void ReplicationCoordinatorImpl::_onDryRunComplete(long long originalTerm) {
invariant(_voteRequester);
LoseElectionDryRunGuardV1 lossGuard(this);
LockGuard lk(_topoMutex);
if (_topCoord->getTerm() != originalTerm) {
log() << "not running for primary, we have been superceded already";
return;
}
const VoteRequester::Result endResult = _voteRequester->getResult();
if (endResult == VoteRequester::Result::kInsufficientVotes) {
log() << "not running for primary, we received insufficient votes";
return;
} else if (endResult == VoteRequester::Result::kStaleTerm) {
log() << "not running for primary, we have been superceded already";
return;
} else if (endResult != VoteRequester::Result::kSuccessfullyElected) {
log() << "not running for primary, we received an unexpected problem";
return;
}
log() << "dry election run succeeded, running for election";
// Stepdown is impossible from this term update.
TopologyCoordinator::UpdateTermResult updateTermResult;
_updateTerm_incallback(originalTerm + 1, &updateTermResult);
invariant(updateTermResult == TopologyCoordinator::UpdateTermResult::kUpdatedTerm);
// Secure our vote for ourself first
_topCoord->voteForMyselfV1();
// Store the vote in persistent storage.
LastVote lastVote;
lastVote.setTerm(originalTerm + 1);
lastVote.setCandidateIndex(_selfIndex);
auto cbStatus = _replExecutor.scheduleDBWork(
[this, lastVote](const ReplicationExecutor::CallbackArgs& cbData) {
_writeLastVoteForMyElection(lastVote, cbData);
});
if (cbStatus.getStatus() == ErrorCodes::ShutdownInProgress) {
return;
}
fassert(34421, cbStatus.getStatus());
lossGuard.dismiss();
}
void ReplicationCoordinatorImpl::_handleHeartbeatResponse(
const ReplicationExecutor::RemoteCommandCallbackData& cbData, int targetIndex) {
// remove handle from queued heartbeats
_untrackHeartbeatHandle(cbData.myHandle);
// Parse and validate the response. At the end of this step, if responseStatus is OK then
// hbResponse is valid.
Status responseStatus = cbData.response.getStatus();
if (responseStatus == ErrorCodes::CallbackCanceled) {
return;
}
const HostAndPort& target = cbData.request.target;
ReplSetHeartbeatResponse hbResponse;
BSONObj resp;
if (responseStatus.isOK()) {
resp = cbData.response.getValue().data;
responseStatus = hbResponse.initialize(resp, _topCoord->getTerm());
}
const Date_t now = _replExecutor.now();
const OpTime lastApplied = getMyLastOptime(); // Locks and unlocks _mutex.
Milliseconds networkTime(0);
StatusWith<ReplSetHeartbeatResponse> hbStatusResponse(hbResponse);
if (responseStatus.isOK()) {
networkTime = cbData.response.getValue().elapsedMillis;
_updateTerm_incallback(hbStatusResponse.getValue().getTerm(), nullptr);
}
else {
log() << "Error in heartbeat request to " << target << "; " << responseStatus;
if (!resp.isEmpty()) {
LOG(3) << "heartbeat response: " << resp;
}
hbStatusResponse = StatusWith<ReplSetHeartbeatResponse>(responseStatus);
}
HeartbeatResponseAction action =
_topCoord->processHeartbeatResponse(
now,
networkTime,
target,
hbStatusResponse,
lastApplied);
if (action.getAction() == HeartbeatResponseAction::NoAction &&
hbStatusResponse.isOK() &&
hbStatusResponse.getValue().hasOpTime() &&
targetIndex >= 0 &&
hbStatusResponse.getValue().hasState() &&
hbStatusResponse.getValue().getState() != MemberState::RS_PRIMARY) {
boost::unique_lock<boost::mutex> lk(_mutex);
if (hbStatusResponse.getValue().getConfigVersion() == _rsConfig.getConfigVersion()) {
_updateOpTimeFromHeartbeat_inlock(targetIndex,
hbStatusResponse.getValue().getOpTime());
// TODO: Enable with Data Replicator
//lk.unlock();
//_dr.slavesHaveProgressed();
}
}
_signalStepDownWaiters();
_scheduleHeartbeatToTarget(
target,
targetIndex,
std::max(now, action.getNextHeartbeatStartDate()));
_handleHeartbeatResponseAction(action, hbStatusResponse);
}
