artdaq  v2_03_03
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Pages
RoutingMasterCore.cc
1 #include <sys/un.h>
2 #include <sys/time.h>
3 #include <sys/epoll.h>
4 #include <arpa/inet.h>
5 #include <netdb.h>
6 #include <pthread.h>
7 #include <sched.h>
8 #include <algorithm>
9 
10 #include "canvas/Utilities/Exception.h"
11 #include "cetlib/exception.h"
12 
13 #include "artdaq-core/Data/Fragment.hh"
14 #include "artdaq-core/Utilities/ExceptionHandler.hh"
15 
16 #include "artdaq/Application/RoutingMasterCore.hh"
17 #include "artdaq/DAQdata/Globals.hh"
18 #include "artdaq/Application/Routing/makeRoutingMasterPolicy.hh"
21 
22 #define TRACE_NAME "RoutingMasterCore"
23 
24 const std::string artdaq::RoutingMasterCore::
25 TABLE_UPDATES_STAT_KEY("RoutingMasterCoreTableUpdates");
26 const std::string artdaq::RoutingMasterCore::
27 TOKENS_RECEIVED_STAT_KEY("RoutingMasterCoreTokensReceived");
28 
29 artdaq::RoutingMasterCore::RoutingMasterCore(int rank, std::string name) :
30  name_(name)
31  , received_token_counter_()
32  , shutdown_requested_(false)
33  , stop_requested_(false)
34  , pause_requested_(false)
35  , token_socket_(-1)
36  , table_socket_(-1)
37  , ack_socket_(-1)
38 {
39  TLOG_DEBUG(name_) << "Constructor" << TLOG_ENDL;
42  metricMan = &metricMan_;
43  my_rank = rank;
44 }
45 
47 {
48  TLOG_DEBUG(name_) << "Destructor" << TLOG_ENDL;
49  if (ev_token_receive_thread_.joinable()) ev_token_receive_thread_.join();
50 }
51 
52 bool artdaq::RoutingMasterCore::initialize(fhicl::ParameterSet const& pset, uint64_t, uint64_t)
53 {
54  TLOG_DEBUG(name_) << "initialize method called with "
55  << "ParameterSet = \"" << pset.to_string()
56  << "\"." << TLOG_ENDL;
57 
58  // pull out the relevant parts of the ParameterSet
59  fhicl::ParameterSet daq_pset;
60  try
61  {
62  daq_pset = pset.get<fhicl::ParameterSet>("daq");
63  }
64  catch (...)
65  {
66  TLOG_ERROR(name_)
67  << "Unable to find the DAQ parameters in the initialization "
68  << "ParameterSet: \"" + pset.to_string() + "\"." << TLOG_ENDL;
69  return false;
70  }
71  try
72  {
73  policy_pset_ = daq_pset.get<fhicl::ParameterSet>("policy");
74  }
75  catch (...)
76  {
77  TLOG_ERROR(name_)
78  << "Unable to find the policy parameters in the DAQ initialization ParameterSet: \"" + daq_pset.to_string() + "\"." << TLOG_ENDL;
79  return false;
80  }
81 
82  // pull out the Metric part of the ParameterSet
83  fhicl::ParameterSet metric_pset;
84  try
85  {
86  metric_pset = daq_pset.get<fhicl::ParameterSet>("metrics");
87  }
88  catch (...) {} // OK if there's no metrics table defined in the FHiCL
89 
90  if (metric_pset.is_empty())
91  {
92  TLOG_INFO(name_) << "No metric plugins appear to be defined" << TLOG_ENDL;
93  }
94  try
95  {
96  metricMan_.initialize(metric_pset, name_);
97  }
98  catch (...)
99  {
100  ExceptionHandler(ExceptionHandlerRethrow::no,
101  "Error loading metrics in RoutingMasterCore::initialize()");
102  }
103 
104  // create the requested CommandableFragmentGenerator
105  auto policy_plugin_spec = policy_pset_.get<std::string>("policy", "");
106  if (policy_plugin_spec.length() == 0)
107  {
108  TLOG_ERROR(name_)
109  << "No fragment generator (parameter name = \"policy\") was "
110  << "specified in the policy ParameterSet. The "
111  << "DAQ initialization PSet was \"" << daq_pset.to_string() << "\"." << TLOG_ENDL;
112  return false;
113  }
114  try
115  {
116  policy_ = artdaq::makeRoutingMasterPolicy(policy_plugin_spec, policy_pset_);
117  }
118  catch (...)
119  {
120  std::stringstream exception_string;
121  exception_string << "Exception thrown during initialization of policy of type \""
122  << policy_plugin_spec << "\"";
123 
124  ExceptionHandler(ExceptionHandlerRethrow::no, exception_string.str());
125 
126  TLOG_DEBUG(name_) << "FHiCL parameter set used to initialize the policy which threw an exception: " << policy_pset_.to_string() << TLOG_ENDL;
127 
128  return false;
129  }
130 
131  rt_priority_ = daq_pset.get<int>("rt_priority", 0);
132  sender_ranks_ = daq_pset.get<std::vector<int>>("sender_ranks");
133  num_receivers_ = policy_->GetReceiverCount();
134 
135  receive_ack_events_ = std::vector<epoll_event>(sender_ranks_.size());
136  receive_token_events_ = std::vector<epoll_event>(num_receivers_);
137 
138  auto mode = daq_pset.get<bool>("senders_send_by_send_count", false);
139  routing_mode_ = mode ? detail::RoutingMasterMode::RouteBySendCount : detail::RoutingMasterMode::RouteBySequenceID;
140  max_table_update_interval_ms_ = daq_pset.get<size_t>("table_update_interval_ms", 1000);
141  current_table_interval_ms_ = max_table_update_interval_ms_;
142  max_ack_cycle_count_ = daq_pset.get<size_t>("table_ack_retry_count", 5);
143  receive_token_port_ = daq_pset.get<int>("routing_token_port", 35555);
144  send_tables_port_ = daq_pset.get<int>("table_update_port", 35556);
145  receive_acks_port_ = daq_pset.get<int>("table_acknowledge_port", 35557);
146  send_tables_address_ = daq_pset.get<std::string>("table_update_address", "227.128.12.28");
147  receive_address_ = daq_pset.get<std::string>("routing_master_hostname", "localhost");
148 
149  // fetch the monitoring parameters and create the MonitoredQuantity instances
150  statsHelper_.createCollectors(daq_pset, 100, 30.0, 60.0, TABLE_UPDATES_STAT_KEY);
151 
152  shutdown_requested_.store(false);
153  start_recieve_token_thread_();
154  return true;
155 }
156 
157 bool artdaq::RoutingMasterCore::start(art::RunID id, uint64_t, uint64_t)
158 {
159  stop_requested_.store(false);
160  pause_requested_.store(false);
161 
162  statsHelper_.resetStatistics();
163  policy_->Reset();
164 
165  metricMan_.do_start();
166  run_id_ = id;
167  table_update_count_ = 0;
168  received_token_count_ = 0;
169 
170  TLOG_DEBUG(name_) << "Started run " << std::to_string(run_id_.run()) << TLOG_ENDL;
171  return true;
172 }
173 
174 bool artdaq::RoutingMasterCore::stop(uint64_t, uint64_t)
175 {
176  TLOG_DEBUG(name_) << "Stopping run " << std::to_string(run_id_.run())
177  << " after " << std::to_string(table_update_count_) << " table updates."
178  << " and " << received_token_count_ << " received tokens." << TLOG_ENDL;
179  stop_requested_.store(true);
180  return true;
181 }
182 
183 bool artdaq::RoutingMasterCore::pause(uint64_t, uint64_t)
184 {
185  TLOG_DEBUG(name_) << "Pausing run " << std::to_string(run_id_.run())
186  << " after " << table_update_count_ << " table updates."
187  << " and " << received_token_count_ << " received tokens." << TLOG_ENDL;
188  pause_requested_.store(true);
189  return true;
190 }
191 
192 bool artdaq::RoutingMasterCore::resume(uint64_t, uint64_t)
193 {
194  TLOG_DEBUG(name_) << "Resuming run " << run_id_.run() << TLOG_ENDL;
195  policy_->Reset();
196  pause_requested_.store(false);
197  metricMan_.do_start();
198  return true;
199 }
200 
202 {
203  policy_.reset(nullptr);
204  metricMan_.shutdown();
205  shutdown_requested_.store(true);
206  return true;
207 }
208 
209 bool artdaq::RoutingMasterCore::soft_initialize(fhicl::ParameterSet const& pset, uint64_t e, uint64_t f)
210 {
211  TLOG_DEBUG(name_) << "soft_initialize method called with "
212  << "ParameterSet = \"" << pset.to_string()
213  << "\"." << TLOG_ENDL;
214  return initialize(pset, e, f);
215 }
216 
217 bool artdaq::RoutingMasterCore::reinitialize(fhicl::ParameterSet const& pset, uint64_t e, uint64_t f)
218 {
219  TLOG_DEBUG(name_) << "reinitialize method called with "
220  << "ParameterSet = \"" << pset.to_string()
221  << "\"." << TLOG_ENDL;
222  return initialize(pset, e, f);
223 }
224 
226 {
227  if (rt_priority_ > 0)
228  {
229 #pragma GCC diagnostic push
230 #pragma GCC diagnostic ignored "-Wmissing-field-initializers"
231  sched_param s_param = {};
232  s_param.sched_priority = rt_priority_;
233  if (pthread_setschedparam(pthread_self(), SCHED_RR, &s_param))
234  TLOG_WARNING(name_) << "setting realtime priority failed" << TLOG_ENDL;
235 #pragma GCC diagnostic pop
236  }
237 
238  // try-catch block here?
239 
240  // how to turn RT PRI off?
241  if (rt_priority_ > 0)
242  {
243 #pragma GCC diagnostic push
244 #pragma GCC diagnostic ignored "-Wmissing-field-initializers"
245  sched_param s_param = {};
246  s_param.sched_priority = rt_priority_;
247  int status = pthread_setschedparam(pthread_self(), SCHED_RR, &s_param);
248  if (status != 0)
249  {
250  TLOG_ERROR(name_)
251  << "Failed to set realtime priority to " << std::to_string(rt_priority_)
252  << ", return code = " << status << TLOG_ENDL;
253  }
254 #pragma GCC diagnostic pop
255  }
256 
257  //MPI_Barrier(local_group_comm_);
258 
259  TLOG_DEBUG(name_) << "Sending initial table." << TLOG_ENDL;
260  auto startTime = artdaq::MonitoredQuantity::getCurrentTime();
261  auto nextSendTime = startTime;
262  double delta_time;
263  while (true)
264  {
265  if (stop_requested_ || pause_requested_) { break; }
266  startTime = artdaq::MonitoredQuantity::getCurrentTime();
267 
268  if (startTime >= nextSendTime)
269  {
270  auto table = policy_->GetCurrentTable();
271  if (table.size() > 0)
272  {
273  send_event_table(table);
274  ++table_update_count_;
275  delta_time = artdaq::MonitoredQuantity::getCurrentTime() - startTime;
276  statsHelper_.addSample(TABLE_UPDATES_STAT_KEY, delta_time);
277  TRACE(16, "%s::process_fragments TABLE_UPDATES_STAT_KEY=%f", name_.c_str(), delta_time);
278  }
279  else
280  {
281  TLOG_WARNING(name_) << "No tokens received in this update interval! This is most likely a Very Bad Thing!" << TLOG_ENDL;
282  }
283  auto max_tokens = policy_->GetMaxNumberOfTokens();
284  if (max_tokens > 0)
285  {
286  auto frac = table.size() / static_cast<double>(max_tokens);
287  if (frac > 0.75) current_table_interval_ms_ = 9 * current_table_interval_ms_ / 10;
288  if (frac < 0.5) current_table_interval_ms_ = 11 * current_table_interval_ms_ / 10;
289  if (current_table_interval_ms_ > max_table_update_interval_ms_) current_table_interval_ms_ = max_table_update_interval_ms_;
290  if (current_table_interval_ms_ < 1) current_table_interval_ms_ = 1;
291  }
292  nextSendTime = startTime + current_table_interval_ms_ / 1000.0;
293  TLOG_DEBUG(name_) << "current_table_interval_ms is now " << current_table_interval_ms_ << TLOG_ENDL;
294  }
295  else
296  {
297  usleep(current_table_interval_ms_ * 10); // 1/100 of the table update interval
298  }
299  }
300 
301  metricMan_.do_stop();
302 
303  policy_.reset(nullptr);
304  return table_update_count_;
305 }
306 
308 {
309  // Reconnect table socket, if necessary
310  if (table_socket_ == -1)
311  {
312  table_socket_ = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
313  if (!table_socket_)
314  {
315  TLOG_ERROR(name_) << "I failed to create the socket for sending Data Requests! Errno: " << std::to_string(errno) << TLOG_ENDL;
316  exit(1);
317  }
318  auto sts = ResolveHost(send_tables_address_.c_str(), send_tables_port_, send_tables_addr_);
319  if (sts == -1)
320  {
321  TLOG_ERROR(name_) << "Unable to resolve table_update_address" << TLOG_ENDL;
322  exit(1);
323  }
324 
325  auto yes = 1;
326  if (receive_address_ != "localhost")
327  {
328  TLOG_DEBUG(name_) << "Making sure that multicast sending uses the correct interface for hostname " << receive_address_ << TLOG_ENDL;
329  struct in_addr addr;
330  sts = ResolveHost(receive_address_.c_str(), addr);
331  if (sts == -1)
332  {
333  throw art::Exception(art::errors::Configuration) << "RoutingMasterCore: Unable to resolve routing_master_address" << std::endl;;
334  }
335 
336  if (setsockopt(table_socket_, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes)) < 0)
337  {
338  throw art::Exception(art::errors::Configuration) <<
339  "RoutingMasterCore: Unable to enable port reuse on table update socket" << std::endl;
340  exit(1);
341  }
342 
343  if (setsockopt(table_socket_, IPPROTO_IP, IP_MULTICAST_IF, &addr, sizeof(addr)) == -1)
344  {
345  TLOG_ERROR(name_) << "Cannot set outgoing interface. Errno: " << std::to_string(errno) << TLOG_ENDL;
346  exit(1);
347  }
348  }
349  if (setsockopt(table_socket_, SOL_SOCKET, SO_BROADCAST, (void*)&yes, sizeof(int)) == -1)
350  {
351  TLOG_ERROR(name_) << "Cannot set request socket to broadcast. Errno: " << std::to_string(errno) << TLOG_ENDL;
352  exit(1);
353  }
354  }
355 
356  // Reconnect ack socket, if necessary
357  if (ack_socket_ == -1)
358  {
359  ack_socket_ = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
360  if (!ack_socket_)
361  {
362  throw art::Exception(art::errors::Configuration) << "RoutingMasterCore: Error creating socket for receiving table update acks!" << std::endl;
363  exit(1);
364  }
365 
366  struct sockaddr_in si_me_request;
367 
368  auto yes = 1;
369  if (setsockopt(ack_socket_, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes)) < 0)
370  {
371  throw art::Exception(art::errors::Configuration) <<
372  "RoutingMasterCore: Unable to enable port reuse on ack socket" << std::endl;
373  exit(1);
374  }
375  memset(&si_me_request, 0, sizeof(si_me_request));
376  si_me_request.sin_family = AF_INET;
377  si_me_request.sin_port = htons(receive_acks_port_);
378  si_me_request.sin_addr.s_addr = htonl(INADDR_ANY);
379  if (bind(ack_socket_, reinterpret_cast<struct sockaddr *>(&si_me_request), sizeof(si_me_request)) == -1)
380  {
381  throw art::Exception(art::errors::Configuration) <<
382  "RoutingMasterCore: Cannot bind request socket to port " << receive_acks_port_ << std::endl;
383  exit(1);
384  }
385  TLOG_DEBUG(name_) << "Listening for acks on 0.0.0.0 port " << receive_acks_port_ << TLOG_ENDL;
386  }
387 
388  auto acks = std::unordered_map<int, bool>();
389  for (auto& r : sender_ranks_)
390  {
391  acks[r] = false;
392  }
393  auto counter = 0U;
394  auto start_time = std::chrono::steady_clock::now();
395  while (std::count_if(acks.begin(), acks.end(), [](std::pair<int, bool> p) {return !p.second; }) > 0)
396  {
397  // Send table update
398  auto header = detail::RoutingPacketHeader(routing_mode_, packet.size());
399  auto packetSize = sizeof(detail::RoutingPacketEntry) * packet.size();
400 
401  TLOG_DEBUG(name_) << "Sending table information for " << std::to_string(header.nEntries) << " events to multicast group " << send_tables_address_ << ", port " << send_tables_port_ << TLOG_ENDL;
402  if (sendto(table_socket_, &header, sizeof(detail::RoutingPacketHeader), 0, reinterpret_cast<struct sockaddr *>(&send_tables_addr_), sizeof(send_tables_addr_)) < 0)
403  {
404  TLOG_ERROR(name_) << "Error sending request message header" << TLOG_ENDL;
405  }
406  if (sendto(table_socket_, &packet[0], packetSize, 0, reinterpret_cast<struct sockaddr *>(&send_tables_addr_), sizeof(send_tables_addr_)) < 0)
407  {
408  TLOG_ERROR(name_) << "Error sending request message data" << TLOG_ENDL;
409  }
410 
411  // Collect acks
412 
413  auto first = packet[0].sequence_id;
414  auto last = packet.rbegin()->sequence_id;
415  TLOG_DEBUG(name_) << "Expecting acks to have first= " << std::to_string(first) << ", and last= " << std::to_string(last) << TLOG_ENDL;
416 
417 
418  auto startTime = std::chrono::steady_clock::now();
419  while (std::count_if(acks.begin(), acks.end(), [](std::pair<int, bool> p) {return !p.second; }) > 0)
420  {
421  auto currentTime = std::chrono::steady_clock::now();
422  auto table_ack_wait_time_ms = current_table_interval_ms_ / max_ack_cycle_count_;
423  if (static_cast<size_t>(std::chrono::duration_cast<std::chrono::milliseconds>(currentTime - startTime).count()) > table_ack_wait_time_ms)
424  {
425  if (counter > max_ack_cycle_count_ && table_update_count_ > 0)
426  {
427  TLOG_ERROR(name_) << "Did not receive acks from all senders after resending table " << std::to_string(counter) << " times during the table_update_interval. Check the status of the senders!" << TLOG_ENDL;
428  break;
429  }
430  TLOG_WARNING(name_) << "Did not receive acks from all senders within the table_ack_wait_time. Resending table update" << TLOG_ENDL;
431  break;
432  }
433 
434  TLOG_ARB(20, name_) << "send_event_table: Polling Request socket for new requests" << TLOG_ENDL;
435  auto ready = true;
436  while (ready)
437  {
439  if (recvfrom(ack_socket_, &buffer, sizeof(detail::RoutingAckPacket), MSG_DONTWAIT, NULL, NULL) < 0)
440  {
441  if (errno == EWOULDBLOCK || errno == EAGAIN)
442  {
443  TLOG_ARB(20, name_) << "send_event_table: No more ack datagrams on ack socket." << TLOG_ENDL;
444  ready = false;
445  }
446  else
447  {
448  TLOG_ERROR(name_) << "An unexpected error occurred during ack packet receive" << TLOG_ENDL;
449  exit(2);
450  }
451  }
452  else
453  {
454  TLOG_DEBUG(name_) << "Ack packet from rank " << buffer.rank << " has first= " << std::to_string(buffer.first_sequence_id) << " and last= " << std::to_string(buffer.last_sequence_id) << TLOG_ENDL;
455  if (acks.count(buffer.rank) && buffer.first_sequence_id == first && buffer.last_sequence_id == last)
456  {
457  TLOG_DEBUG(name_) << "Received table update acknowledgement from sender with rank " << std::to_string(buffer.rank) << "." << TLOG_ENDL;
458  acks[buffer.rank] = true;
459  TLOG_DEBUG(name_) << "There are now " << std::count_if(acks.begin(), acks.end(), [](std::pair<int, bool> p) {return !p.second; }) << " acks outstanding" << TLOG_ENDL;
460  }
461  else
462  {
463  if (!acks.count(buffer.rank)) { TLOG_ERROR(name_) << "Received acknowledgement from invalid rank " << buffer.rank << "! Cross-talk between RoutingMasters means there's a configuration error!" << TLOG_ENDL; }
464  else { TLOG_WARNING(name_) << "Received acknowledgement from rank " << buffer.rank << " that had incorrect sequence ID information. Discarding." << TLOG_ENDL; }
465  }
466  }
467  }
468  usleep(table_ack_wait_time_ms * 1000 / 10);
469  }
470  }
471  if (metricMan)
472  {
473  std::chrono::duration<double, std::ratio<1>> delta = std::chrono::steady_clock::now() - start_time;
474  metricMan->sendMetric("Avg Table Acknowledge Time", delta.count(), "seconds", 3, MetricMode::Average);
475  }
476 }
477 
478 void artdaq::RoutingMasterCore::receive_tokens_()
479 {
480  while (!shutdown_requested_)
481  {
482  TLOG_DEBUG(name_) << "Receive Token loop start" << TLOG_ENDL;
483  if (token_socket_ == -1)
484  {
485  TLOG_DEBUG(name_) << "Opening token listener socket" << TLOG_ENDL;
486  token_socket_ = TCP_listen_fd(receive_token_port_, 3 * sizeof(detail::RoutingToken));
487 
488  if (token_epoll_fd_ != -1) close(token_epoll_fd_);
489  struct epoll_event ev;
490  token_epoll_fd_ = epoll_create1(0);
491  ev.events = EPOLLIN | EPOLLPRI;
492  ev.data.fd = token_socket_;
493  if (epoll_ctl(token_epoll_fd_, EPOLL_CTL_ADD, token_socket_, &ev) == -1)
494  {
495  TLOG_ERROR(name_) << "Could not register listen socket to epoll fd" << TLOG_ENDL;
496  exit(3);
497  }
498  }
499  if (token_socket_ == -1 || token_epoll_fd_ == -1)
500  {
501  TLOG_DEBUG(name_) << "One of the listen sockets was not opened successfully." << TLOG_ENDL;
502  return;
503  }
504 
505  auto nfds = epoll_wait(token_epoll_fd_, &receive_token_events_[0], receive_token_events_.size(), current_table_interval_ms_);
506  if (nfds == -1)
507  {
508  perror("epoll_wait");
509  exit(EXIT_FAILURE);
510  }
511 
512  TLOG_DEBUG(name_) << "Received " << std::to_string(nfds) << " events" << TLOG_ENDL;
513  for (auto n = 0; n < nfds; ++n)
514  {
515  if (receive_token_events_[n].data.fd == token_socket_)
516  {
517  TLOG_DEBUG(name_) << "Accepting new connection on token_socket" << TLOG_ENDL;
518  sockaddr_in addr;
519  socklen_t arglen = sizeof(addr);
520  auto conn_sock = accept(token_socket_, (struct sockaddr *)&addr, &arglen);
521 
522  if (conn_sock == -1)
523  {
524  perror("accept");
525  exit(EXIT_FAILURE);
526  }
527 
528  receive_token_addrs_[conn_sock] = std::string(inet_ntoa(addr.sin_addr));
529  struct epoll_event ev;
530  ev.events = EPOLLIN | EPOLLET;
531  ev.data.fd = conn_sock;
532  if (epoll_ctl(token_epoll_fd_, EPOLL_CTL_ADD, conn_sock, &ev) == -1)
533  {
534  perror("epoll_ctl: conn_sock");
535  exit(EXIT_FAILURE);
536  }
537  }
538  else
539  {
540  auto startTime = artdaq::MonitoredQuantity::getCurrentTime();
541  detail::RoutingToken buff;
542  auto sts = read(receive_token_events_[n].data.fd, &buff, sizeof(detail::RoutingToken));
543  if (sts != sizeof(detail::RoutingToken) || buff.header != TOKEN_MAGIC)
544  {
545  TLOG_ERROR(name_) << "Received invalid token from " << receive_token_addrs_[receive_token_events_[n].data.fd] << TLOG_ENDL;
546  }
547  else
548  {
549  TLOG_DEBUG(name_) << "Received token from " << std::to_string(buff.rank) << " indicating " << buff.new_slots_free << " slots are free." << TLOG_ENDL;
550  received_token_count_ += buff.new_slots_free;
551  if (routing_mode_ == detail::RoutingMasterMode::RouteBySequenceID)
552  {
553  policy_->AddReceiverToken(buff.rank, buff.new_slots_free);
554  }
555  else if (routing_mode_ == detail::RoutingMasterMode::RouteBySendCount)
556  {
557  if (!received_token_counter_.count(buff.rank)) received_token_counter_[buff.rank] = 0;
558  received_token_counter_[buff.rank] += buff.new_slots_free;
559  TLOG_DEBUG(name_) << "RoutingMasterMode is RouteBySendCount. I have " << received_token_counter_[buff.rank] << " tokens for rank " << buff.rank << " and I need " << sender_ranks_.size() << "." << TLOG_ENDL;
560  while (received_token_counter_[buff.rank] >= sender_ranks_.size())
561  {
562  TLOG_DEBUG(name_) << "RoutingMasterMode is RouteBySendCount. I have " << received_token_counter_[buff.rank] << " tokens for rank " << buff.rank << " and I need " << sender_ranks_.size()
563  << "... Sending token to policy" << TLOG_ENDL;
564  policy_->AddReceiverToken(buff.rank, 1);
565  received_token_counter_[buff.rank] -= sender_ranks_.size();
566  }
567  }
568  }
569  auto delta_time = artdaq::MonitoredQuantity::getCurrentTime() - startTime;
570  statsHelper_.addSample(TOKENS_RECEIVED_STAT_KEY, delta_time);
571 
572  }
573  }
574  }
575 }
576 
577 void artdaq::RoutingMasterCore::start_recieve_token_thread_()
578 {
579  if (ev_token_receive_thread_.joinable()) ev_token_receive_thread_.join();
580  TLOG_INFO(name_) << "Starting Token Reception Thread" << TLOG_ENDL;
581  ev_token_receive_thread_ = std::thread(&RoutingMasterCore::receive_tokens_, this);
582 }
583 
584 std::string artdaq::RoutingMasterCore::report(std::string const&) const
585 {
586  std::string resultString;
587 
588  // if we haven't been able to come up with any report so far, say so
589  auto tmpString = name_ + " run number = " + std::to_string(run_id_.run())
590  + ", table updates sent = " + std::to_string(table_update_count_)
591  + ", Receiver tokens received = " + std::to_string(received_token_count_);
592  return tmpString;
593 }
594 
595 std::string artdaq::RoutingMasterCore::buildStatisticsString_() const
596 {
597  std::ostringstream oss;
598  oss << name_ << " statistics:" << std::endl;
599 
600  auto mqPtr = artdaq::StatisticsCollection::getInstance().getMonitoredQuantity(TABLE_UPDATES_STAT_KEY);
601  if (mqPtr.get() != nullptr)
602  {
603  artdaq::MonitoredQuantityStats stats;
604  mqPtr->getStats(stats);
605  oss << " Table Update statistics: "
606  << stats.recentSampleCount << " table updates sent at "
607  << stats.recentSampleRate << " table updates/sec, , monitor window = "
608  << stats.recentDuration << " sec" << std::endl;
609  oss << " Average times per table update: ";
610  if (stats.recentSampleRate > 0.0)
611  {
612  oss << " elapsed time = "
613  << (1.0 / stats.recentSampleRate) << " sec";
614  }
615  oss << ", avg table acknowledgement wait time = "
616  << (mqPtr->getRecentValueSum() / sender_ranks_.size()) << " sec" << std::endl;
617  }
618 
619  mqPtr = artdaq::StatisticsCollection::getInstance().getMonitoredQuantity(TOKENS_RECEIVED_STAT_KEY);
620  if (mqPtr.get() != nullptr)
621  {
622  artdaq::MonitoredQuantityStats stats;
623  mqPtr->getStats(stats);
624  oss << " Received Token statistics: "
625  << stats.recentSampleCount << " tokens received at "
626  << stats.recentSampleRate << " tokens/sec, , monitor window = "
627  << stats.recentDuration << " sec" << std::endl;
628  oss << " Average times per token: ";
629  if (stats.recentSampleRate > 0.0)
630  {
631  oss << " elapsed time = "
632  << (1.0 / stats.recentSampleRate) << " sec";
633  }
634  oss << ", input token wait time = "
635  << mqPtr->getRecentValueSum() << " sec" << std::endl;
636  }
637 
638  return oss.str();
639 }
640 
641 void artdaq::RoutingMasterCore::sendMetrics_()
642 {
643  auto mqPtr = artdaq::StatisticsCollection::getInstance().getMonitoredQuantity(TABLE_UPDATES_STAT_KEY);
644  if (mqPtr.get() != nullptr)
645  {
646  artdaq::MonitoredQuantityStats stats;
647  mqPtr->getStats(stats);
648  metricMan_.sendMetric("Table Update Count", static_cast<unsigned long>(stats.fullSampleCount), "updates", 1, MetricMode::Accumulate);
649  metricMan_.sendMetric("Table Update Rate", stats.recentSampleRate, "updates/sec", 1, MetricMode::Average);
650  metricMan_.sendMetric("Average Sender Acknowledgement Time", (mqPtr->getRecentValueSum() / sender_ranks_.size()), "seconds", 3, MetricMode::Average);
651  }
652 
653  mqPtr = artdaq::StatisticsCollection::getInstance().getMonitoredQuantity(TOKENS_RECEIVED_STAT_KEY);
654  if (mqPtr.get() != nullptr)
655  {
656  artdaq::MonitoredQuantityStats stats;
657  mqPtr->getStats(stats);
658  metricMan_.sendMetric("Receiver Token Count", static_cast<unsigned long>(stats.fullSampleCount), "updates", 1, MetricMode::Accumulate);
659  metricMan_.sendMetric("Receiver Token Rate", stats.recentSampleRate, "updates/sec", 1, MetricMode::Average);
660  metricMan_.sendMetric("Total Receiver Token Wait Time", mqPtr->getRecentValueSum(), "seconds", 3, MetricMode::Average);
661  }
662 }
bool resume(uint64_t, uint64_t)
Resumes the RoutingMasterCore.
size_t process_event_table()
Main loop of the RoutingMasterCore. Determines when to send the next table update, asks the RoutingMasterPolicy for the table to send, and sends it.
void addMonitoredQuantityName(std::string const &statKey)
Add a MonitoredQuantity name to the list.
int ResolveHost(char const *host_in, in_addr &addr)
Convert a string hostname to a in_addr suitable for socket communication.
Definition: TCPConnect.cc:29
bool soft_initialize(fhicl::ParameterSet const &pset, uint64_t timeout, uint64_t timestamp)
Soft-Initializes the RoutingMasterCore.
static const std::string TOKENS_RECEIVED_STAT_KEY
Key for the Tokens Received MonitoredQuantity.
A row of the Routing Table.
bool reinitialize(fhicl::ParameterSet const &pset, uint64_t timeout, uint64_t timestamp)
Reinitializes the RoutingMasterCore.
bool start(art::RunID id, uint64_t, uint64_t)
Start the RoutingMasterCore.
A RoutingAckPacket contains the rank of the table receiver, plus the first and last sequence IDs in t...
RoutingMasterCore(int rank, std::string name)
RoutingMasterCore Constructor.
bool initialize(fhicl::ParameterSet const &pset, uint64_t, uint64_t)
Processes the initialize request.
The RoutingToken contains the magic bytes, the rank of the token sender, and the number of slots free...
int TCP_listen_fd(int port, int rcvbuf)
Create a TCP listening socket on the given port and INADDR_ANY, with the given receive buffer...
Fragment::sequence_id_t first_sequence_id
The first sequence ID in the received RoutingPacket.
bool pause(uint64_t, uint64_t)
Pauses the RoutingMasterCore.
The header of the Routing Table, containing the magic bytes and the number of entries.
Fragment::sequence_id_t last_sequence_id
The last sequence ID in the received RoutingPacket.
std::vector< RoutingPacketEntry > RoutingPacket
A RoutingPacket is simply a vector of RoutingPacketEntry objects. It is not suitable for network tran...
bool stop(uint64_t, uint64_t)
Stops the RoutingMasterCore.
int rank
The rank from which the RoutingAckPacket came.
std::unique_ptr< RoutingMasterPolicy > makeRoutingMasterPolicy(std::string const &policy_plugin_spec, fhicl::ParameterSet const &ps)
Load a RoutingMasterPolicy plugin.
std::string report(std::string const &) const
Send a report on the current status of the RoutingMasterCore.
static const std::string TABLE_UPDATES_STAT_KEY
Key for Table Update count MonnitoredQuantity.
bool shutdown(uint64_t)
Shuts Down the RoutingMasterCore.
void send_event_table(detail::RoutingPacket table)
Sends a detail::RoutingPacket to the table receivers.