LogSystemDiskQueueAdapter::CommitMessage>, Future<Void>>>;

	ProxyCommitData* const pProxyCommitData;
	std::vector<CommitTransactionRequest> trs;
	const int currentBatchMemBytesCount;

	double startTime;

	// The current stage of batch commit
	std::string_view stage = UNSET;

	Optional<UID> debugID;

	bool forceRecovery = false;
	bool rejected = false; // If rejected due to long queue length

	int64_t localBatchNumber;
	LogPushData toCommit;

	int batchOperations = 0;

	Span span;

	int64_t batchBytes = 0;

	int latencyBucket = 0;

	Version commitVersion;
	Version prevVersion;

	int64_t maxTransactionBytes;
	std::vector<std::vector<int>> transactionResolverMap;
	std::vector<std::vector<std::vector<int>>> txReadConflictRangeIndexMap;

	Future<Void> releaseDelay;
	Future<Void> releaseFuture;

	std::vector<ResolveTransactionBatchReply> resolution;

	double computeStart;
	double computeDuration = 0;

	Arena arena;

	/// true if the batch is the 1st batch for this proxy, additional metadata
	/// processing is involved for this batch.
	bool isMyFirstBatch;
	bool firstStateMutations;

	Optional<Value> previousCoordinators;

	StoreCommit_t storeCommits;

	std::vector<uint8_t> committed;

	Optional<Key> lockedKey;
	bool locked;

	int commitCount = 0;

	std::vector<int> nextTr;

	bool lockedAfter;

	Optional<Value> metadataVersionAfter;

	int mutationCount = 0;
	int mutationBytes = 0;

	std::map<Key, MutationListRef> logRangeMutations;
	Arena logRangeMutationsArena;

	int transactionNum = 0;
	int yieldBytes = 0;

	LogSystemDiskQueueAdapter::CommitMessage msg;

	Future<Version> loggingComplete;

	double commitStartTime;

	std::unordered_map<uint16_t, Version> tpcvMap; // obtained from resolver
	std::set<Tag> writtenTags; // final set tags written to in the batch
	std::set<Tag> writtenTagsPreResolution; // tags written to in the batch not including any changes from the resolver.

	IdempotencyIdKVBuilder idempotencyKVBuilder;

	CommitBatchContext(ProxyCommitData*, const std::vector<CommitTransactionRequest>*, const int);

	void setupTraceBatch();

	std::set<Tag> getWrittenTagsPreResolution();

	void checkHotShards();

	bool rangeLockEnabled();

	Version lastShardMove;

private:
	void evaluateBatchSize();
};

bool CommitBatchContext::rangeLockEnabled() {
	return pProxyCommitData->rangeLockEnabled();
}

void CommitBatchContext::checkHotShards() {
	// removed expired hot shards
	for (auto it = pProxyCommitData->hotShards.begin(); it != pProxyCommitData->hotShards.end();) {
		if (now() > it->second) {
			it = pProxyCommitData->hotShards.erase(it);
		} else {
			++it;
		}
	}

	if (pProxyCommitData->hotShards.empty()) {
		return;
	}

	auto trsBegin = trs.begin();

	std::vector<size_t> transactionsToRemove;
	for (int transactionNum = 0; transactionNum < trs.size(); transactionNum++) {
		VectorRef<MutationRef>* pMutations = &trs[transactionNum].transaction.mutations;
		bool abortTransaction = false;
		for (int mutationNum = 0; mutationNum < pMutations->size(); mutationNum++) {
			auto& m = (*pMutations)[mutationNum];
			if (isSingleKeyMutation((MutationRef::Type)m.type)) {
				for (const auto& shard : pProxyCommitData->hotShards) {
					if (shard.first.contains(KeyRef(m.param1))) {
						abortTransaction = true;
						break;
					}
				}
			} else if (m.type == MutationRef::ClearRange) {
				for (const auto& shard : pProxyCommitData->hotShards) {
					if (shard.first.intersects(KeyRangeRef(m.param1, m.param2))) {
						abortTransaction = true;
						break;
					}
				}
			} else {
				UNREACHABLE();
			}
		}
		if (abortTransaction) {
			trs[transactionNum].reply.sendError(transaction_throttled_hot_shard());
			transactionsToRemove.push_back(transactionNum);
		}
	}
	// Remove transactions marked for removal in reverse order to avoid shifting indices
	for (auto it = transactionsToRemove.rbegin(); it != transactionsToRemove.rend(); ++it) {
		trs.erase(trsBegin + *it);
	}
	committed.resize(trs.size());
	return;
}

// Check whether the mutation intersects any legal backup ranges
// If so, it will be clamped to the intersecting range(s) later
inline bool shouldBackup(MutationRef const& m) {
	if (normalKeys.contains(m.param1) || m.param1 == metadataVersionKey) {
		return true;
	} else if (m.type != MutationRef::Type::ClearRange) {
		return systemBackupMutationMask().rangeContaining(m.param1).value();
	} else {
		for (auto& r : systemBackupMutationMask().intersectingRanges(KeyRangeRef(m.param1, m.param2))) {
			if (r->value()) {
				return true;
			}
		}
	}
	return false;
}

// Find the set of logs the batch is sent to. An empty set indicates it cannot be
// determined. In version vector, this means the batch should be sent to all logs.
std::set<Tag> CommitBatchContext::getWrittenTagsPreResolution() {
	std::set<Tag> transactionTags;
	lastShardMove = pProxyCommitData->lastShardMove;
	if (pProxyCommitData->txnStateStore->getReplaceContent()) {
		return std::set<Tag>();
	}
	if (!pProxyCommitData->idempotencyClears.empty()) {
		return std::set<Tag>();
	}
	for (int transactionNum = 0; transactionNum < trs.size(); transactionNum++) {
		int mutationNum = 0;
		VectorRef<MutationRef>* pMutations = &trs[transactionNum].transaction.mutations;
		if (trs[transactionNum].idempotencyId.valid()) {
			return std::set<Tag>();
		}
		for (; mutationNum < pMutations->size(); mutationNum++) {
			auto& m = (*pMutations)[mutationNum];
			// disable version vector's effect if any mutation in the batch is backed up.
			// TODO: make backup work with version vector.
			if (pProxyCommitData->vecBackupKeys.size() > 1 && shouldBackup(m)) {
				return std::set<Tag>();
			}
			if (isSingleKeyMutation((MutationRef::Type)m.type)) {
				auto& tags = pProxyCommitData->tagsForKey(m.param1);
				transactionTags.insert(tags.begin(), tags.end());
				if (!pProxyCommitData->cdcRouting.empty()) {
					const auto& cdcTags = pProxyCommitData->cdcRouting.tagsForKey(m.param1);
					transactionTags.insert(cdcTags.begin(), cdcTags.end());
				}
			} else if (m.type == MutationRef::ClearRange) {
				auto range = pProxyCommitData->keyInfo.rangeContaining(m.param1);
				if (range.end() >= m.param2) {
					range.value().populateTags();
					transactionTags.insert(range.value().tags.begin(), range.value().tags.end());
				} else {
					std::set<Tag> allSources;
					while (range.begin() < m.param2) {
						range.value().populateTags();
						allSources.insert(range.value().tags.begin(), range.value().tags.end());
						transactionTags.insert(range.value().tags.begin(), range.value().tags.end());
						++range;
					}
				}
				KeyRangeRef clearRange(KeyRangeRef(m.param1, m.param2));
				if (!pProxyCommitData->cdcRouting.empty()) {
					const auto cdcTags = pProxyCommitData->cdcRouting.tagsForRange(clearRange);
					transactionTags.insert(cdcTags.begin(), cdcTags.end());
				}
			} else {
				UNREACHABLE();
			}
		}
	}

	if (toCommit.getLogRouterTags()) {
		toCommit.storeRandomRouterTag();
		transactionTags.insert(toCommit.savedRandomRouterTag.get());
	}

	return transactionTags;
}

CommitBatchContext::CommitBatchContext(ProxyCommitData* const pProxyCommitData_,
                                       const std::vector<CommitTransactionRequest>* trs_,
                                       const int currentBatchMemBytesCount)
  : pProxyCommitData(pProxyCommitData_), trs(std::move(*const_cast<std::vector<CommitTransactionRequest>*>(trs_))),
    currentBatchMemBytesCount(currentBatchMemBytesCount), startTime(g_network->now()),
    localBatchNumber(++pProxyCommitData->localCommitBatchesStarted),
    toCommit(pProxyCommitData->logSystem, pProxyCommitData->localTLogCount), span("MP:commitBatch"_loc),
    committed(trs.size()), lastShardMove(invalidVersion) {

	evaluateBatchSize();

	if (batchOperations != 0) {
		latencyBucket =
		    std::min<int>(SERVER_KNOBS->PROXY_COMPUTE_BUCKETS - 1,
		                  SERVER_KNOBS->PROXY_COMPUTE_BUCKETS * batchBytes /
		                      (batchOperations * (CLIENT_KNOBS->VALUE_SIZE_LIMIT + CLIENT_KNOBS->KEY_SIZE_LIMIT)));
	}

	// since we are using just the former to limit the number of versions actually in flight!
	ASSERT(SERVER_KNOBS->MAX_READ_TRANSACTION_LIFE_VERSIONS <= SERVER_KNOBS->MAX_VERSIONS_IN_FLIGHT);
}

void CommitBatchContext::setupTraceBatch() {
	for (const auto& tr : trs) {
		if (tr.debugID.present()) {
			if (!debugID.present()) {
				debugID = nondeterministicRandom()->randomUniqueID();
			}

			g_traceBatch.addAttach("CommitAttachID", tr.debugID.get().first(), debugID.get().first());
		}
		span.addLink(tr.spanContext);
	}

	if (debugID.present()) {
		g_traceBatch.addEvent("CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.Before");
	}
}

void CommitBatchContext::evaluateBatchSize() {
	for (const auto& tr : trs) {
		const auto& mutations = tr.transaction.mutations;
		batchOperations += mutations.size();
		batchBytes += mutations.expectedSize();
	}
}

// Try to identify recovery transaction and backup's apply mutations (blind writes).
// Both cannot be rejected and are approximated by looking at first mutation
// starting with 0xff.
bool canReject(const std::vector<CommitTransactionRequest>& trs) {
	for (const auto& tr : trs) {
		if (tr.transaction.mutations.empty())
			continue;
		if (tr.transaction.mutations[0].param1.startsWith("\xff"_sr) || tr.transaction.read_conflict_ranges.empty()) {
			return false;
		}
	}
	return true;
}

double computeReleaseDelay(CommitBatchContext* self, double latencyBucket) {
	return std::min(SERVER_KNOBS->MAX_PROXY_COMPUTE,
	                self->batchOperations * self->pProxyCommitData->commitComputePerOperation[latencyBucket]);
}

Future<Void> preresolutionProcessing(CommitBatchContext* self) {

	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	std::vector<CommitTransactionRequest>& trs = self->trs;
	const int64_t localBatchNumber = self->localBatchNumber;
	const int latencyBucket = self->latencyBucket;
	const Optional<UID>& debugID = self->debugID;
	Span span("MP:preresolutionProcessing"_loc, self->span.context);
	double startTime = g_network->timer_monotonic();

	if (self->localBatchNumber - self->pProxyCommitData->latestLocalCommitBatchResolving.get() >
	        SERVER_KNOBS->RESET_MASTER_BATCHES &&
	    now() - self->pProxyCommitData->lastMasterReset > SERVER_KNOBS->RESET_MASTER_DELAY) {
		TraceEvent(SevWarnAlways, "ResetMasterNetwork", self->pProxyCommitData->dbgid)
		    .detail("CurrentBatch", self->localBatchNumber)
		    .detail("InProcessBatch", self->pProxyCommitData->latestLocalCommitBatchResolving.get());
		FlowTransport::transport().resetConnection(self->pProxyCommitData->master.address());
		self->pProxyCommitData->lastMasterReset = now();
	}

	// Pre-resolution the commits
	CODE_PROBE(pProxyCommitData->latestLocalCommitBatchResolving.get() < localBatchNumber - 1, "Wait for local batch");
	co_await pProxyCommitData->latestLocalCommitBatchResolving.whenAtLeast(localBatchNumber - 1);
	double queuingDelay = g_network->timer_monotonic() - startTime;
	pProxyCommitData->stats.computeLatency.addMeasurement(queuingDelay);
	pProxyCommitData->stats.commitBatchQueuingDist->sampleSeconds(queuingDelay);
	if ((queuingDelay > (double)SERVER_KNOBS->MAX_READ_TRANSACTION_LIFE_VERSIONS / SERVER_KNOBS->VERSIONS_PER_SECOND ||
	     (g_network->isSimulated() && buggify(0.01))) &&
	    SERVER_KNOBS->PROXY_REJECT_BATCH_QUEUED_TOO_LONG && canReject(trs)) {
		// Disabled for the recovery transaction. otherwise, recovery can't finish and keeps doing more recoveries.
		CODE_PROBE(true, "Reject transactions in the batch");
		TraceEvent(g_network->isSimulated() ? SevInfo : SevWarnAlways, "ProxyReject", pProxyCommitData->dbgid)
		    .suppressFor(0.1)
		    .detail("QDelay", queuingDelay)
		    .detail("Transactions", trs.size())
		    .detail("BatchNumber", localBatchNumber);
		ASSERT(pProxyCommitData->latestLocalCommitBatchResolving.get() == localBatchNumber - 1);
		pProxyCommitData->latestLocalCommitBatchResolving.set(localBatchNumber);

		co_await pProxyCommitData->latestLocalCommitBatchLogging.whenAtLeast(localBatchNumber - 1);
		ASSERT(pProxyCommitData->latestLocalCommitBatchLogging.get() == localBatchNumber - 1);
		pProxyCommitData->latestLocalCommitBatchLogging.set(localBatchNumber);
		for (const auto& tr : trs) {
			tr.reply.sendError(transaction_too_old());
		}
		++pProxyCommitData->stats.commitBatchOut;
		pProxyCommitData->stats.txnCommitOut += trs.size();
		pProxyCommitData->stats.txnRejectedForQueuedTooLong += trs.size();
		self->rejected = true;
		co_return;
	}

	self->releaseDelay = delay(computeReleaseDelay(self, latencyBucket), TaskPriority::ProxyMasterVersionReply);

	if (debugID.present()) {
		g_traceBatch.addEvent(
		    "CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.GettingCommitVersion");
	}

	if (SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST) {
		self->writtenTagsPreResolution = self->getWrittenTagsPreResolution();
	}

	if (SERVER_KNOBS->HOT_SHARD_THROTTLING_ENABLED && !pProxyCommitData->hotShards.empty()) {
		self->checkHotShards();
	}

	GetCommitVersionRequest req(span.context,
	                            pProxyCommitData->commitVersionRequestNumber++,
	                            pProxyCommitData->mostRecentProcessedRequestNumber,
	                            pProxyCommitData->dbgid);
	double beforeGettingCommitVersion = g_network->timer_monotonic();
	GetCommitVersionReply versionReply = co_await brokenPromiseToNever(
	    pProxyCommitData->master.getCommitVersion.getReply(req, TaskPriority::ProxyMasterVersionReply));

	pProxyCommitData->mostRecentProcessedRequestNumber = versionReply.requestNum;

	pProxyCommitData->stats.txnCommitVersionAssigned += trs.size();
	pProxyCommitData->stats.lastCommitVersionAssigned = versionReply.version;
	pProxyCommitData->stats.getCommitVersionDist->sampleSeconds(g_network->timer_monotonic() -
	                                                            beforeGettingCommitVersion);

	self->commitVersion = versionReply.version;
	self->prevVersion = versionReply.prevVersion;

	// TraceEvent("CPGetVersion", pProxyCommitData->dbgid).detail("Master", pProxyCommitData->master.id().toString()).detail("CommitVersion", self->commitVersion).detail("PrvVersion", self->prevVersion);

	for (auto it : versionReply.resolverChanges) {
		auto rs = pProxyCommitData->keyResolvers.modify(it.range);
		for (auto r = rs.begin(); r != rs.end(); ++r)
			r->value().emplace_back(versionReply.resolverChangesVersion, it.dest);
	}

	//TraceEvent("ProxyGotVer", pProxyContext->dbgid).detail("Commit", commitVersion).detail("Prev", prevVersion);

	if (debugID.present()) {
		g_traceBatch.addEvent("CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.GotCommitVersion");
	}
}

Future<Void> getResolution(CommitBatchContext* self) {
	double resolutionStart = g_network->timer_monotonic();
	// Sending these requests is the fuzzy border between phase 1 and phase 2; it could conceivably overlap with
	// resolution processing but is still using CPU
	ProxyCommitData* pProxyCommitData = self->pProxyCommitData;
	std::vector<CommitTransactionRequest>& trs = self->trs;
	Span span("MP:getResolution"_loc, self->span.context);

	ResolutionRequestBuilder requests(pProxyCommitData,
	                                  self->commitVersion,
	                                  self->prevVersion,
	                                  pProxyCommitData->version.get(),
	                                  self->lastShardMove,
	                                  span);
	int conflictRangeCount = 0;
	self->maxTransactionBytes = 0;
	for (int t = 0; t < trs.size(); t++) {
		requests.addTransaction(trs[t], self->commitVersion, t);
		conflictRangeCount +=
		    trs[t].transaction.read_conflict_ranges.size() + trs[t].transaction.write_conflict_ranges.size();
		//TraceEvent("MPTransactionDump", self->dbgid).detail("Snapshot", trs[t].transaction.read_snapshot);
		// for(auto& m : trs[t].transaction.mutations)
		self->maxTransactionBytes = std::max<int64_t>(self->maxTransactionBytes, trs[t].transaction.expectedSize());
		//	TraceEvent("MPTransactionsDump", self->dbgid).detail("Mutation", m.toString());
	}
	pProxyCommitData->stats.conflictRanges += conflictRangeCount;

	for (int r = 1; r < pProxyCommitData->resolvers.size(); r++)
		ASSERT(requests.requests[r].txnStateTransactions.size() == requests.requests[0].txnStateTransactions.size());

	pProxyCommitData->stats.txnCommitResolving += trs.size();
	std::vector<Future<ResolveTransactionBatchReply>> replies;
	Future<ResolveTransactionBatchReply> singleResolverReply;
	double singleResolverStart = 0;
	if (pProxyCommitData->resolvers.size() == 1) {
		requests.requests[0].debugID = self->debugID;
		requests.requests[0].writtenTags = self->writtenTagsPreResolution;
		singleResolverStart = g_network->timer_monotonic();
		singleResolverReply = brokenPromiseToNever(
		    pProxyCommitData->resolvers[0].resolve.getReply(requests.requests[0], TaskPriority::ProxyResolverReply));
	} else {
		for (int r = 0; r < pProxyCommitData->resolvers.size(); r++) {
			requests.requests[r].debugID = self->debugID;
			requests.requests[r].writtenTags = self->writtenTagsPreResolution;
			replies.push_back(
			    trackResolutionMetrics(pProxyCommitData->stats.resolverDist[r],
			                           brokenPromiseToNever(pProxyCommitData->resolvers[r].resolve.getReply(
			                               requests.requests[r], TaskPriority::ProxyResolverReply))));
		}
	}

	self->transactionResolverMap.swap(requests.transactionResolverMap);
	// Used to report conflicting keys
	self->txReadConflictRangeIndexMap.swap(requests.txReadConflictRangeIndexMap);

	self->releaseFuture = releaseResolvingAfter(pProxyCommitData, self->releaseDelay, self->localBatchNumber);

	if (self->localBatchNumber - self->pProxyCommitData->latestLocalCommitBatchLogging.get() >
	        SERVER_KNOBS->RESET_RESOLVER_BATCHES &&
	    now() - self->pProxyCommitData->lastResolverReset > SERVER_KNOBS->RESET_RESOLVER_DELAY) {
		for (int r = 0; r < self->pProxyCommitData->resolvers.size(); r++) {
			TraceEvent(SevWarnAlways, "ResetResolverNetwork", self->pProxyCommitData->dbgid)
			    .detail("PeerAddr", self->pProxyCommitData->resolvers[r].address())
			    .detail("PeerAddress", self->pProxyCommitData->resolvers[r].address())
			    .detail("CurrentBatch", self->localBatchNumber)
			    .detail("InProcessBatch", self->pProxyCommitData->latestLocalCommitBatchLogging.get());
			FlowTransport::transport().resetConnection(self->pProxyCommitData->resolvers[r].address());
		}
		self->pProxyCommitData->lastResolverReset = now();
	}

	// Wait for the final resolution
	if (pProxyCommitData->resolvers.size() == 1) {
		ResolveTransactionBatchReply resolutionResp = co_await singleResolverReply;
		pProxyCommitData->stats.resolverDist[0]->sampleSeconds(g_network->timer_monotonic() - singleResolverStart);
		self->resolution.clear();
		self->resolution.push_back(std::move(resolutionResp));
	} else {
		std::vector<ResolveTransactionBatchReply> resolutionResp = co_await getAll(replies);
		self->resolution = std::move(resolutionResp);
	}

	self->pProxyCommitData->stats.resolutionDist->sampleSeconds(g_network->timer_monotonic() - resolutionStart);
	if (self->debugID.present()) {
		g_traceBatch.addEvent(
		    "CommitDebug", self->debugID.get().first(), "CommitProxyServer.commitBatch.AfterResolution");
	}
}

void assertResolutionStateMutationsSizeConsistent(const std::vector<ResolveTransactionBatchReply>& resolution) {
	for (int r = 1; r < resolution.size(); r++) {
		ASSERT(resolution[r].stateMutations.size() == resolution[0].stateMutations.size());
		for (int s = 0; s < resolution[r].stateMutations.size(); s++) {
			ASSERT(resolution[r].stateMutations[s].size() == resolution[0].stateMutations[s].size());
		}
	}
}

// If the splitMutations is not empty, which means some clear range in mutations are split into multiple clear range
// ops. Modify mutations by replace the old clear range with the split clear ranges
void replaceRawClearRanges(Arena& arena,
                           VectorRef<MutationRef>& mutations,
                           std::vector<std::pair<int, std::vector<MutationRef>>>& splitMutations,
                           size_t totalSize,
                           Optional<UID> debugId = Optional<UID>()) {
	if (splitMutations.empty())
		return;

	int i = mutations.size() - 1;
	mutations.resize(arena, totalSize);
	// place from back
	int curr = totalSize - 1;
	for (; i >= 0; --i) {
		if (splitMutations.empty()) {
			ASSERT_EQ(curr, i);
			break;
		}

		if (splitMutations.back().first == i) {
			ASSERT_EQ(mutations[i].type, MutationRef::ClearRange);
			// TODO(gglass): legacy comment below references tenant. Possibly some
			// opportunity for simplification here. Legacy comment:
			// replace with tenant aligned mutations
			auto& currMutations = splitMutations.back().second;
			while (!currMutations.empty()) {
				mutations[curr] = currMutations.back();
				currMutations.pop_back();
				curr--;
			}
			splitMutations.pop_back();
		} else {
			ASSERT_GT(curr, i);
			mutations[curr] = mutations[i];
			curr--;
		}
	}

	ASSERT_EQ(splitMutations.size(), 0);
}
// Acknowledge transaction state store commits.
// Note: This acknowledgement will cause the transaction state store's popped version ("poppedUpTo", that's
// maintained in LogSystemDiskQueueAdapter) to get updated.
void acknowledgeTransactionStateStoreCommits(CommitBatchContext* self) {
	for (auto& p : self->storeCommits) {
		ASSERT(!p.second.isReady());
		p.first.get().acknowledge.send(Void());
		ASSERT(p.second.isReady());
	}
}

// Compute and apply "metadata" effects of each other proxy's most recent batch
void applyMetadataEffect(CommitBatchContext* self) {
	bool initialState = self->isMyFirstBatch;
	self->firstStateMutations = self->isMyFirstBatch;
	for (int versionIndex = 0; versionIndex < self->resolution[0].stateMutations.size(); versionIndex++) {
		// pProxyCommitData->logAdapter->setNextVersion( ??? );  << Ideally we would be telling the log adapter that the
		// pushes in this commit will be in the version at which these state mutations were committed by another proxy,
		// but at present we don't have that information here.  So the disk queue may be unnecessarily conservative
		// about popping.

		for (int transactionIndex = 0;
		     transactionIndex < self->resolution[0].stateMutations[versionIndex].size() && !self->forceRecovery;
		     transactionIndex++) {
			bool committed = true;
			for (int resolver = 0; resolver < self->resolution.size(); resolver++) {
				committed =
				    committed && self->resolution[resolver].stateMutations[versionIndex][transactionIndex].committed;
			}

			if (committed) {
				applyMetadataMutations(SpanContext(),
				                       self->pProxyCommitData->getApplyMetadataProxyContext(),
				                       self->arena,
				                       self->pProxyCommitData->logSystemConsumer,
				                       self->resolution[0].stateMutations[versionIndex][transactionIndex].mutations,
				                       /* pToCommit= */ nullptr,
				                       self->forceRecovery,
				                       /* version= */ self->commitVersion,
				                       /* popVersion= */ 0,
				                       /* initialCommit */ false,
				                       /* provisionalCommitProxy */ self->pProxyCommitData->provisional);
			}

			if (!self->resolution[0].stateMutations[versionIndex][transactionIndex].mutations.empty() &&
			    self->firstStateMutations) {
				ASSERT(committed);
				self->firstStateMutations = false;
				self->forceRecovery = false;
			}
		}
		// These changes to txnStateStore will be committed by the other proxy, so we simply discard the commit message
		auto fcm = self->pProxyCommitData->logAdapter->getCommitMessage();
		self->storeCommits.emplace_back(fcm, self->pProxyCommitData->txnStateStore->commit());

		if (initialState) {
			initialState = false;
			self->forceRecovery = false;
			self->pProxyCommitData->txnStateStore->resyncLog();

			acknowledgeTransactionStateStoreCommits(self);
			self->storeCommits.clear();
		}
	}
}

/// Determine which transactions actually committed (conservatively) by combining results from the resolvers
void determineCommittedTransactions(CommitBatchContext* self) {
	auto pProxyCommitData = self->pProxyCommitData;
	const auto& trs = self->trs;

	ASSERT(self->transactionResolverMap.size() == self->committed.size());
	// For each commitTransactionRef, it is only sent to resolvers specified in transactionResolverMap
	// Thus, we use this nextTr to track the correct transaction index on each resolver.
	self->nextTr.resize(self->resolution.size());
	for (int t = 0; t < trs.size(); t++) {
		uint8_t commit = ConflictBatchStatus::TransactionCommitted;
		for (int r : self->transactionResolverMap[t]) {
			commit = std::min(self->resolution[r].committed[self->nextTr[r]++], commit);
		}
		self->committed[t] = commit;
	}
	for (int r = 0; r < self->resolution.size(); r++)
		ASSERT(self->nextTr[r] == self->resolution[r].committed.size());

	pProxyCommitData->logAdapter->setNextVersion(self->commitVersion);

	self->lockedKey = pProxyCommitData->txnStateStore->readValue(databaseLockedKey).get();
	self->locked = self->lockedKey.present() && !self->lockedKey.get().empty();

	const Optional<Value> mustContainSystemKey =
	    pProxyCommitData->txnStateStore->readValue(mustContainSystemMutationsKey).get();
	if (mustContainSystemKey.present() && !mustContainSystemKey.get().empty()) {
		for (int t = 0; t < trs.size(); t++) {
			if (self->committed[t] == ConflictBatchStatus::TransactionCommitted) {
				bool foundSystem = false;
				for (auto& m : trs[t].transaction.mutations) {
					if ((m.type == MutationRef::ClearRange ? m.param2 : m.param1) >= nonMetadataSystemKeys.end) {
						foundSystem = true;
						break;
					}
				}
				if (!foundSystem) {
					self->committed[t] = ConflictBatchStatus::TransactionConflict;
				}
			}
		}
	}
}

// This first pass through committed transactions deals with "metadata" effects (modifications of txnStateStore, changes
// to storage servers' responsibilities)
Future<Void> changeCoordinatorsForMetadata(CommitBatchContext* self) {
	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	co_await brokenPromiseToNever(pProxyCommitData->db->get().clusterInterface.changeCoordinators.getReply(
	    ChangeCoordinatorsRequest(pProxyCommitData->txnStateStore->readValue(coordinatorsKey).get().get(),
	                              self->pProxyCommitData->master.id())));
	ASSERT(false); // ChangeCoordinatorsRequest should always throw
}

Future<Void> applyMetadataToCommittedTransactions(CommitBatchContext* self) {
	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	auto& trs = self->trs;

	int t;
	for (t = 0; t < trs.size() && !self->forceRecovery; t++) {
		if (self->committed[t] == ConflictBatchStatus::TransactionCommitted &&
		    (!self->locked || trs[t].isLockAware())) {
			self->commitCount++;
			applyMetadataMutations(trs[t].spanContext,
			                       pProxyCommitData->getApplyMetadataProxyContext(),
			                       self->arena,
			                       pProxyCommitData->logSystemConsumer,
			                       trs[t].transaction.mutations,
			                       SERVER_KNOBS->PROXY_USE_RESOLVER_PRIVATE_MUTATIONS ? nullptr : &self->toCommit,
			                       self->forceRecovery,
			                       self->commitVersion,
			                       self->commitVersion + 1,
			                       /* initialCommit= */ false,
			                       /* provisionalCommitProxy */ self->pProxyCommitData->provisional);
		}

		if (self->firstStateMutations) {
			ASSERT(self->committed[t] == ConflictBatchStatus::TransactionCommitted);
			self->firstStateMutations = false;
			self->forceRecovery = false;
		}
	}

	if (self->forceRecovery) {
		for (; t < trs.size(); t++)
			self->committed[t] = ConflictBatchStatus::TransactionConflict;
		TraceEvent(SevWarn, "RestartingTxnSubsystem", pProxyCommitData->dbgid).detail("Stage", "AwaitCommit");
	}
	if (SERVER_KNOBS->PROXY_USE_RESOLVER_PRIVATE_MUTATIONS) {
		// Resolver also calculates forceRecovery and only applies metadata mutations
		// in the same set of transactions as this proxy.
		ResolveTransactionBatchReply& reply = self->resolution[0];
		self->toCommit.setMutations(reply.privateMutationCount, reply.privateMutations);
		if (SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST) {
			// TraceEvent("ResolverReturn").detail("ReturnTags",reply.writtenTags).detail("TPCVsize",reply.tpcvMap.size()).detail("ReqTags",self->writtenTagsPreResolution);
			self->tpcvMap = reply.tpcvMap;
			self->pProxyCommitData->lastShardMove = reply.lastShardMove;

			// extract push locations from tpcv
			std::vector<int> fromLocations;
			fromLocations.reserve(reply.tpcvMap.size());
			for (const auto& pair : self->tpcvMap) {
				fromLocations.push_back(pair.first);
			}
			// save push locations for each tag
			self->toCommit.setPushLocationsForTags(fromLocations);
		}
		self->toCommit.addWrittenTags(reply.writtenTags);
	}

	self->lockedKey = pProxyCommitData->txnStateStore->readValue(databaseLockedKey).get();
	self->lockedAfter = self->lockedKey.present() && !self->lockedKey.get().empty();

	self->metadataVersionAfter = pProxyCommitData->txnStateStore->readValue(metadataVersionKey).get();

	auto fcm = pProxyCommitData->logAdapter->getCommitMessage();
	self->storeCommits.emplace_back(fcm, pProxyCommitData->txnStateStore->commit());
	pProxyCommitData->version.set(self->commitVersion);
	if (!pProxyCommitData->validState.isSet())
		pProxyCommitData->validState.send(Void());
	ASSERT(self->commitVersion);

	if (!self->isMyFirstBatch &&
	    pProxyCommitData->txnStateStore->readValue(coordinatorsKey).get().get() != self->previousCoordinators.get()) {
		return changeCoordinatorsForMetadata(self);
	}

	return Void();
}

WriteMutationRefVar writeMutation(CommitBatchContext* self, const MutationRef* mutation) {
	self->toCommit.writeTypedMessage(*mutation);
	return std::variant<MutationRef, VectorRef<MutationRef>>{ *mutation };
}

void pushToBackupMutations(CommitBatchContext* self,
                           ProxyCommitData* const pProxyCommitData,
                           Arena& arena,
                           MutationRef const& m,
                           MutationRef const& writtenMutation) {
	if (m.type != MutationRef::Type::ClearRange) {
		// Add the mutation to the relevant backup tag
		for (const auto& backupName : pProxyCommitData->vecBackupKeys[m.param1]) {
			self->logRangeMutations[backupName].push_back_deep(self->logRangeMutationsArena, writtenMutation);
		}
	} else {
		KeyRangeRef mutationRange(m.param1, m.param2);
		KeyRangeRef intersectionRange;

		// Identify and add the intersecting ranges of the mutation to the array of mutations to serialize
		for (auto backupRange : pProxyCommitData->vecBackupKeys.intersectingRanges(mutationRange)) {
			// Get the backup sub range
			const auto& backupSubrange = backupRange.range();

			// Determine the intersecting range
			intersectionRange = mutationRange & backupSubrange;

			// Create the custom mutation for the specific backup tag
			MutationRef backupMutation(MutationRef::Type::ClearRange, intersectionRange.begin, intersectionRange.end);

			// Add the mutation to the relevant backup tag
			for (const auto& backupName : backupRange.value()) {
				self->logRangeMutations[backupName].push_back_deep(self->logRangeMutationsArena, backupMutation);
			}
		}
	}
}

void addAccumulativeChecksumMutations(CommitBatchContext* self) {
	ASSERT(self->pProxyCommitData->acsBuilder != nullptr);
	const uint16_t acsIndex = getCommitProxyAccumulativeChecksumIndex(self->pProxyCommitData->commitProxyIndex);
	for (const auto& [tag, acsState] : self->pProxyCommitData->acsBuilder->getAcsTable()) {
		ASSERT(tagSupportAccumulativeChecksum(tag));
		ASSERT(acsState.version <= self->commitVersion);
		if (acsState.version < self->commitVersion) {
			// Have not updated in the current commit batch
			// So, need not send acs mutation for this tag
			continue;
		}
		ASSERT(acsState.epoch == self->pProxyCommitData->epoch);
		MutationRef acsMutation;
		acsMutation.type = MutationRef::SetValue;
		acsMutation.param1 = accumulativeChecksumKey; // private mutation
		AccumulativeChecksumState acsToSend(acsIndex, acsState.acs, self->commitVersion, self->pProxyCommitData->epoch);
		Value acsValue = accumulativeChecksumValue(acsToSend);
		acsMutation.param2 = acsValue;
		acsMutation.setAccumulativeChecksumIndex(acsIndex);
		if (CLIENT_KNOBS->ENABLE_ACCUMULATIVE_CHECKSUM_LOGGING) {
			TraceEvent(SevInfo, "AcsBuilderIssueAccumulativeChecksumMutation", self->pProxyCommitData->dbgid)
			    .detail("AcsTag", tag)
			    .detail("AcsIndex", acsIndex)
			    .detail("AcsToSend", acsToSend.toString())
			    .detail("Mutation", acsMutation)
			    .detail("Version", self->commitVersion)
			    .detail("CommitProxyIndex", self->pProxyCommitData->commitProxyIndex);
		}
		DEBUG_MUTATION("ProxyCommit", self->commitVersion, acsMutation, self->pProxyCommitData->dbgid);
		self->toCommit.addTag(tag);
		self->toCommit.writeTypedMessage(acsMutation);
	}
}

void rejectMutationsForReadLockOnRange(CommitBatchContext* self) {
	ASSERT(self->rangeLockEnabled());
	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	ASSERT(pProxyCommitData->rangeLock != nullptr);
	// Fast path: no exclusive locks held -> nothing to check, skip the
	// per-mutation loop entirely. Steady state when no bulkload is running.
	if (!pProxyCommitData->rangeLock->anyExclusiveLockHeld()) {
		++pProxyCommitData->stats.rangeLockFastPath;
		return;
	}
	++pProxyCommitData->stats.rangeLockSlowPath;
	std::vector<CommitTransactionRequest>& trs = self->trs;
	for (int i = self->transactionNum; i < trs.size(); i++) {
		if (self->committed[i] != ConflictBatchStatus::TransactionCommitted) {
			continue;
		} else if (trs[i].isLockAware()) {
			continue; // rangeLock is transparent to lock-aware transactions
		}
		VectorRef<MutationRef>* pMutations = &trs[i].transaction.mutations;
		for (int j = 0; j < pMutations->size(); j++) {
			MutationRef m = (*pMutations)[j];
			KeyRange rangeToCheck;
			if (isSingleKeyMutation((MutationRef::Type)m.type)) {
				rangeToCheck = singleKeyRange(m.param1);
			} else if (m.type == MutationRef::ClearRange) {
				rangeToCheck = KeyRangeRef(m.param1, m.param2);
			}
			bool shouldReject = pProxyCommitData->rangeLock->isLocked(rangeToCheck);
			if (shouldReject) {
				self->committed[i] = ConflictBatchStatus::TransactionLockReject;
				trs[i].reply.sendError(transaction_rejected_range_locked());
				break;
			}
		}
	}
}

/// This second pass through committed transactions assigns the actual mutations to the appropriate storage servers'
/// tags
Future<Void> assignMutationsToStorageServers(CommitBatchContext* self) {
	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	std::vector<CommitTransactionRequest>& trs = self->trs;

	for (; self->transactionNum < trs.size(); self->transactionNum++) {
		if (!(self->committed[self->transactionNum] == ConflictBatchStatus::TransactionCommitted &&
		      (!self->locked || trs[self->transactionNum].isLockAware()))) {
			continue;
		}

		bool checkSample = trs[self->transactionNum].commitCostEstimation.present();
		Optional<ClientTrCommitCostEstimation>* trCost = &trs[self->transactionNum].commitCostEstimation;
		int mutationNum = 0;
		VectorRef<MutationRef>* pMutations = &trs[self->transactionNum].transaction.mutations;

		self->toCommit.addTransactionInfo(trs[self->transactionNum].spanContext);

		for (; mutationNum < pMutations->size(); mutationNum++) {
			if (self->yieldBytes > SERVER_KNOBS->DESIRED_TOTAL_BYTES) {
				self->yieldBytes = 0;
				if (g_network->check_yield(TaskPriority::ProxyCommitYield1)) {
					self->computeDuration += g_network->timer_monotonic() - self->computeStart;
					co_await delay(0, TaskPriority::ProxyCommitYield1);
					self->computeStart = g_network->timer_monotonic();
				}
			}

			MutationRef m = (*pMutations)[mutationNum];
			Arena arena;
			MutationRef writtenMutation;
			self->mutationCount++;
			self->mutationBytes += m.expectedSize();
			self->yieldBytes += m.expectedSize();

			// Determine the set of tags (responsible storage servers) for the mutation, splitting it
			// if necessary.  Serialize (splits of) the mutation into the message buffer and add the tags.
			if (isSingleKeyMutation((MutationRef::Type)m.type)) {
				auto& tags = pProxyCommitData->tagsForKey(m.param1);

				// sample single key mutation based on cost
				// the expectation of sampling is every COMMIT_SAMPLE_COST sample once
				if (checkSample) {
					double totalCosts = trCost->get().writeCosts;
					double cost = getWriteOperationCost(m.expectedSize());
					double mul = std::max(1.0, totalCosts / std::max(1.0, (double)CLIENT_KNOBS->COMMIT_SAMPLE_COST));
					ASSERT(totalCosts > 0);
					double prob = mul * cost / totalCosts;

					if (deterministicRandom()->random01() < prob) {
						const auto& storageServers = pProxyCommitData->keyInfo[m.param1].src_info;
						for (const auto& ssInfo : storageServers) {
							auto id = ssInfo->interf.id();
							// scale cost
							cost = cost < CLIENT_KNOBS->COMMIT_SAMPLE_COST ? CLIENT_KNOBS->COMMIT_SAMPLE_COST : cost;
							pProxyCommitData->updateSSTagCost(
							    id, trs[self->transactionNum].tagSet.get(), m, cost / storageServers.size());
						}
					}
				}

				DEBUG_MUTATION("ProxyCommit", self->commitVersion, m, pProxyCommitData->dbgid).detail("To", tags);
				self->toCommit.addTags(tags);
				if (!pProxyCommitData->cdcRouting.empty()) {
					self->toCommit.addTags(pProxyCommitData->cdcRouting.tagsForKey(m.param1));
				}

				if (pProxyCommitData->acsBuilder != nullptr) {
					updateMutationWithAcsAndAddMutationToAcsBuilder(
					    pProxyCommitData->acsBuilder,
					    m,
					    tags,
					    getCommitProxyAccumulativeChecksumIndex(pProxyCommitData->commitProxyIndex),
					    pProxyCommitData->epoch,
					    self->commitVersion,
					    pProxyCommitData->dbgid);
				}

				WriteMutationRefVar var = writeMutation(self, &m);
				// FIXME: Remove assert once ClearRange RAW_ACCESS usecase handling is done
				ASSERT(std::holds_alternative<MutationRef>(var));
				writtenMutation = std::get<MutationRef>(var);
			} else if (m.type == MutationRef::ClearRange) {
				auto range = pProxyCommitData->keyInfo.rangeContaining(m.param1);
				if (range.end() >= m.param2) {
					// Fast path
					DEBUG_MUTATION("ProxyCommit", self->commitVersion, m, pProxyCommitData->dbgid)
					    .detail("To", range.value().tags);
					range.value().populateTags();
					self->toCommit.addTags(range.value().tags);

					if (pProxyCommitData->acsBuilder != nullptr) {
						updateMutationWithAcsAndAddMutationToAcsBuilder(
						    pProxyCommitData->acsBuilder,
						    m,
						    range.value().tags,
						    getCommitProxyAccumulativeChecksumIndex(pProxyCommitData->commitProxyIndex),
						    pProxyCommitData->epoch,
						    self->commitVersion,
						    pProxyCommitData->dbgid);
					}

					// check whether clear is sampled
					if (checkSample && !trCost->get().clearIdxCosts.empty() &&
					    trCost->get().clearIdxCosts[0].first == mutationNum) {
						auto const& ssInfos = range.value().src_info;
						for (auto const& ssInfo : ssInfos) {
							auto id = ssInfo->interf.id();
							pProxyCommitData->updateSSTagCost(id,
							                                  trs[self->transactionNum].tagSet.get(),
							                                  m,
							                                  trCost->get().clearIdxCosts[0].second / ssInfos.size());
						}
						trCost->get().clearIdxCosts.pop_front();
					}
				} else {
					CODE_PROBE(true, "A clear range extends past a shard boundary");
					std::set<Tag> allSources;
					while (range.begin() < m.param2) {
						range.value().populateTags();
						allSources.insert(range.value().tags.begin(), range.value().tags.end());

						// check whether clear is sampled
						if (checkSample && !trCost->get().clearIdxCosts.empty() &&
						    trCost->get().clearIdxCosts[0].first == mutationNum) {
							auto const& ssInfos = range.value().src_info;
							for (auto const& ssInfo : ssInfos) {
								auto id = ssInfo->interf.id();
								pProxyCommitData->updateSSTagCost(id,
								                                  trs[self->transactionNum].tagSet.get(),
								                                  m,
								                                  trCost->get().clearIdxCosts[0].second /
								                                      ssInfos.size());
							}
							trCost->get().clearIdxCosts.pop_front();
						}
						++range;
					}

					DEBUG_MUTATION("ProxyCommit", self->commitVersion, m)
					    .detail("Dbgid", pProxyCommitData->dbgid)
					    .detail("To", allSources);
					self->toCommit.addTags(allSources);

					if (self->pProxyCommitData->acsBuilder != nullptr) {
						updateMutationWithAcsAndAddMutationToAcsBuilder(
						    pProxyCommitData->acsBuilder,
						    m,
						    allSources,
						    getCommitProxyAccumulativeChecksumIndex(pProxyCommitData->commitProxyIndex),
						    pProxyCommitData->epoch,
						    self->commitVersion,
						    pProxyCommitData->dbgid);
					}
				}

				KeyRangeRef clearRange(KeyRangeRef(m.param1, m.param2));
				if (!pProxyCommitData->cdcRouting.empty()) {
					self->toCommit.addTags(pProxyCommitData->cdcRouting.tagsForRange(clearRange));
				}
				WriteMutationRefVar var = writeMutation(self, &m);
				// FIXME: Remove assert once ClearRange RAW_ACCESS usecase handling is done
				ASSERT(std::holds_alternative<MutationRef>(var));
				writtenMutation = std::get<MutationRef>(var);
			} else if (m.type == MutationRef::NoOp) {
				// TODO(gglass): what is the deal with MutationRef::NoOp? Is it needed?
				// This used to be the following:
				// ASSERT_EQ(pProxyCommitData->getTenantMode(), TenantMode::REQUIRED);
				ASSERT(false);
				continue;
			} else {
				UNREACHABLE();
			}

			DisabledTraceEvent(SevDebug, "BeforeBackup", pProxyCommitData->dbgid)
			    .detail("M1", m.param1)
			    .detail("M2", m.param2)
			    .detail("MT", getTypeString(m.type))
			    .detail("VecBackupKeys", pProxyCommitData->vecBackupKeys.size())
			    .detail("ShouldBackup", shouldBackup(m));

			if (pProxyCommitData->vecBackupKeys.size() <= 1 || !shouldBackup(m)) {
				continue;
			}

			pushToBackupMutations(self, pProxyCommitData, arena, m, writtenMutation);
		}

		if (checkSample) {
			self->pProxyCommitData->stats.txnExpensiveClearCostEstCount +=
			    trs[self->transactionNum].commitCostEstimation.get().expensiveCostEstCount;
		}
	}
}

Future<Void> postResolution(CommitBatchContext* self) {
	double postResolutionStart = g_network->timer_monotonic();
	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	std::vector<CommitTransactionRequest>& trs = self->trs;
	const int64_t localBatchNumber = self->localBatchNumber;
	const Optional<UID>& debugID = self->debugID;
	Span span("MP:postResolution"_loc, self->span.context);

	bool queuedCommits = pProxyCommitData->latestLocalCommitBatchLogging.get() < localBatchNumber - 1;
	CODE_PROBE(queuedCommits, "Queuing post-resolution commit processing");
	co_await pProxyCommitData->latestLocalCommitBatchLogging.whenAtLeast(localBatchNumber - 1);
	double postResolutionQueuing = g_network->timer_monotonic();
	pProxyCommitData->stats.postResolutionDist->sampleSeconds(postResolutionQueuing - postResolutionStart);
	co_await yield(TaskPriority::ProxyCommitYield1);

	self->computeStart = g_network->timer_monotonic();

	pProxyCommitData->stats.txnCommitResolved += trs.size();

	if (debugID.present()) {
		g_traceBatch.addEvent(
		    "CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.ProcessingMutations");
	}

	self->isMyFirstBatch = !pProxyCommitData->version.get();
	self->previousCoordinators = pProxyCommitData->txnStateStore->readValue(coordinatorsKey).get();

	assertResolutionStateMutationsSizeConsistent(self->resolution);

	applyMetadataEffect(self);

	if (debugID.present()) {
		g_traceBatch.addEvent(
		    "CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.ApplyMetadataEffect");
	}

	determineCommittedTransactions(self);

	if (debugID.present()) {
		g_traceBatch.addEvent(
		    "CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.DetermineCommittedTransactions");
	}

	if (self->forceRecovery) {
		co_await Future<Void>(Never());
	}

	// First pass
	co_await applyMetadataToCommittedTransactions(self);

	if (debugID.present()) {
		g_traceBatch.addEvent(
		    "CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.ApplyMetadataToCommittedTxn");
	}

	// After applyed metadata change, this commit proxy has the latest view of locked ranges.
	// If a transaction has any mutation accessing to the locked range, reject the transaction with
	// error_code_transaction_rejected_range_locked
	if (self->rangeLockEnabled()) {
		rejectMutationsForReadLockOnRange(self);
	}

	// Second pass
	co_await assignMutationsToStorageServers(self);

	if (debugID.present()) {
		g_traceBatch.addEvent("CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.AssignMutationToSS");
	}

	// Serialize and backup the mutations as a single mutation
	if ((pProxyCommitData->vecBackupKeys.size() > 1) && !self->logRangeMutations.empty()) {
		co_await addBackupMutations(pProxyCommitData,
		                            &self->logRangeMutations,
		                            &self->toCommit,
		                            self->commitVersion,
		                            &self->computeDuration,
		                            &self->computeStart);
	}

	// When version vector is enabled, idempotency entries should only be created or cleared
	// if the operation was detected at pre resolution time. This ensures that the
	// operation is broadcast to all logs, and does not lead to logs being included
	// that are not part of the expected tag set (tpcv).
	buildIdempotencyIdMutations(self->trs,
	                            self->idempotencyKVBuilder,
	                            self->commitVersion,
	                            self->committed,
	                            ConflictBatchStatus::TransactionCommitted,
	                            self->locked,
	                            [&](const KeyValue& kv) {
		                            MutationRef idempotencyIdSet;
		                            idempotencyIdSet.type = MutationRef::Type::SetValue;
		                            idempotencyIdSet.param1 = kv.key;
		                            idempotencyIdSet.param2 = kv.value;
		                            auto& tags = pProxyCommitData->tagsForKey(kv.key);
		                            ASSERT(!SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST ||
		                                   pProxyCommitData->db->get().logSystemConfig.numLogs() ==
		                                       self->tpcvMap.size());
		                            self->toCommit.addTags(tags);
		                            if (pProxyCommitData->acsBuilder != nullptr) {
			                            updateMutationWithAcsAndAddMutationToAcsBuilder(
			                                pProxyCommitData->acsBuilder,
			                                idempotencyIdSet,
			                                tags,
			                                getCommitProxyAccumulativeChecksumIndex(pProxyCommitData->commitProxyIndex),
			                                pProxyCommitData->epoch,
			                                self->commitVersion,
			                                pProxyCommitData->dbgid);
		                            }
		                            self->toCommit.writeTypedMessage(idempotencyIdSet);
	                            });

	if (!SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST ||
	    pProxyCommitData->db->get().logSystemConfig.numLogs() == self->tpcvMap.size()) {
		for (int i = 0; i < pProxyCommitData->idempotencyClears.size(); i++) {
			auto& tags = pProxyCommitData->tagsForKey(pProxyCommitData->idempotencyClears[i].param1);
			self->toCommit.addTags(tags);
			if (pProxyCommitData->acsBuilder != nullptr) {
				updateMutationWithAcsAndAddMutationToAcsBuilder(
				    pProxyCommitData->acsBuilder,
				    pProxyCommitData->idempotencyClears[i],
				    tags,
				    getCommitProxyAccumulativeChecksumIndex(pProxyCommitData->commitProxyIndex),
				    pProxyCommitData->epoch,
				    self->commitVersion,
				    pProxyCommitData->dbgid);
			}
			WriteMutationRefVar var = writeMutation(self, &pProxyCommitData->idempotencyClears[i]);
			ASSERT(std::holds_alternative<MutationRef>(var));
		}
		pProxyCommitData->idempotencyClears = Standalone<VectorRef<MutationRef>>();
	}

	self->toCommit.saveTags(self->writtenTags);

	pProxyCommitData->stats.mutations += self->mutationCount;
	pProxyCommitData->stats.mutationBytes += self->mutationBytes;

	// Storage servers mustn't make durable versions which are not fully committed (because then they are impossible
	// to roll back) We prevent this by limiting the number of versions which are semi-committed but not fully
	// committed to be less than the MVCC window
	if (pProxyCommitData->committedVersion.get() <
	    self->commitVersion - SERVER_KNOBS->MAX_READ_TRANSACTION_LIFE_VERSIONS) {
		self->computeDuration += g_network->timer_monotonic() - self->computeStart;
		Span waitVersionSpan;
		while (pProxyCommitData->committedVersion.get() <
		       self->commitVersion - SERVER_KNOBS->MAX_READ_TRANSACTION_LIFE_VERSIONS) {
			// This should be *extremely* rare in the real world, but knob buggification should make it happen in
			// simulation
			CODE_PROBE(true, "Semi-committed pipeline limited by MVCC window");
			//TraceEvent("ProxyWaitingForCommitted", pProxyCommitData->dbgid).detail("CommittedVersion", pProxyCommitData->committedVersion.get()).detail("NeedToCommit", commitVersion);
			waitVersionSpan = Span("MP:overMaxReadTransactionLifeVersions"_loc, span.context);
			// @todo probably there is no need to get the (entire) version vector from the sequencer
			// in this case, and if so, consider adding a flag to the request to tell the sequencer
			// to not send the version vector information.
			auto res =
			    co_await race(pProxyCommitData->committedVersion.whenAtLeast(
			                      self->commitVersion - SERVER_KNOBS->MAX_READ_TRANSACTION_LIFE_VERSIONS),
			                  pProxyCommitData->cx->onProxiesChanged(),
			                  pProxyCommitData->master.getLiveCommittedVersion.getReply(
			                      GetRawCommittedVersionRequest(waitVersionSpan.context, debugID, invalidVersion),
			                      TaskPriority::GetLiveCommittedVersionReply));
			if (res.index() == 0) {
				co_await yield();
				break;
			} else if (res.index() == 1) {
			} else if (res.index() == 2) {
				GetRawCommittedVersionReply v = std::get<2>(std::move(res));

				if (v.version > pProxyCommitData->committedVersion.get()) {
					pProxyCommitData->locked = v.locked;
					pProxyCommitData->metadataVersion = v.metadataVersion;
					pProxyCommitData->committedVersion.set(v.version);
				}

				if (pProxyCommitData->committedVersion.get() <
				    self->commitVersion - SERVER_KNOBS->MAX_READ_TRANSACTION_LIFE_VERSIONS) {
					co_await delay(SERVER_KNOBS->PROXY_SPIN_DELAY);
				}
			} else {
				UNREACHABLE();
			}
		}
		waitVersionSpan = Span{};
		self->computeStart = g_network->timer_monotonic();
	}

	self->msg = self->storeCommits.back().first.get();

	if (self->debugID.present())
		g_traceBatch.addEvent(
		    "CommitDebug", self->debugID.get().first(), "CommitProxyServer.commitBatch.AfterStoreCommits");

	// txnState (transaction subsystem state) tag: message extracted from log adapter
	bool firstMessage = true;
	for (auto m : self->msg.messages) {
		if (firstMessage) {
			ASSERT(!SERVER_KNOBS->ENABLE_VERSION_VECTOR ||
			       pProxyCommitData->db->get().logSystemConfig.numLogs() == self->tpcvMap.size());
			self->toCommit.addTxsTag();
		}
		self->toCommit.writeMessage(StringRef(m.begin(), m.size()), !firstMessage);
		firstMessage = false;
	}

	if (self->prevVersion && self->commitVersion - self->prevVersion < SERVER_KNOBS->MAX_VERSIONS_IN_FLIGHT / 2)
		debug_advanceMaxCommittedVersion(UID(), self->commitVersion); //< Is this valid?

	// TraceEvent("ProxyPush", pProxyCommitData->dbgid)
	// 	    .detail("PrevVersion", self->prevVersion)
	// 	    .detail("Version", self->commitVersion)
	// 	    .detail("TransactionsSubmitted", trs.size())
	//	    .detail("TransactionsCommitted", self->commitCount)
	//	    .detail("TxsPopTo", self->msg.popTo);

	if (self->prevVersion && self->commitVersion - self->prevVersion < SERVER_KNOBS->MAX_VERSIONS_IN_FLIGHT / 2)
		debug_advanceMaxCommittedVersion(UID(), self->commitVersion);

	self->commitStartTime = now();
	pProxyCommitData->lastStartCommit = self->commitStartTime;
	Optional<std::unordered_map<uint16_t, Version>> tpcvMap = Optional<std::unordered_map<uint16_t, Version>>();
	if (SERVER_KNOBS->ENABLE_VERSION_VECTOR) {
		tpcvMap = self->tpcvMap;
	}
	if (self->pProxyCommitData->acsBuilder != nullptr) {
		// Issue acs mutation at the end of this commit batch
		addAccumulativeChecksumMutations(self);
	}
	const auto versionSet = LogPushVersionSet{ self->prevVersion,
		                                       self->commitVersion,
		                                       pProxyCommitData->committedVersion.get(),
		                                       pProxyCommitData->minKnownCommittedVersion };
	self->loggingComplete =
	    pProxyCommitData->logSystem->push(versionSet, self->toCommit, span.context, self->debugID, tpcvMap);

	float ratio = self->toCommit.getEmptyMessageRatio();
	pProxyCommitData->stats.commitBatchingEmptyMessageRatio.addMeasurement(ratio);

	if (!self->forceRecovery) {
		ASSERT(pProxyCommitData->latestLocalCommitBatchLogging.get() == self->localBatchNumber - 1);
		pProxyCommitData->latestLocalCommitBatchLogging.set(self->localBatchNumber);
	}

	self->computeDuration += g_network->timer_monotonic() - self->computeStart;
	if (self->batchOperations > 0) {
		double estimatedDelay = computeReleaseDelay(self, self->latencyBucket);
		double computePerOperation =
		    std::min(SERVER_KNOBS->MAX_COMPUTE_PER_OPERATION, self->computeDuration / self->batchOperations);

		if (computePerOperation <= pProxyCommitData->commitComputePerOperation[self->latencyBucket]) {
			pProxyCommitData->commitComputePerOperation[self->latencyBucket] = computePerOperation;
		} else {
			pProxyCommitData->commitComputePerOperation[self->latencyBucket] =
			    SERVER_KNOBS->PROXY_COMPUTE_GROWTH_RATE * computePerOperation +
			    ((1.0 - SERVER_KNOBS->PROXY_COMPUTE_GROWTH_RATE) *
			     pProxyCommitData->commitComputePerOperation[self->latencyBucket]);
		}
		pProxyCommitData->stats.maxComputeNS =
		    std::max<int64_t>(pProxyCommitData->stats.maxComputeNS,
		                      1e9 * pProxyCommitData->commitComputePerOperation[self->latencyBucket]);
		pProxyCommitData->stats.minComputeNS =
		    std::min<int64_t>(pProxyCommitData->stats.minComputeNS,
		                      1e9 * pProxyCommitData->commitComputePerOperation[self->latencyBucket]);

		if (estimatedDelay >= SERVER_KNOBS->MAX_COMPUTE_DURATION_LOG_CUTOFF ||
		    self->computeDuration >= SERVER_KNOBS->MAX_COMPUTE_DURATION_LOG_CUTOFF) {
			TraceEvent(SevInfo, "LongComputeDuration", pProxyCommitData->dbgid)
			    .suppressFor(10.0)
			    .detail("EstimatedComputeDuration", estimatedDelay)
			    .detail("ComputeDuration", self->computeDuration)
			    .detail("ComputePerOperation", computePerOperation)
			    .detail("LatencyBucket", self->latencyBucket)
			    .detail("UpdatedComputePerOperationEstimate",
			            pProxyCommitData->commitComputePerOperation[self->latencyBucket])
			    .detail("BatchBytes", self->batchBytes)
			    .detail("BatchOperations", self->batchOperations);
		}
	}

	pProxyCommitData->stats.processingMutationDist->sampleSeconds(g_network->timer_monotonic() - postResolutionQueuing);
}

Future<Void> transactionLogging(CommitBatchContext* self) {
	double tLoggingStart = g_network->timer_monotonic();
	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	Span span("MP:transactionLogging"_loc, self->span.context);

	try {
		auto res = co_await race(self->loggingComplete,
		                         pProxyCommitData->committedVersion.whenAtLeast(self->commitVersion + 1));
		if (res.index() == 0) {
			Version ver = std::get<0>(res);
			if (!SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST) {
				pProxyCommitData->minKnownCommittedVersion = std::max(pProxyCommitData->minKnownCommittedVersion, ver);
			}
		}
	} catch (Error& e) {
		if (e.code() == error_code_broken_promise) {
			throw tlog_failed();
		}
		throw;
	}

	pProxyCommitData->lastCommitLatency = now() - self->commitStartTime;
	pProxyCommitData->lastCommitTime = std::max(pProxyCommitData->lastCommitTime.get(), self->commitStartTime);

	co_await yield(TaskPriority::ProxyCommitYield2);

	if (pProxyCommitData->popRemoteTxs &&
	    self->msg.popTo > (!pProxyCommitData->txsPopVersions.empty() ? pProxyCommitData->txsPopVersions.back().second
	                                                                 : pProxyCommitData->lastTxsPop)) {
		if (pProxyCommitData->txsPopVersions.size() >= SERVER_KNOBS->MAX_TXS_POP_VERSION_HISTORY) {
			TraceEvent(SevWarnAlways, "DiscardingTxsPopHistory").suppressFor(1.0);
			pProxyCommitData->txsPopVersions.pop_front();
		}

		pProxyCommitData->txsPopVersions.emplace_back(self->commitVersion, self->msg.popTo);
	}
	pProxyCommitData->logSystemConsumer->popTxs(self->msg.popTo);
	pProxyCommitData->stats.tlogLoggingDist->sampleSeconds(g_network->timer_monotonic() - tLoggingStart);
}

Future<Void> reply(CommitBatchContext* self) {
	double replyStart = g_network->timer_monotonic();
	ProxyCommitData* const pProxyCommitData = self->pProxyCommitData;
	Span span("MP:reply"_loc, self->span.context);

	const Optional<UID>& debugID = self->debugID;

	if (!SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST) {
		// Version vector/unicast is disabled: Logging completed, so the current version (and all versions prior to
		// the current versions) can be treated as commited.

		// Advance min committed version.
		if (self->prevVersion && self->commitVersion - self->prevVersion < SERVER_KNOBS->MAX_VERSIONS_IN_FLIGHT / 2) {
			//TraceEvent("CPAdvanceMinVersion", self->pProxyCommitData->dbgid).detail("PrvVersion", self->prevVersion).detail("CommitVersion", self->commitVersion).detail("Master", self->pProxyCommitData->master.id().toString()).detail("TxSize", self->trs.size());
			debug_advanceMinCommittedVersion(UID(), self->commitVersion);
		}

		// Acknowledge transaction state store commits.
		acknowledgeTransactionStateStoreCommits(self);
	}

	// TraceEvent("ProxyPushed", pProxyCommitData->dbgid)
	//     .detail("PrevVersion", self->prevVersion)
	//     .detail("Version", self->commitVersion);
	if (debugID.present())
		g_traceBatch.addEvent("CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.AfterLogPush");

	// After logging finishes, we report the commit version to master so that every other proxy can get the most
	// up-to-date live committed version. We also maintain the invariant that master's committed version >=
	// self->committedVersion by reporting commit version first before updating self->committedVersion. Otherwise, a
	// client may get a commit version that the master is not aware of, and next GRV request may get a version less
	// than self->committedVersion.

	CODE_PROBE(pProxyCommitData->committedVersion.get() > self->commitVersion,
	           "later version was reported committed first");

	if (self->commitVersion >= pProxyCommitData->committedVersion.get()) {
		Optional<std::set<Tag>> writtenTags;
		if (SERVER_KNOBS->ENABLE_VERSION_VECTOR) {
			writtenTags = self->writtenTags;
		}
		co_await pProxyCommitData->master.reportLiveCommittedVersion.getReply(
		    ReportRawCommittedVersionRequest(self->commitVersion,
		                                     self->lockedAfter,
		                                     self->metadataVersionAfter,
		                                     pProxyCommitData->minKnownCommittedVersion,
		                                     self->prevVersion,
		                                     writtenTags),
		    TaskPriority::ProxyMasterVersionReply);
	}

	if (debugID.present()) {
		g_traceBatch.addEvent(
		    "CommitDebug", debugID.get().first(), "CommitProxyServer.commitBatch.AfterReportRawCommittedVersion");
	}

	if (SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST) {
		// Version vector/unicast is enabled: Received a reply from the sequencer, so we can treat the
		// current version (and all versions prior to the current version) as committed.

		// Advance min committed version now.
		if (self->prevVersion && self->commitVersion - self->prevVersion < SERVER_KNOBS->MAX_VERSIONS_IN_FLIGHT / 2) {
			//TraceEvent("CPAdvanceMinVersion", self->pProxyCommitData->dbgid).detail("PrvVersion", self->prevVersion).detail("CommitVersion", self->commitVersion).detail("Master", self->pProxyCommitData->master.id().toString()).detail("TxSize", self->trs.size());
			debug_advanceMinCommittedVersion(UID(), self->commitVersion);
		}

		// Acknowledge transaction state store commits.
		acknowledgeTransactionStateStoreCommits(self);
	}

	if (self->commitVersion > pProxyCommitData->committedVersion.get()) {
		pProxyCommitData->locked = self->lockedAfter;
		pProxyCommitData->metadataVersion = self->metadataVersionAfter;
		pProxyCommitData->committedVersion.set(self->commitVersion);
		if (SERVER_KNOBS->ENABLE_VERSION_VECTOR_TLOG_UNICAST) {
			ASSERT(self->loggingComplete.isReady());
			pProxyCommitData->minKnownCommittedVersion =
			    std::max(pProxyCommitData->minKnownCommittedVersion, self->loggingComplete.get());
		}
	}

	if (self->forceRecovery) {
		TraceEvent(SevWarn, "RestartingTxnSubsystem", pProxyCommitData->dbgid).detail("Stage", "ProxyShutdown");
		throw worker_removed();
	}

	// Send replies to clients
	// TODO: should be timer_monotonic(), but gets compared to request time, which uses g_network->timer().
	double endTime = g_network->timer();
	// Reset all to zero, used to track the correct index of each commitTransacitonRef on each resolver

	std::fill(self->nextTr.begin(), self->nextTr.end(), 0);
	std::unordered_map<uint8_t, int16_t> idCountsForKey;
	for (int t = 0; t < self->trs.size(); t++) {
		auto& tr = self->trs[t];
		if (self->committed[t] == ConflictBatchStatus::TransactionCommitted && (!self->locked || tr.isLockAware())) {
			ASSERT_WE_THINK(self->commitVersion != invalidVersion);
			if (self->trs[t].idempotencyId.valid()) {
				idCountsForKey[uint8_t(t >> 8)] += 1;
			}
			tr.reply.send(CommitID(self->commitVersion, t, self->metadataVersionAfter));
		} else if (self->committed[t] == ConflictBatchStatus::TransactionTooOld) {
			tr.reply.sendError(transaction_too_old());
		} else if (self->committed[t] == ConflictBatchStatus::TransactionLockReject) {
			// We already sent the error
			ASSERT(tr.reply.isSet());
		} else {
			// If enable the option to report conflicting keys from resolvers, we send back all keyranges' indices
			// through CommitID
			if (tr.transaction.report_conflicting_keys) {
				Standalone<VectorRef<int>> conflictingKRIndices;
				for (int resolverInd : self->transactionResolverMap[t]) {
					auto const& cKRs =
					    self->resolution[resolverInd]
					        .conflictingKeyRangeMap[self->nextTr[resolverInd]]; // nextTr[resolverInd] -> index of
					                                                            // this trs[t] on the resolver
					for (auto const& rCRIndex : cKRs) {
						// read_conflict_range can change when sent to resolvers, mapping the index from
						// resolver-side to original index in commitTransactionRef
						conflictingKRIndices.push_back(conflictingKRIndices.arena(),
						                               self->txReadConflictRangeIndexMap[t][resolverInd][rCRIndex]);
					}
				}
				// At least one keyRange index should be returned
				ASSERT(!conflictingKRIndices.empty());
				tr.reply.send(CommitID(
				    invalidVersion, t, Optional<Value>(), Optional<Standalone<VectorRef<int>>>(conflictingKRIndices)));
			} else {
				tr.reply.sendError(not_committed());
			}
		}

		// Update corresponding transaction indices on each resolver
		for (int resolverInd : self->transactionResolverMap[t])
			self->nextTr[resolverInd]++;

		// TODO: filter if pipelined with large commit
		const double duration = endTime - tr.requestTime();
		pProxyCommitData->stats.commitLatencySample.addMeasurement(duration);
		if (pProxyCommitData->latencyBandConfig.present()) {
			bool filter = self->maxTransactionBytes >
			              pProxyCommitData->latencyBandConfig.get().commitConfig.maxCommitBytes.orDefault(
			                  std::numeric_limits<int>::max());
			pProxyCommitData->stats.commitLatencyBands.addMeasurement(duration, 1, Filtered(filter));
		}
	}

	for (auto [highOrderBatchIndex, count] : idCountsForKey) {
		pProxyCommitData->expectedIdempotencyIdCountForKey.send(
		    ExpectedIdempotencyIdCountForKey{ self->commitVersion, count, highOrderBatchIndex });
	}

	++pProxyCommitData->stats.commitBatchOut;
	pProxyCommitData->stats.txnCommitOut += self->trs.size();
	pProxyCommitData->stats.txnConflicts += self->trs.size() - self->commitCount;
	pProxyCommitData->stats.txnCommitOutSuccess += self->commitCount;

	if (now() - pProxyCommitData->lastCoalesceTime > SERVER_KNOBS->RESOLVER_COALESCE_TIME) {
		pProxyCommitData->lastCoalesceTime = now();
		int lastSize = pProxyCommitData->keyResolvers.size();
		auto rs = pProxyCommitData->keyResolvers.ranges();
		Version oldestVersion = self->prevVersion - SERVER_KNOBS->MAX_WRITE_TRANSACTION_LIFE_VERSIONS;
		for (auto r = rs.begin(); r != rs.end(); ++r) {
			while (r->value().size() > 1 && r->value()[1].first < oldestVersion)
				r->value().pop_front();
			if (!r->value().empty() && r->value().front().first < oldestVersion)
				r->value().front().first = 0;
		}
		if (SERVER_KNOBS->PROXY_USE_RESOLVER_PRIVATE_MUTATIONS) {
			// Only normal key space, because \xff key space is processed by all resolvers.
			pProxyCommitData->keyResolvers.coalesce(normalKeys);
			auto& versions = pProxyCommitData->systemKeyVersions;
			while (versions.size() > 1 && versions[1] < oldestVersion) {
				versions.pop_front();
			}
			if (!versions.empty() && versions[0] < oldestVersion) {
				versions[0] = 0;
			}
		} else {
			pProxyCommitData->keyResolvers.coalesce(allKeys);
		}
		if (pProxyCommitData->keyResolvers.size() != lastSize)
			TraceEvent("KeyResolverSize", pProxyCommitData->dbgid)
			    .detail("Size", pProxyCommitData->keyResolvers.size());
	}

	// Dynamic batching for commits
	double target_latency =
	    (now() - self->startTime) * SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_INTERVAL_LATENCY_FRACTION;
	pProxyCommitData->commitBatchInterval =
	    std::max(SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_INTERVAL_MIN,
	             std::min(SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_INTERVAL_MAX,
	                      target_latency * SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_INTERVAL_SMOOTHER_ALPHA +
	                          pProxyCommitData->commitBatchInterval *
	                              (1 - SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_INTERVAL_SMOOTHER_ALPHA)));
	pProxyCommitData->stats.commitBatchingWindowSize.addMeasurement(pProxyCommitData->commitBatchInterval);
	pProxyCommitData->commitBatchesMemBytesCount -= self->currentBatchMemBytesCount;
	ASSERT_ABORT(pProxyCommitData->commitBatchesMemBytesCount >= 0);
	co_await self->releaseFuture;
	pProxyCommitData->stats.replyCommitDist->sampleSeconds(g_network->timer_monotonic() - replyStart);
}

// Commit one batch of transactions trs
Future<Void> commitBatchImpl(CommitBatchContext* pContext) {
	// WARNING: this code is run at a high priority (until the first delay(0)), so it needs to do as little work as
	// possible

	pContext->stage = INITIALIZE;
	getCurrentLineage()->modify(&TransactionLineage::operation) = TransactionLineage::Operation::Commit;

	// Active load balancing runs at a very high priority (to obtain accurate estimate of memory used by commit batches)
	// so we need to downgrade here
	co_await delay(0, TaskPriority::ProxyCommit);

	pContext->pProxyCommitData->lastVersionTime = pContext->startTime;
	++pContext->pProxyCommitData->stats.commitBatchIn;
	pContext->setupTraceBatch();

	/////// Phase 1: Pre-resolution processing (CPU bound except waiting for a version # which is separately pipelined
	/// and *should* be available by now (unless empty commit); ordered; currently atomic but could yield)
	pContext->stage = PRE_RESOLUTION;
	co_await CommitBatch::preresolutionProcessing(pContext);
	if (pContext->rejected) {
		pContext->pProxyCommitData->commitBatchesMemBytesCount -= pContext->currentBatchMemBytesCount;
		co_return;
	}

	/////// Phase 2: Resolution (waiting on the network; pipelined)
	pContext->stage = RESOLUTION;
	co_await CommitBatch::getResolution(pContext);

	////// Phase 3: Post-resolution processing (CPU bound except for very rare situations; ordered; currently atomic but
	/// doesn't need to be)
	pContext->stage = POST_RESOLUTION;
	co_await CommitBatch::postResolution(pContext);

	/////// Phase 4: Logging (network bound; pipelined up to MAX_READ_TRANSACTION_LIFE_VERSIONS (limited by loop above))
	pContext->stage = TRANSACTION_LOGGING;
	co_await CommitBatch::transactionLogging(pContext);

	/////// Phase 5: Replies (CPU bound; no particular order required, though ordered execution would be best for
	/// latency)
	pContext->stage = REPLY;
	co_await CommitBatch::reply(pContext);

	pContext->stage = COMPLETE;
}

} // namespace CommitBatch

Future<Void> commitBatch(ProxyCommitData* pCommitData,
                         std::vector<CommitTransactionRequest>* trs,
                         int currentBatchMemBytesCount) {

	CommitBatch::CommitBatchContext context(pCommitData, trs, currentBatchMemBytesCount);
	Future<Void> commit = CommitBatch::commitBatchImpl(&context);

	try {
		co_await timeoutError(commit, SERVER_KNOBS->COMMIT_PROXY_LIVENESS_TIMEOUT);
	} catch (Error& err) {
		if (err.code() == error_code_actor_cancelled) {
			throw;
		}
		TraceEvent(SevInfo, "CommitBatchFailed", pCommitData->dbgid)
		    .detail("Stage", context.stage)
		    .detail("ErrorCode", err.code());
		throw failed_to_progress();
	}
}

// Add tss mapping data to the reply, if any of the included storage servers have a TSS pair
void maybeAddTssMapping(GetKeyServerLocationsReply& reply,
                        ProxyCommitData* commitData,
                        std::unordered_set<UID>& included,
                        UID ssId) {
	if (!included.contains(ssId)) {
		auto mappingItr = commitData->tssMapping.find(ssId);
		if (mappingItr != commitData->tssMapping.end()) {
			reply.resultsTssMapping.push_back(*mappingItr);
		}
		included.insert(ssId);
	}
}

void addTagMapping(GetKeyServerLocationsReply& reply, ProxyCommitData* commitData) {
	for (const auto& [_, shard] : reply.results) {
		for (auto& ssi : shard) {
			auto iter = commitData->storageCache.find(ssi.id());
			ASSERT_WE_THINK(iter != commitData->storageCache.end());
			reply.resultsTagMapping.emplace_back(ssi.id(), iter->second->tag);
		}
	}
}

static Future<Void> doKeyServerLocationRequest(GetKeyServerLocationsRequest req, ProxyCommitData* commitData) {
	// We can't respond to these requests until we have valid txnStateStore
	getCurrentLineage()->modify(&TransactionLineage::operation) = TransactionLineage::Operation::GetKeyServersLocations;
	getCurrentLineage()->modify(&TransactionLineage::txID) = req.spanContext.traceID;

	co_await commitData->validState.getFuture();

	co_await delay(0, TaskPriority::DefaultEndpoint);

	std::unordered_set<UID> tssMappingsIncluded;
	GetKeyServerLocationsReply rep;

	if (!req.end.present()) {
		auto r = req.reverse ? commitData->keyInfo.rangeContainingKeyBefore(req.begin)
		                     : commitData->keyInfo.rangeContaining(req.begin);
		std::vector<StorageServerInterface> ssis;
		ssis.reserve(r.value().src_info.size());
		for (auto& it : r.value().src_info) {
			ssis.push_back(it->interf);
			maybeAddTssMapping(rep, commitData, tssMappingsIncluded, it->interf.id());
		}
		rep.results.emplace_back(r.range(), ssis);
	} else if (!req.reverse) {
		int count = 0;
		for (auto r = commitData->keyInfo.rangeContaining(req.begin);
		     r != commitData->keyInfo.ranges().end() && count < req.limit && r.begin() < req.end.get();
		     ++r) {
			std::vector<StorageServerInterface> ssis;
			ssis.reserve(r.value().src_info.size());
			for (auto& it : r.value().src_info) {
				ssis.push_back(it->interf);
				maybeAddTssMapping(rep, commitData, tssMappingsIncluded, it->interf.id());
			}
			rep.results.emplace_back(r.range(), ssis);
			count++;
		}
	} else {
		int count = 0;
		auto r = commitData->keyInfo.rangeContainingKeyBefore(req.end.get());
		while (count < req.limit && req.begin < r.end()) {
			std::vector<StorageServerInterface> ssis;
			ssis.reserve(r.value().src_info.size());
			for (auto& it : r.value().src_info) {
				ssis.push_back(it->interf);
				maybeAddTssMapping(rep, commitData, tssMappingsIncluded, it->interf.id());
			}
			rep.results.emplace_back(r.range(), ssis);
			if (r == commitData->keyInfo.ranges().begin()) {
				break;
			}
			count++;
			--r;
		}
	}
	addTagMapping(rep, commitData);
	req.reply.send(rep);
	++commitData->stats.keyServerLocationOut;
}

static Future<Void> readRequestServer(CommitProxyInterface proxy,
                                      PromiseStream<Future<Void>> addActor,
                                      ProxyCommitData* commitData) {
	while (true) {
		GetKeyServerLocationsRequest req = co_await proxy.getKeyServersLocations.getFuture();
		// WARNING: this code is run at a high priority, so it needs to do as little work as possible
		if (req.limit != CLIENT_KNOBS->STORAGE_METRICS_SHARD_LIMIT && // Always do data distribution requests
		    (commitData->stats.keyServerLocationIn.getValue() - commitData->stats.keyServerLocationOut.getValue() >
		         SERVER_KNOBS->KEY_LOCATION_MAX_QUEUE_SIZE ||
		     (g_network->isSimulated() && buggify(0.001)))) {
			++commitData->stats.keyServerLocationErrors;
			req.reply.sendError(commit_proxy_memory_limit_exceeded());
			TraceEvent(SevWarnAlways, "ProxyLocationRequestThresholdExceeded").suppressFor(60);
		} else {
			++commitData->stats.keyServerLocationIn;
			addActor.send(doKeyServerLocationRequest(req, commitData));
		}
	}
}

static Future<Void> rejoinServer(CommitProxyInterface proxy, ProxyCommitData* commitData) {
	// We can't respond to these requests until we have valid txnStateStore
	co_await commitData->validState.getFuture();

	TraceEvent("ProxyReadyForReads", proxy.id()).log();

	while (true) {
		GetStorageServerRejoinInfoRequest req = co_await proxy.getStorageServerRejoinInfo.getFuture();
		if (commitData->txnStateStore->readValue(serverListKeyFor(req.id)).get().present()) {
			GetStorageServerRejoinInfoReply rep;
			rep.version = commitData->version.get();
			rep.tag = decodeServerTagValue(commitData->txnStateStore->readValue(serverTagKeyFor(req.id)).get().get());
			RangeResult history = commitData->txnStateStore->readRange(serverTagHistoryRangeFor(req.id)).get();
			for (int i = history.size() - 1; i >= 0; i--) {
				rep.history.push_back(
				    std::make_pair(decodeServerTagHistoryKey(history[i].key), decodeServerTagValue(history[i].value)));
			}
			auto localityKey = commitData->txnStateStore->readValue(tagLocalityListKeyFor(req.dcId)).get();
			rep.newLocality = false;
			if (localityKey.present()) {
				int8_t locality = decodeTagLocalityListValue(localityKey.get());
				if (locality != rep.tag.locality) {
					TraceEvent(SevWarnAlways, "SSRejoinedWithChangedLocality")
					    .detail("Tag", rep.tag.toString())
					    .detail("DcId", req.dcId)
					    .detail("NewLocality", locality);
				} else if (locality != rep.tag.locality) {
					uint16_t tagId = 0;
					std::vector<uint16_t> usedTags;
					auto tagKeys = commitData->txnStateStore->readRange(serverTagKeys).get();
					for (auto& kv : tagKeys) {
						Tag t = decodeServerTagValue(kv.value);
						if (t.locality == locality) {
							usedTags.push_back(t.id);
						}
					}
					auto historyKeys = commitData->txnStateStore->readRange(serverTagHistoryKeys).get();
					for (auto& kv : historyKeys) {
						Tag t = decodeServerTagValue(kv.value);
						if (t.locality == locality) {
							usedTags.push_back(t.id);
						}
					}
					std::sort(usedTags.begin(), usedTags.end());

					int usedIdx = 0;
					for (; !usedTags.empty() && tagId <= usedTags.end()[-1]; tagId++) {
						if (tagId < usedTags[usedIdx]) {
							break;
						} else {
							usedIdx++;
						}
					}
					rep.newTag = Tag(locality, tagId);
				}
			} else {
				ASSERT_WE_THINK(rep.tag.locality != tagLocalityUpgraded);
				TraceEvent(SevWarnAlways, "SSRejoinedWithUnknownLocality")
				    .detail("Tag", rep.tag.toString())
				    .detail("DcId", req.dcId);
			}
			req.reply.send(rep);
		} else {
			req.reply.sendError(worker_removed());
		}
	}
}

Future<Void> ddMetricsRequestServer(CommitProxyInterface proxy, Reference<AsyncVar<ServerDBInfo> const> db) {
	while (true) {
		GetDDMetricsRequest req = co_await proxy.getDDMetrics.getFuture();
		if (!db->get().distributor.present()) {
			req.reply.sendError(dd_not_found());
			continue;
		}
		ErrorOr<GetDataDistributorMetricsReply> reply =
		    co_await errorOr(db->get().distributor.get().dataDistributorMetrics.getReply(
		        GetDataDistributorMetricsRequest(req.keys, req.shardLimit)));
		if (reply.isError()) {
			req.reply.sendError(reply.getError());
		} else {
			GetDDMetricsReply newReply;
			newReply.storageMetricsList = reply.get().storageMetricsList;
			req.reply.send(newReply);
		}
	}
}

Future<Void> monitorRemoteCommitted(ProxyCommitData* self) {
	while (true) {
		co_await delay(0); // allow this actor to be cancelled if we are removed after db changes.
		Optional<std::vector<OptionalInterface<TLogInterface>>> remoteLogs;
		if (self->db->get().recoveryState >= RecoveryState::ALL_LOGS_RECRUITED) {
			for (auto& logSet : self->db->get().logSystemConfig.tLogs) {
				if (!logSet.isLocal) {
					remoteLogs = logSet.tLogs;
					for (auto& tLog : logSet.tLogs) {
						if (!tLog.present()) {
							remoteLogs = Optional<std::vector<OptionalInterface<TLogInterface>>>();
							break;
						}
					}
					break;
				}
			}
		}

		if (!remoteLogs.present()) {
			co_await self->db->onChange();
			continue;
		}
		self->popRemoteTxs = true;

		Future<Void> onChange = self->db->onChange();
		while (true) {
			std::vector<Future<TLogQueuingMetricsReply>> replies;
			for (auto& it : remoteLogs.get()) {
				replies.push_back(
				    brokenPromiseToNever(it.interf().getQueuingMetrics.getReply(TLogQueuingMetricsRequest())));
			}
			co_await (waitForAll(replies) || onChange);

			if (onChange.isReady()) {
				break;
			}

			// FIXME: use the configuration to calculate a more precise minimum recovery version.
			Version minVersion = std::numeric_limits<Version>::max();
			for (auto& it : replies) {
				minVersion = std::min(minVersion, it.get().v);
			}

			while (!self->txsPopVersions.empty() && self->txsPopVersions.front().first <= minVersion) {
				self->lastTxsPop = self->txsPopVersions.front().second;
				self->logSystemConsumer->popTxs(self->txsPopVersions.front().second, tagLocalityRemoteLog);
				self->txsPopVersions.pop_front();
			}

			co_await (delay(SERVER_KNOBS->UPDATE_REMOTE_LOG_VERSION_INTERVAL) || onChange);
			if (onChange.isReady()) {
				break;
			}
		}
	}
}

Future<Void> proxySnapCreate(ProxySnapRequest snapReq, ProxyCommitData* commitData) {
	TraceEvent("SnapCommitProxy_SnapReqEnter")
	    .detail("SnapPayload", snapReq.snapPayload)
	    .detail("SnapUID", snapReq.snapUID);
	try {
		// whitelist check
		ExecCmdValueString execArg(snapReq.snapPayload);
		StringRef binPath = execArg.getBinaryPath();
		if (!isWhitelisted(commitData->whitelistedBinPathVec, binPath)) {
			TraceEvent("SnapCommitProxy_WhiteListCheckFailed")
			    .detail("SnapPayload", snapReq.snapPayload)
			    .detail("SnapUID", snapReq.snapUID);
			throw snap_path_not_whitelisted();
		}
		// db fully recovered check
		if (commitData->db->get().recoveryState != RecoveryState::FULLY_RECOVERED) {
			// Cluster is not fully recovered and needs TLogs
			// from previous generation for full recovery.
			// Currently, snapshot of old tlog generation is not
			// supported and hence failing the snapshot request until
			// cluster is fully_recovered.
			TraceEvent("SnapCommitProxy_ClusterNotFullyRecovered")
			    .detail("SnapPayload", snapReq.snapPayload)
			    .detail("SnapUID", snapReq.snapUID);
			throw snap_not_fully_recovered_unsupported();
		}

		auto result = commitData->txnStateStore->readValue("log_anti_quorum"_sr.withPrefix(configKeysPrefix)).get();
		int logAntiQuorum = 0;
		if (result.present()) {
			logAntiQuorum = atoi(result.get().toString().c_str());
		}
		// FIXME: logAntiQuorum not supported, remove it later,
		// In version2, we probably don't need this limitation, but this needs to be tested.
		if (logAntiQuorum > 0) {
			TraceEvent("SnapCommitProxy_LogAntiQuorumNotSupported")
			    .detail("SnapPayload", snapReq.snapPayload)
			    .detail("SnapUID", snapReq.snapUID);
			throw snap_log_anti_quorum_unsupported();
		}

		int snapReqRetry = 0;
		double snapRetryBackoff = FLOW_KNOBS->PREVENT_FAST_SPIN_DELAY;
		while (true) {
			// send a snap request to DD
			if (!commitData->db->get().distributor.present()) {
				TraceEvent(SevWarnAlways, "DataDistributorNotPresent").detail("Operation", "SnapRequest");
				throw dd_not_found();
			}
			Error err;
			try {
				Future<ErrorOr<Void>> ddSnapReq =
				    commitData->db->get().distributor.get().distributorSnapReq.tryGetReply(
				        DistributorSnapRequest(snapReq.snapPayload, snapReq.snapUID));
				co_await throwErrorOr(ddSnapReq);
				break;
			} catch (Error& e) {
				err = e;
			}
			TraceEvent("SnapCommitProxy_DDSnapResponseError")
			    .errorUnsuppressed(err)
			    .detail("SnapPayload", snapReq.snapPayload)
			    .detail("SnapUID", snapReq.snapUID)
			    .detail("Retry", snapReqRetry);
			// Retry if we have network issues
			if (err.code() != error_code_request_maybe_delivered ||
			    ++snapReqRetry > SERVER_KNOBS->SNAP_NETWORK_FAILURE_RETRY_LIMIT)
				throw err;
			co_await delay(snapRetryBackoff);
			snapRetryBackoff = snapRetryBackoff * 2; // exponential backoff
		}
		snapReq.reply.send(Void());
	} catch (Error& e) {
		TraceEvent("SnapCommitProxy_SnapReqError")
		    .errorUnsuppressed(e)
		    .detail("SnapPayload", snapReq.snapPayload)
		    .detail("SnapUID", snapReq.snapUID);
		if (e.code() != error_code_operation_cancelled) {
			snapReq.reply.sendError(e);
		} else {
			throw e;
		}
	}
	TraceEvent("SnapCommitProxy_SnapReqExit")
	    .detail("SnapPayload", snapReq.snapPayload)
	    .detail("SnapUID", snapReq.snapUID);
}

Future<Void> proxyCheckSafeExclusion(Reference<AsyncVar<ServerDBInfo> const> db, ExclusionSafetyCheckRequest req) {
	TraceEvent("SafetyCheckCommitProxyBegin").log();
	ExclusionSafetyCheckReply reply(false);
	if (!db->get().distributor.present()) {
		TraceEvent(SevWarnAlways, "DataDistributorNotPresent").detail("Operation", "ExclusionSafetyCheck");
		req.reply.send(reply);
		co_return;
	}
	try {
		Future<ErrorOr<DistributorExclusionSafetyCheckReply>> ddSafeFuture =
		    db->get().distributor.get().distributorExclCheckReq.tryGetReply(
		        DistributorExclusionSafetyCheckRequest(req.exclusions));
		DistributorExclusionSafetyCheckReply ddReply = co_await throwErrorOr(ddSafeFuture);
		reply.safe = ddReply.safe;
	} catch (Error& e) {
		TraceEvent("SafetyCheckCommitProxyResponseError").error(e);
		if (e.code() != error_code_operation_cancelled) {
			req.reply.sendError(e);
			co_return;
		} else {
			throw e;
		}
	}
	TraceEvent("SafetyCheckCommitProxyFinish").log();
	req.reply.send(reply);
}

class TxnTagCommitCostReporter {
	UID myID;
	Reference<AsyncVar<ServerDBInfo> const> db;
	UIDTransactionTagMap<TransactionCommitCostEstimation>* ssTrTagCommitCost;

	Future<Void> traceRatekeeperChanges() {
		while (true) {
			co_await db->onChange();
			if (db->get().ratekeeper.present()) {
				TraceEvent("ProxyRatekeeperChanged", myID).detail("RKID", db->get().ratekeeper.get().id());
			} else {
				TraceEvent("ProxyRatekeeperDied", myID).log();
			}
		}
	}

	Future<Void> reportCommitCosts() {
		bool sendImmediately = db->get().ratekeeper.present();
		while (true) {
			if (!sendImmediately) {
				co_await db->onChange();
			}
			sendImmediately = false;

			if (!db->get().ratekeeper.present()) {
				continue;
			}

			auto reply = brokenPromiseToNever(db->get().ratekeeper.get().reportCommitCostEstimation.getReply(
			    ReportCommitCostEstimationRequest(std::move(*ssTrTagCommitCost))));
			ssTrTagCommitCost->clear();

			auto dbChanged = db->onChange();
			auto replyOrChange = co_await race(reply, dbChanged);
			if (replyOrChange.index() == 1) {
				sendImmediately = true;
				continue;
			}

			co_await race(delay(SERVER_KNOBS->REPORT_TRANSACTION_COST_ESTIMATION_DELAY), dbChanged);
			sendImmediately = true;
		}
	}

public:
	TxnTagCommitCostReporter(UID myID,
	                         Reference<AsyncVar<ServerDBInfo> const> db,
	                         UIDTransactionTagMap<TransactionCommitCostEstimation>* ssTrTagCommitCost)
	  : myID(myID), db(db), ssTrTagCommitCost(ssTrTagCommitCost) {}

	Future<Void> run() { co_await race(traceRatekeeperChanges(), reportCommitCosts()); }
};

namespace {
struct ExpireServerEntry {
	int64_t timeReceived;
	int expectedCount = 0;
	int receivedCount = 0;
	bool initialized = false;
};

struct IdempotencyKey {
	Version version;
	uint8_t highOrderBatchIndex;
	bool operator==(const IdempotencyKey& other) const {
		return version == other.version && highOrderBatchIndex == other.highOrderBatchIndex;
	}
};

} // namespace

namespace std {
template <>
struct hash<IdempotencyKey> {
	std::size_t operator()(const IdempotencyKey& key) const {
		std::size_t seed = 0;
		boost::hash_combine(seed, std::hash<Version>{}(key.version));
		boost::hash_combine(seed, std::hash<uint8_t>{}(key.highOrderBatchIndex));
		return seed;
	}
};

} // namespace std

namespace {

class IdempotencyIdsExpireServer {
	PublicRequestStream<ExpireIdempotencyIdRequest> expireIdempotencyId;
	PromiseStream<ExpectedIdempotencyIdCountForKey> expectedIdempotencyIdCountForKey;
	Standalone<VectorRef<MutationRef>>* idempotencyClears;
	std::unordered_map<IdempotencyKey, ExpireServerEntry> idStatus;

	void updateStatus(IdempotencyKey key, ExpireServerEntry* status) {
		if (status->initialized) {
			if (status->receivedCount == status->expectedCount) {
				auto keyRange =
				    makeIdempotencySingleKeyRange(idempotencyClears->arena(), key.version, key.highOrderBatchIndex);
				idempotencyClears->push_back(idempotencyClears->arena(),
				                             MutationRef(MutationRef::ClearRange, keyRange.begin, keyRange.end));
				idStatus.erase(key);
			}
		} else {
			status->timeReceived = now();
			status->initialized = true;
		}
	}

	Future<Void> receiveExpireRequests() {
		while (true) {
			ExpireIdempotencyIdRequest req = co_await expireIdempotencyId.getFuture();
			IdempotencyKey key{ req.commitVersion, req.batchIndexHighByte };
			ExpireServerEntry* status = &idStatus[key];
			status->receivedCount += 1;
			CODE_PROBE(status->expectedCount == 0, "ExpireIdempotencyIdRequest received before count is known");
			if (status->expectedCount > 0) {
				ASSERT_LE(status->receivedCount, status->expectedCount);
			}
			updateStatus(key, status);
		}
	}

	Future<Void> receiveExpectedCounts() {
		while (true) {
			ExpectedIdempotencyIdCountForKey req = co_await expectedIdempotencyIdCountForKey.getFuture();
			IdempotencyKey key{ req.commitVersion, req.batchIndexHighByte };
			ExpireServerEntry* status = &idStatus[key];
			ASSERT_EQ(status->expectedCount, 0);
			status->expectedCount = req.idempotencyIdCount;
			updateStatus(key, status);
		}
	}

	Future<Void> purgeOldEntries() {
		while (true) {
			co_await delay(SERVER_KNOBS->IDEMPOTENCY_ID_IN_MEMORY_LIFETIME);
			int64_t purgeBefore = now() - SERVER_KNOBS->IDEMPOTENCY_ID_IN_MEMORY_LIFETIME;
			std::vector<IdempotencyKey> keys;
			keys.reserve(idStatus.size());
			for (const auto& entry : idStatus) {
				keys.push_back(entry.first);
			}

			std::vector<IdempotencyKey> keysToErase;
			for (const auto& key : keys) {
				co_await yield();
				auto status = idStatus.find(key);
				if (status != idStatus.end() && status->second.timeReceived < purgeBefore) {
					keysToErase.push_back(key);
				}
			}
			for (const auto& key : keysToErase) {
				idStatus.erase(key);
			}
		}
	}

public:
	IdempotencyIdsExpireServer(PublicRequestStream<ExpireIdempotencyIdRequest> expireIdempotencyId,
	                           PromiseStream<ExpectedIdempotencyIdCountForKey> expectedIdempotencyIdCountForKey,
	                           Standalone<VectorRef<MutationRef>>* idempotencyClears)
	  : expireIdempotencyId(expireIdempotencyId), expectedIdempotencyIdCountForKey(expectedIdempotencyIdCountForKey),
	    idempotencyClears(idempotencyClears) {}

	Future<Void> run() { co_await race(receiveExpireRequests(), receiveExpectedCounts(), purgeOldEntries()); }
};

TEST_CASE("/fdbserver/commitproxy/IdempotencyIdsExpireServer/ExpireBeforeExpectedCount") {
	constexpr Version version = 100;
	constexpr uint8_t firstBatch = 1;
	constexpr uint8_t secondBatch = 2;
	PublicRequestStream<ExpireIdempotencyIdRequest> expireRequests;
	PromiseStream<ExpectedIdempotencyIdCountForKey> expectedCounts;
	Standalone<VectorRef<MutationRef>> clears;
	IdempotencyIdsExpireServer expireServer(expireRequests, expectedCounts, &clears);
	Future<Void> server = expireServer.run();

	// Expire requests can arrive before their expected counts; neither batch may be cleared yet.
	expireRequests.send(ExpireIdempotencyIdRequest(version, firstBatch));
	expireRequests.send(ExpireIdempotencyIdRequest(version, firstBatch));
	expireRequests.send(ExpireIdempotencyIdRequest(version, secondBatch));
	co_await yield();
	ASSERT(!server.isReady());
	ASSERT(clears.empty());

	// The first batch is complete once counts arrive, while the second is still one request short.
	expectedCounts.send(ExpectedIdempotencyIdCountForKey(version, 2, firstBatch));
	expectedCounts.send(ExpectedIdempotencyIdCountForKey(version, 2, secondBatch));
	co_await yield();
	ASSERT(!server.isReady());
	ASSERT_EQ(clears.size(), 1);

	Arena expectedArena;
	auto firstRange = makeIdempotencySingleKeyRange(expectedArena, version, firstBatch);
	ASSERT_EQ(clears[0].type, MutationRef::ClearRange);
	ASSERT(clears[0].param1 == firstRange.begin);
	ASSERT(clears[0].param2 == firstRange.end);

	// Completing the second batch must emit its distinct clear range.
	expireRequests.send(ExpireIdempotencyIdRequest(version, secondBatch));
	co_await yield();
	ASSERT(!server.isReady());
	ASSERT_EQ(clears.size(), 2);

	auto secondRange = makeIdempotencySingleKeyRange(expectedArena, version, secondBatch);
	ASSERT_EQ(clears[1].type, MutationRef::ClearRange);
	ASSERT(clears[1].param1 == secondRange.begin);
	ASSERT(clears[1].param2 == secondRange.end);
	server.cancel();
}

struct TransactionStateResolveContext {
	// Maximum sequence for txnStateRequest, this is defined when the request last flag is set.
	Sequence maxSequence = std::numeric_limits<Sequence>::max();

	// Flags marks received transaction state requests, we only process the transaction request when *all* requests are
	// received.
	std::unordered_set<Sequence> receivedSequences;

	ProxyCommitData* pCommitData = nullptr;

	// Pointer to transaction state store, shortcut for commitData.txnStateStore
	IKeyValueStore* pTxnStateStore = nullptr;

	Future<Void> txnRecovery;

	// Actor streams
	PromiseStream<Future<Void>>* pActors = nullptr;

	// Flag reports if the transaction state request is complete. This request should only happen during recover, i.e.
	// once per commit proxy.
	bool processed = false;

	TransactionStateResolveContext() = default;

	TransactionStateResolveContext(ProxyCommitData* pCommitData_, PromiseStream<Future<Void>>* pActors_)
	  : pCommitData(pCommitData_), pTxnStateStore(pCommitData_->txnStateStore), pActors(pActors_) {
		ASSERT(pTxnStateStore != nullptr);
	}
};

Future<Void> processCompleteTransactionStateRequest(TransactionStateResolveContext* pContext) {
	KeyRange txnKeys = allKeys;
	std::map<Tag, UID> tag_uid;

	RangeResult UIDtoTagMap = pContext->pTxnStateStore->readRange(serverTagKeys).get();
	for (const KeyValueRef& kv : UIDtoTagMap) {
		tag_uid[decodeServerTagValue(kv.value)] = decodeServerTagKey(kv.key);
	}

	while (true) {
		co_await yield();

		RangeResult data =
		    pContext->pTxnStateStore
		        ->readRange(txnKeys, SERVER_KNOBS->BUGGIFIED_ROW_LIMIT, SERVER_KNOBS->APPLY_MUTATION_BYTES)
		        .get();
		if (data.empty())
			break;

		((KeyRangeRef&)txnKeys) = KeyRangeRef(keyAfter(data.back().key, txnKeys.arena()), txnKeys.end);

		Standalone<VectorRef<MutationRef>> mutations;
		std::vector<std::pair<MapPair<Key, ServerCacheInfo>, int>> keyInfoData;
		std::vector<UID> src, dest;
		ServerCacheInfo info;
		auto updateTagInfo = [pContext = pContext](const std::vector<UID>& uids,
		                                           std::vector<Tag>& tags,
		                                           std::vector<Reference<StorageInfo>>& storageInfoItems) {
			for (const auto& id : uids) {
				auto storageInfo = getStorageInfo(id, &pContext->pCommitData->storageCache, pContext->pTxnStateStore);
				ASSERT(storageInfo->tag != invalidTag);
				tags.push_back(storageInfo->tag);
				storageInfoItems.push_back(storageInfo);
			}
		};
		for (auto& kv : data) {
			if (kv.key.startsWith(keyServersPrefix)) {
				KeyRef k = kv.key.removePrefix(keyServersPrefix);
				if (k == allKeys.end) {
					continue;
				}
				decodeKeyServersValue(tag_uid, kv.value, src, dest);

				info.tags.clear();

				info.src_info.clear();
				updateTagInfo(src, info.tags, info.src_info);

				info.dest_info.clear();
				updateTagInfo(dest, info.tags, info.dest_info);

				uniquify(info.tags);
				keyInfoData.emplace_back(MapPair<Key, ServerCacheInfo>(k, info), 1);
			} else if (kv.key.startsWith(rangeLockPrefix)) {
				if (pContext->pCommitData->rangeLockEnabled()) {
					ASSERT(pContext->pCommitData->rangeLock != nullptr);
					Key keyInsert = kv.key.removePrefix(rangeLockPrefix);
					pContext->pCommitData->rangeLock->initKeyPoint(keyInsert, kv.value);
				}
			} else {
				mutations.emplace_back(mutations.arena(), MutationRef::SetValue, kv.key, kv.value);
				continue;
			}
		}

		// insert keyTag data separately from metadata mutations so that we can do one bulk insert which
		// avoids a lot of map lookups.
		pContext->pCommitData->keyInfo.rawInsert(keyInfoData);

		Arena arena;
		bool confChanges;
		applyMetadataMutations(SpanContext(),
		                       pContext->pCommitData->getApplyMetadataProxyContext(),
		                       arena,
		                       Reference<LogSystemConsumer>(),
		                       mutations,
		                       /* pToCommit= */ nullptr,
		                       confChanges,
		                       /* version= */ 0,
		                       /* popVersion= */ 0,
		                       /* initialCommit= */ true,
		                       /* provisionalCommitProxy */ pContext->pCommitData->provisional);
	}

	auto lockedKey = pContext->pTxnStateStore->readValue(databaseLockedKey).get();
	pContext->pCommitData->locked = lockedKey.present() && !lockedKey.get().empty();
	pContext->pCommitData->metadataVersion = pContext->pTxnStateStore->readValue(metadataVersionKey).get();
	pContext->pCommitData->cdcRouting.reload(pContext->pTxnStateStore);

	pContext->pTxnStateStore->enableSnapshot();
}

Future<Void> processTransactionStateRequestPart(TransactionStateResolveContext* pContext, TxnStateRequest request) {
	ASSERT(pContext->pCommitData != nullptr);
	ASSERT(pContext->pActors != nullptr);

	if (pContext->receivedSequences.contains(request.sequence)) {
		if (pContext->receivedSequences.size() == pContext->maxSequence) {
			co_await pContext->txnRecovery;
		}
		// This part is already received. Still we will re-broadcast it to other CommitProxies
		pContext->pActors->send(broadcastTxnRequest(request, SERVER_KNOBS->TXN_STATE_SEND_AMOUNT, true));
		co_await yield();
		co_return;
	}

	if (request.last) {
		// This is the last piece of subsequence, yet other pieces might still on the way.
		pContext->maxSequence = request.sequence + 1;
	}
	pContext->receivedSequences.insert(request.sequence);

	// Although we may receive the CommitTransactionRequest for the recovery transaction before all of the
	// TxnStateRequest, we will not get a resolution result from any resolver until the master has submitted its initial
	// (sequence 0) resolution request, which it doesn't do until we have acknowledged all TxnStateRequests
	ASSERT(!pContext->pCommitData->validState.isSet());

	for (auto& kv : request.data) {
		pContext->pTxnStateStore->set(kv, &request.arena);
	}
	pContext->pTxnStateStore->commit(true);

	if (pContext->receivedSequences.size() == pContext->maxSequence) {
		// Received all components of the txnStateRequest
		ASSERT(!pContext->processed);
		pContext->txnRecovery = processCompleteTransactionStateRequest(pContext);
		co_await pContext->txnRecovery;
		pContext->processed = true;
	}

	pContext->pActors->send(broadcastTxnRequest(request, SERVER_KNOBS->TXN_STATE_SEND_AMOUNT, true));
	co_await yield();
}

} // anonymous namespace

//
// Metrics related to the commit proxy are logged on a five second interval in
// the `ProxyMetrics` trace. However, it can be hard to determine workload
// burstiness when looking at such a large time range. This function adds much
// more frequent logging for certain metrics to provide fine-grained insight
// into workload patterns. The metrics logged by this function break down into
// two categories:
//
//   * existing counters reported by `ProxyMetrics`
//   * new counters that are only reported by this function
//
// Neither is implemented optimally, but the data collected should be helpful
// in identifying workload patterns on the server.
//
// Metrics reporting by this function can be disabled by setting the
// `BURSTINESS_METRICS_ENABLED` knob to false. The reporting interval can be
// adjusted by modifying the knob `BURSTINESS_METRICS_LOG_INTERVAL`.
//
Future<Void> logDetailedMetrics(ProxyCommitData* commitData) {
	while (true) {
		if (!SERVER_KNOBS->BURSTINESS_METRICS_ENABLED) {
			co_return;
		}

		double startTime = now();
		int64_t commitBatchInBaseline = commitData->stats.commitBatchIn.getValue();
		int64_t commitBatchFlushByteLimitBaseline = commitData->stats.commitBatchFlushByteLimit.getValue();
		int64_t commitBatchFlushCountLimitBaseline = commitData->stats.commitBatchFlushCountLimit.getValue();
		int64_t commitBatchFlushTimeoutBaseline = commitData->stats.commitBatchFlushTimeout.getValue();
		int64_t commitBatchFlushFirstInBatchBaseline = commitData->stats.commitBatchFlushFirstInBatch.getValue();
		int64_t commitBatchFlushTransactionSizeLimitBaseline =
		    commitData->stats.commitBatchFlushTransactionSizeLimit.getValue();
		int64_t txnCommitInBaseline = commitData->stats.txnCommitIn.getValue();
		int64_t mutationsBaseline = commitData->stats.mutations.getValue();
		int64_t mutationBytesBaseline = commitData->stats.mutationBytes.getValue();

		co_await delay(SERVER_KNOBS->BURSTINESS_METRICS_LOG_INTERVAL);

		int64_t commitBatchInReal = commitData->stats.commitBatchIn.getValue();
		int64_t commitBatchFlushByteLimitReal = commitData->stats.commitBatchFlushByteLimit.getValue();
		int64_t commitBatchFlushCountLimitReal = commitData->stats.commitBatchFlushCountLimit.getValue();
		int64_t commitBatchFlushTimeoutReal = commitData->stats.commitBatchFlushTimeout.getValue();
		int64_t commitBatchFlushFirstInBatchReal = commitData->stats.commitBatchFlushFirstInBatch.getValue();
		int64_t commitBatchFlushTransactionSizeLimitReal =
		    commitData->stats.commitBatchFlushTransactionSizeLimit.getValue();
		int64_t txnCommitInReal = commitData->stats.txnCommitIn.getValue();
		int64_t mutationsReal = commitData->stats.mutations.getValue();
		int64_t mutationBytesReal = commitData->stats.mutationBytes.getValue();

		// Don't log anything if any of the counters got reset during the wait
		// interval. Assume that typically all the counters get reset at once.
		if (commitBatchInReal < commitBatchInBaseline ||
		    commitBatchFlushByteLimitReal < commitBatchFlushByteLimitBaseline ||
		    commitBatchFlushCountLimitReal < commitBatchFlushCountLimitBaseline ||
		    commitBatchFlushTimeoutReal < commitBatchFlushTimeoutBaseline ||
		    commitBatchFlushFirstInBatchReal < commitBatchFlushFirstInBatchBaseline ||
		    commitBatchFlushTransactionSizeLimitReal < commitBatchFlushTransactionSizeLimitBaseline ||
		    txnCommitInReal < txnCommitInBaseline || mutationsReal < mutationsBaseline ||
		    mutationBytesReal < mutationBytesBaseline) {
			continue;
		}

		TraceEvent("ProxyDetailedMetrics")
		    .detail("Elapsed", now() - startTime)
		    .detail("CommitBatchIn", commitBatchInReal - commitBatchInBaseline)
		    .detail("CommitBatchFlushByteLimit", commitBatchFlushByteLimitReal - commitBatchFlushByteLimitBaseline)
		    .detail("CommitBatchFlushCountLimit", commitBatchFlushCountLimitReal - commitBatchFlushCountLimitBaseline)
		    .detail("CommitBatchFlushTimeout", commitBatchFlushTimeoutReal - commitBatchFlushTimeoutBaseline)
		    .detail("CommitBatchFlushFirstInBatch",
		            commitBatchFlushFirstInBatchReal - commitBatchFlushFirstInBatchBaseline)
		    .detail("CommitBatchFlushTransactionSizeLimit",
		            commitBatchFlushTransactionSizeLimitReal - commitBatchFlushTransactionSizeLimitBaseline)
		    .detail("TxnCommitIn", txnCommitInReal - txnCommitInBaseline)
		    .detail("Mutations", mutationsReal - mutationsBaseline)
		    .detail("MutationBytes", mutationBytesReal - mutationBytesBaseline)
		    .detail("UniqueClients", commitData->stats.getSizeAndResetUniqueClients());
	}
}

class CommitProxyServerCore {
	CommitProxyInterface proxy;
	LifetimeToken masterLifetime;
	Reference<AsyncVar<ServerDBInfo> const> db;
	bool firstProxy;
	std::string whitelistBinPaths;

	ProxyCommitData commitData;
	PromiseStream<std::pair<std::vector<CommitTransactionRequest>, int>> batchedCommits;
	Future<Void> commitBatcherActor;
	Future<Void> lastCommitComplete = Void();
	TxnTagCommitCostReporter txnTagCommitCostReporter;
	IdempotencyIdsExpireServer idempotencyIdsExpireServer;
	PromiseStream<Future<Void>> addActor;
	Future<Void> onError;
	Future<Void> dbInfoChange;
	TransactionStateResolveContext transactionStateResolveContext;

	Future<Void> serveDbInfoChanges() {
		while (true) {
			co_await dbInfoChange;
			dbInfoChange = commitData.db->onChange();
			if (masterLifetime.isEqual(commitData.db->get().masterLifetime) &&
			    commitData.db->get().recoveryState >= RecoveryState::RECOVERY_TRANSACTION) {
				commitData.logSystem = makeLogSystemFromServerDBInfo(proxy.id(), commitData.db->get(), false, addActor);
				commitData.logSystemConsumer = commitData.logSystem->makeConsumer();
				for (auto it : commitData.tag_popped) {
					commitData.logSystemConsumer->pop(it.second, it.first);
				}
				commitData.logSystemConsumer->popTxs(commitData.lastTxsPop, tagLocalityRemoteLog);
			}

			commitData.updateLatencyBandConfig(commitData.db->get().latencyBandConfig);
		}
	}

	Future<Void> serveActorErrors() { co_await onError; }

	Future<Void> serveBatchedCommits() {
		while (true) {
			std::pair<std::vector<CommitTransactionRequest>, int> batchedRequests = co_await batchedCommits.getFuture();

			// WARNING: this code is run at a high priority, so it needs to do as little work as possible
			/*
			TraceEvent("CommitProxyCTR", proxy.id())
			    .detail("CommitTransactions", trs.size())
			    .detail("TransactionRate", transactionRate)
			    .detail("TransactionQueue", transactionQueue.size())
			    .detail("ReleasedTransactionCount", transactionCount);
			TraceEvent("CommitProxyCore", commitData.dbgid)
			    .detail("TxSize", trs.size())
			    .detail("MasterLifetime", masterLifetime.toString())
			    .detail("DbMasterLifetime", commitData.db->get().masterLifetime.toString())
			    .detail("RecoveryState", commitData.db->get().recoveryState)
			    .detail("CCInf", commitData.db->get().clusterInterface.id().toString());
			*/
			const std::vector<CommitTransactionRequest>& trs = batchedRequests.first;
			const int batchBytes = batchedRequests.second;
			if (!trs.empty() ||
			    (commitData.db->get().recoveryState >= RecoveryState::ACCEPTING_COMMITS &&
			     masterLifetime.isEqual(commitData.db->get().masterLifetime) && lastCommitComplete.isReady())) {

				lastCommitComplete =
				    commitBatch(&commitData,
				                const_cast<std::vector<CommitTransactionRequest>*>(&batchedRequests.first),
				                batchBytes);

				addActor.send(lastCommitComplete);
			}
		}
	}

	Future<Void> serveProxySnapRequests() {
		while (true) {
			ProxySnapRequest snapReq = co_await proxy.proxySnapReq.getFuture();
			TraceEvent(SevDebug, "SnapMasterEnqueue").log();
			addActor.send(proxySnapCreate(snapReq, &commitData));
		}
	}

	Future<Void> serveExclusionSafetyCheckRequests() {
		while (true) {
			ExclusionSafetyCheckRequest exclCheckReq = co_await proxy.exclusionSafetyCheckReq.getFuture();
			addActor.send(proxyCheckSafeExclusion(db, exclCheckReq));
		}
	}

	Future<Void> serveTxnStateRequests() {
		while (true) {
			TxnStateRequest request = co_await proxy.txnState.getFuture();
			addActor.send(processTransactionStateRequestPart(&transactionStateResolveContext, request));
		}
	}

	Future<Void> serveSetThrottledShardRequests() {
		while (true) {
			SetThrottledShardRequest request = co_await proxy.setThrottledShard.getFuture();
			for (auto& shard : request.throttledShards) {
				auto it = commitData.hotShards.begin();
				for (; it != commitData.hotShards.end(); ++it) {
					if (it->first == shard) {
						it->second = request.expirationTime;
						break;
					}
				}
				if (it == commitData.hotShards.end()) {
					commitData.hotShards.emplace_back(std::make_pair(shard, request.expirationTime));
				}
			}
			// TraceEvent(SevDebug, "ReceivedSetThrottledShards").detail("NumHotShards", commitData.hotShards.size());
		}
	}

public:
	CommitProxyServerCore(CommitProxyInterface proxy,
	                      MasterInterface master,
	                      LifetimeToken masterLifetime,
	                      Reference<AsyncVar<ServerDBInfo> const> db,
	                      LogEpoch epoch,
	                      Version recoveryTransactionVersion,
	                      bool firstProxy,
	                      std::string whitelistBinPaths,
	                      bool provisional,
	                      uint16_t commitProxyIndex)
	  : proxy(proxy), masterLifetime(masterLifetime), db(db), firstProxy(firstProxy),
	    whitelistBinPaths(std::move(whitelistBinPaths)), commitData(proxy.id(),
	                                                                master,
	                                                                recoveryTransactionVersion,
	                                                                proxy.commit,
	                                                                db,
	                                                                firstProxy,
	                                                                provisional,
	                                                                commitProxyIndex,
	                                                                epoch),
	    txnTagCommitCostReporter(proxy.id(), db, &commitData.ssTrTagCommitCost),
	    idempotencyIdsExpireServer(proxy.expireIdempotencyId,
	                               commitData.expectedIdempotencyIdCountForKey,
	                               &commitData.idempotencyClears),
	    onError(transformError(actorCollection(addActor.getFuture()), broken_promise(), tlog_failed())) {}

	Future<Void> run() {
		addActor.send(waitFailureServer(proxy.waitFailure.getFuture()));
		addActor.send(traceRole(Role::COMMIT_PROXY, proxy.id()));

		//TraceEvent("CommitProxyInit1", proxy.id());

		// Wait until we can load the "real" logsystem, since we don't support switching them currently
		while (!(masterLifetime.isEqual(commitData.db->get().masterLifetime) &&
		         commitData.db->get().recoveryState >= RecoveryState::RECOVERY_TRANSACTION)) {
			//TraceEvent("ProxyInit2", proxy.id()).detail("LSEpoch", db->get().logSystemConfig.epoch).detail("Need", epoch);
			co_await commitData.db->onChange();
		}
		dbInfoChange = commitData.db->onChange();
		//TraceEvent("ProxyInit3", proxy.id());

		commitData.resolvers = commitData.db->get().resolvers;
		commitData.localTLogCount = commitData.db->get().logSystemConfig.numLogs();
		ASSERT(!commitData.resolvers.empty());
		for (int i = 0; i < commitData.resolvers.size(); ++i) {
			commitData.stats.resolverDist.push_back(
			    Histogram::getHistogram("CommitProxy"_sr,
			                            "ToResolver_" + commitData.resolvers[i].id().toString(),
			                            Histogram::Unit::milliseconds));
		}

		// Initialize keyResolvers map
		auto rs =
		    commitData.keyResolvers.modify(SERVER_KNOBS->PROXY_USE_RESOLVER_PRIVATE_MUTATIONS ? normalKeys : allKeys);
		for (auto r = rs.begin(); r != rs.end(); ++r)
			r->value().emplace_back(0, 0);
		commitData.systemKeyVersions.push_back(0);

		commitData.logSystem = makeLogSystemFromServerDBInfo(proxy.id(), commitData.db->get(), false, addActor);
		commitData.logSystemConsumer = commitData.logSystem->makeConsumer();
		commitData.logAdapter =
		    new LogSystemDiskQueueAdapter(commitData.logSystem, Reference<AsyncVar<PeekTxsInfo>>(), 1, false);
		commitData.txnStateStore = keyValueStoreLogSystem(commitData.logAdapter,
		                                                  commitData.db,
		                                                  proxy.id(),
		                                                  2e9,
		                                                  DisableSnapshot::True,
		                                                  ReplaceContent::True,
		                                                  ExactRecovery::True);
		createWhitelistBinPathVec(whitelistBinPaths, commitData.whitelistedBinPathVec);

		commitData.updateLatencyBandConfig(commitData.db->get().latencyBandConfig);

		// ((SERVER_MEM_LIMIT * COMMIT_BATCHES_MEM_FRACTION_OF_TOTAL) / COMMIT_BATCHES_MEM_TO_TOTAL_MEM_SCALE_FACTOR) is
		// only a approximate formula for limiting the memory used. COMMIT_BATCHES_MEM_TO_TOTAL_MEM_SCALE_FACTOR is an
		// estimate based on experiments and not an accurate one.
		int64_t commitBatchesMemoryLimit = SERVER_KNOBS->COMMIT_BATCHES_MEM_BYTES_HARD_LIMIT;
		if (SERVER_KNOBS->SERVER_MEM_LIMIT > 0) {
			commitBatchesMemoryLimit =
			    std::min(commitBatchesMemoryLimit,
			             static_cast<int64_t>(
			                 (SERVER_KNOBS->SERVER_MEM_LIMIT * SERVER_KNOBS->COMMIT_BATCHES_MEM_FRACTION_OF_TOTAL) /
			                 SERVER_KNOBS->COMMIT_BATCHES_MEM_TO_TOTAL_MEM_SCALE_FACTOR));
		}
		TraceEvent(SevInfo, "CommitBatchesMemoryLimit").detail("BytesLimit", commitBatchesMemoryLimit);

		// Initialize RangeLock
		if (commitData.rangeLockEnabled()) {
			commitData.rangeLock = std::make_shared<RangeLock>(&commitData);
			TraceEvent(SevInfo, "CommitProxyRangeLockEnabled", commitData.dbgid);
		}

		addActor.send(monitorRemoteCommitted(&commitData));
		addActor.send(readRequestServer(proxy, addActor, &commitData));
		addActor.send(rejoinServer(proxy, &commitData));
		addActor.send(ddMetricsRequestServer(proxy, db));
		addActor.send(txnTagCommitCostReporter.run());
		addActor.send(logDetailedMetrics(&commitData));

		auto openDb = openDBOnServer(db);

		if (firstProxy) {
			addActor.send(recurringAsync(
			    [openDb = openDb]() {
				    return cleanIdempotencyIds(openDb, SERVER_KNOBS->IDEMPOTENCY_IDS_MIN_AGE_SECONDS);
			    },
			    SERVER_KNOBS->IDEMPOTENCY_IDS_CLEANER_POLLING_INTERVAL,
			    true,
			    SERVER_KNOBS->IDEMPOTENCY_IDS_CLEANER_POLLING_INTERVAL));
		}
		addActor.send(idempotencyIdsExpireServer.run());

		// wait for txnStateStore recovery
		co_await commitData.txnStateStore->readValue(StringRef());

		int commitBatchByteLimit =
		    (int)std::min<double>(SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_BYTES_MAX,
		                          std::max<double>(SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_BYTES_MIN,
		                                           SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_BYTES_SCALE_BASE *
		                                               pow(commitData.db->get().client.commitProxies.size(),
		                                                   SERVER_KNOBS->COMMIT_TRANSACTION_BATCH_BYTES_SCALE_POWER)));

		commitBatcherActor = commitBatcher(
		    &commitData, batchedCommits, proxy.commit.getFuture(), commitBatchByteLimit, commitBatchesMemoryLimit);

		// This has to be initialized after the commitData.txnStateStore gets initialized.
		transactionStateResolveContext = TransactionStateResolveContext(&commitData, &addActor);

		co_await race(serveDbInfoChanges(),
		              serveActorErrors(),
		              serveBatchedCommits(),
		              serveProxySnapRequests(),
		              serveExclusionSafetyCheckRequests(),
		              serveTxnStateRequests(),
		              serveSetThrottledShardRequests());
	}
};

// only update the local Db info if the CP is not removed
Future<Void> updateLocalDbInfo(Reference<AsyncVar<ServerDBInfo> const> in,
                               Reference<AsyncVar<ServerDBInfo>> out,
                               uint64_t recoveryCount,
                               CommitProxyInterface myInterface) {
	// whether this CP already receive the db info including itself
	bool firstValidDbInfo = false;

	while (true) {
		bool isIncluded =
		    std::count(in->get().client.commitProxies.begin(), in->get().client.commitProxies.end(), myInterface);
		if (in->get().recoveryCount >= recoveryCount && !isIncluded) {
			throw worker_removed();
		}

		if (isIncluded) {
			firstValidDbInfo = true;
		}

		// only update the db info if this is the current CP, or before we received first one including current CP.
		// Several db infos at the beginning just contain the provisional CP
		if (isIncluded || !firstValidDbInfo) {
			DisabledTraceEvent("UpdateLocalDbInfo", myInterface.id())
			    .detail("Provisional", myInterface.provisional)
			    .detail("Included", isIncluded)
			    .detail("FirstValid", firstValidDbInfo)
			    .detail("ReceivedRC", in->get().recoveryCount)
			    .detail("RecoveryCount", recoveryCount);
			if (in->get().recoveryCount >= out->get().recoveryCount) {
				out->set(in->get());
			}
		}

		co_await in->onChange();
	}
}

Future<Void> commitProxyServer(CommitProxyInterface proxy,
                               InitializeCommitProxyRequest req,
                               Reference<AsyncVar<ServerDBInfo> const> db,
                               std::string whitelistBinPaths) {
	try {
		auto localDb = makeReference<AsyncVar<ServerDBInfo>>();
		CommitProxyServerCore core(proxy,
		                           req.master,
		                           req.masterLifetime,
		                           localDb,
		                           req.recoveryCount,
		                           req.recoveryTransactionVersion,
		                           req.firstProxy,
		                           std::move(whitelistBinPaths),
		                           proxy.provisional,
		                           req.commitProxyIndex);
		co_await race(core.run(), updateLocalDbInfo(db, localDb, req.recoveryCount, proxy));
	} catch (Error& e) {
		Severity sev = e.code() == error_code_failed_to_progress ? SevWarnAlways : SevInfo;
		TraceEvent(sev, "CommitProxyTerminated", proxy.id()).errorUnsuppressed(e);

		if (e.code() != error_code_worker_removed && e.code() != error_code_tlog_stopped &&
		    e.code() != error_code_tlog_failed && e.code() != error_code_coordinators_changed &&
		    e.code() != error_code_coordinated_state_conflict && e.code() != error_code_new_coordinators_timed_out &&
		    e.code() != error_code_failed_to_progress) {
			throw;
		}
		CODE_PROBE(e.code() == error_code_failed_to_progress, "Commit proxy failed to progress");
	}
}

void forceLinkCommitProxyTests() {}