XRootD
XrdPfcResourceMonitor.cc
Go to the documentation of this file.
2 #include "XrdPfc.hh"
4 #include "XrdPfcFsTraversal.hh"
5 #include "XrdPfcDirState.hh"
8 #include "XrdPfcTrace.hh"
9 #include "XrdPfcPurgePin.hh"
10 
11 #include "XrdOss/XrdOss.hh"
12 
13 #include <algorithm>
14 
15 // #define RM_DEBUG
16 #ifdef RM_DEBUG
17 #define dprintf(...) printf(__VA_ARGS__)
18 #else
19 #define dprintf(...) (void(0))
20 #endif
21 
22 using namespace XrdPfc;
23 
24 namespace
25 {
26  XrdSysTrace* GetTrace() { return Cache::GetInstance().GetTrace(); }
27  const char *m_traceID = "ResourceMonitor";
28 }
29 
30 //------------------------------------------------------------------------------
31 
33  m_fs_state(* new DataFsState),
34  m_oss(oss)
35 {}
36 
38 {
39  delete &m_fs_state;
40 }
41 
42 //------------------------------------------------------------------------------
43 // Initial scan
44 //------------------------------------------------------------------------------
45 
47 {
48  m_dir_scan_mutex.Lock();
49  if (m_dir_scan_in_progress) {
50  bool dir_checked = false;
51  m_dir_scan_open_requests.push_back({lfn, cond, dir_checked});
52  cond.Lock();
53  m_dir_scan_mutex.UnLock();
54  while ( ! dir_checked)
55  cond.Wait();
56  cond.UnLock();
57  } else {
58  m_dir_scan_mutex.UnLock();
59  }
60 }
61 
62 void ResourceMonitor::process_inter_dir_scan_open_requests(FsTraversal &fst)
63 {
64  m_dir_scan_mutex.Lock();
65  while ( ! m_dir_scan_open_requests.empty())
66  {
67  LfnCondRecord &lcr = m_dir_scan_open_requests.front();
68  m_dir_scan_mutex.UnLock();
69 
70  cross_check_or_process_oob_lfn(lcr.f_lfn, fst);
71  lcr.f_cond.Lock();
72  lcr.f_checked = true;
73  lcr.f_cond.Signal();
74  lcr.f_cond.UnLock();
75 
76  m_dir_scan_mutex.Lock();
77  m_dir_scan_open_requests.pop_front();
78  }
79  m_dir_scan_mutex.UnLock();
80 }
81 
82 void ResourceMonitor::cross_check_or_process_oob_lfn(const std::string &lfn, FsTraversal &fst)
83 {
84  // Check if lfn has already been processed ... or process it now and mark
85  // the DirState accordingly (partially processed oob).
86  static const char *trc_pfx = "cross_check_or_process_oob_lfn() ";
87 
88  // last-existing-dir-state is currently not used. It was expected some work would need
89  // to be done along the path ... but we know they are all empty as they were just
90  // created. We could tag them all as already scanned.
91  DirState *last_existing_ds = nullptr;
92  DirState *ds = m_fs_state.find_dirstate_for_lfn(lfn, &last_existing_ds);
93  if (ds->m_scanned)
94  return;
95 
96  size_t pos = lfn.find_last_of("/");
97  std::string dir = (pos == std::string::npos) ? "" : lfn.substr(0, pos);
98 
99  XrdOssDF *dhp = m_oss.newDir(trc_pfx);
100  if (dhp->Opendir(dir.c_str(), fst.default_env()) == XrdOssOK)
101  {
102  fst.slurp_dir_ll(*dhp, ds->m_depth, dir.c_str(), trc_pfx);
103 
104  // XXXX clone of function below .... move somewhere? Esp. removal of non-paired files?
105  DirUsage &here = ds->m_here_usage;
106  for (auto it = fst.m_current_files.begin(); it != fst.m_current_files.end(); ++it)
107  {
108  if (it->second.has_data && it->second.has_cinfo) {
109  here.m_StBlocks += it->second.stat_data.st_blocks;
110  here.m_NFiles += 1;
111  }
112  }
113  }
114  delete dhp;
115  ds->m_scanned = true;
116 }
117 
119 {
120  dprintf("In scan_dir_and_recurse for '%s', size of dir_vec = %d, file_stat_map = %d\n",
121  fst.m_current_path.c_str(),
122  (int)fst.m_current_dirs.size(), (int)fst.m_current_files.size());
123 
124  // Breadth first, accumulate into "here", unless it was already scanned via an
125  // OOB open file request.
126  if ( ! fst.m_dir_state->m_scanned)
127  {
128  DirUsage &here = fst.m_dir_state->m_here_usage;
129  for (auto it = fst.m_current_files.begin(); it != fst.m_current_files.end(); ++it)
130  {
131  dprintf("would be doing something with %s ... has_data=%d, has_cinfo=%d\n",
132  it->first.c_str(), it->second.has_data, it->second.has_cinfo);
133 
134  // XXX Make some of these optional?
135  // Remove files that do not have both cinfo and data?
136  // Remove empty directories before even descending?
137  // Leave this for some consistency pass?
138  // Note that FsTraversal supports ignored paths ... some details (config, N2N) to be clarified.
139 
140  if (it->second.has_data && it->second.has_cinfo) {
141  here.m_StBlocks += it->second.stat_data.st_blocks;
142  here.m_NFiles += 1;
143  }
144  }
145  fst.m_dir_state->m_scanned = true;
146  }
147 
148  // Swap-out directories as inter_dir_scan can use the FsTraversal.
149  std::vector<std::string> dirs;
150  dirs.swap(fst.m_current_dirs);
151 
152  if (++m_dir_scan_check_counter >= 100)
153  {
154  process_inter_dir_scan_open_requests(fst);
155  m_dir_scan_check_counter = 0;
156  }
157 
158  // Descend into sub-dirs, do not accumulate into recursive_subdir_usage yet. This is done
159  // in a separate pass to allow for proper accounting of files being opened during the initial scan.
160  for (auto &dname : dirs)
161  {
162  if (fst.cd_down(dname))
163  {
165  fst.cd_up();
166  }
167  // XXX else try to remove it?
168  }
169 }
170 
172 {
173  // Called after PFC configuration is complete, but before full startup of the daemon.
174  // Base line usages are accumulated as part of the file-system, traversal.
175 
177 
178  DirState *root_ds = m_fs_state.get_root();
179  FsTraversal fst(m_oss);
180  fst.m_protected_top_dirs.insert("pfc-stats"); // XXXX This should come from config. Also: N2N?
181 
182  if ( ! fst.begin_traversal(root_ds, "/"))
183  return false;
184 
185  // The following are initialized in ResourceMonitor.hh to avoid a race at startup:
186  // m_dir_scan_in_progress = true;
187  // m_dir_scan_check_counter = 0;
188 
190 
191  fst.end_traversal();
192 
193  // We have all directories scanned, available in DirState tree, let all remaining files go
194  // and then we shall do the upward propagation of usages.
195  {
196  XrdSysMutexHelper _lock(m_dir_scan_mutex);
197  m_dir_scan_in_progress = false;
198  m_dir_scan_check_counter = 0;
199  }
200  // m_dir_scan_open_requests should now be final, ie, no new entries will be added.
201  while ( ! m_dir_scan_open_requests.empty())
202  {
203  LfnCondRecord &lcr = m_dir_scan_open_requests.front();
204  lcr.f_cond.Lock();
205  lcr.f_checked = true;
206  lcr.f_cond.Signal();
207  lcr.f_cond.UnLock();
208 
209  m_dir_scan_open_requests.pop_front();
210  }
211 
212  // Do upward propagation of usages.
214  m_current_usage_in_st_blocks = root_ds->m_here_usage.m_StBlocks +
217 
218  return true;
219 }
220 
221 //------------------------------------------------------------------------------
222 // Processing of queues
223 //------------------------------------------------------------------------------
224 
226 {
227  static const char *trc_pfx = "process_queues() ";
228 
229  // Assure that we pick up only entries that are present now.
230  // We really want all open records to be processed before file-stats updates
231  // and all those before the close records.
232  // Purges are sort of tangential as they really just modify bytes / number
233  // of files in a directory and do not deal with any persistent file id tokens.
234 
235  int n_records = 0;
236  {
237  XrdSysMutexHelper _lock(&m_queue_mutex);
238  n_records += m_file_open_q.swap_queues();
239  n_records += m_file_update_stats_q.swap_queues();
240  n_records += m_file_close_q.swap_queues();
241  n_records += m_file_purge_q1.swap_queues();
242  n_records += m_file_purge_q2.swap_queues();
243  n_records += m_file_purge_q3.swap_queues();
244  ++m_queue_swap_u1;
245  }
246 
247  for (auto &i : m_file_open_q.read_queue())
248  {
249  // i.id: LFN, i.record: OpenRecord
250  AccessToken &at = token(i.id);
251  dprintf("process file open for token %d, time %ld -- %s\n",
252  i.id, i.record.m_open_time, at.m_filename.c_str());
253 
254  // Resolve fname into DirState.
255  // We could clear the filename after this ... or keep it, should we need it later on.
256  // For now it is just used for debug printouts.
257  DirState *last_existing_ds = nullptr;
258  DirState *ds = m_fs_state.find_dirstate_for_lfn(at.m_filename, &last_existing_ds);
259  at.m_dir_state = ds;
260  ds->m_here_stats.m_NFilesOpened += 1;
261 
262  // If this is a new file figure out how many new parent dirs got created along the way.
263  if ( ! i.record.m_existing_file) {
264  ds->m_here_stats.m_NFilesCreated += 1;
265  DirState *pp = ds;
266  while (pp != last_existing_ds) {
267  pp = pp->get_parent();
269  }
270  }
271 
272  ds->m_here_usage.m_LastOpenTime = i.record.m_open_time;
273  }
274 
275  for (auto &i : m_file_update_stats_q.read_queue())
276  {
277  // i.id: token, i.record: Stats
278  AccessToken &at = token(i.id);
279  // Stats
280  DirState *ds = at.m_dir_state;
281  dprintf("process file update for token %d, %p -- %s\n",
282  i.id, ds, at.m_filename.c_str());
283 
284  ds->m_here_stats.AddUp(i.record);
285  m_current_usage_in_st_blocks += i.record.m_StBlocksAdded;
286  }
287 
288  for (auto &i : m_file_close_q.read_queue())
289  {
290  // i.id: token, i.record: CloseRecord
291  AccessToken &at = token(i.id);
292  dprintf("process file close for token %d, time %ld -- %s\n",
293  i.id, i.record.m_close_time, at.m_filename.c_str());
294 
295  DirState *ds = at.m_dir_state;
296  ds->m_here_stats.m_NFilesClosed += 1;
297  ds->m_here_usage.m_LastCloseTime = i.record.m_close_time;
298 
299  at.clear();
300  }
301  { // Release the AccessToken slots under lock.
302  XrdSysMutexHelper _lock(&m_queue_mutex);
303  for (auto &i : m_file_close_q.read_queue())
304  m_access_tokens_free_slots.push_back(i.id);
305  }
306 
307  for (auto &i : m_file_purge_q1.read_queue())
308  {
309  // i.id: DirState*, i.record: PurgeRecord
310  DirState *ds = i.id;
311  ds->m_here_stats.m_StBlocksRemoved += i.record.m_size_in_st_blocks;
312  ds->m_here_stats.m_NFilesRemoved += i.record.m_n_files;
313  m_current_usage_in_st_blocks -= i.record.m_size_in_st_blocks;
314  }
315  for (auto &i : m_file_purge_q2.read_queue())
316  {
317  // i.id: directory-path, i.record: PurgeRecord
318  DirState *ds = m_fs_state.get_root()->find_path(i.id, -1, false, false);
319  if ( ! ds) {
320  TRACE(Error, trc_pfx << "DirState not found for directory path '" << i.id << "'.");
321  // find_path can return the last dir found ... but this clearly isn't a valid purge record.
322  continue;
323  }
324  ds->m_here_stats.m_StBlocksRemoved += i.record.m_size_in_st_blocks;
325  ds->m_here_stats.m_NFilesRemoved += i.record.m_n_files;
326  m_current_usage_in_st_blocks -= i.record.m_size_in_st_blocks;
327  }
328  for (auto &i : m_file_purge_q3.read_queue())
329  {
330  // i.id: LFN, i.record: size of file in st_blocks
331  DirState *ds = m_fs_state.get_root()->find_path(i.id, -1, true, false);
332  if ( ! ds) {
333  TRACE(Error, trc_pfx << "DirState not found for LFN path '" << i.id << "'.");
334  continue;
335  }
336  ds->m_here_stats.m_StBlocksRemoved += i.record;
337  ds->m_here_stats.m_NFilesRemoved += 1;
338  m_current_usage_in_st_blocks -= i.record;
339  }
340 
341  // Read queues / vectors are cleared at swap time.
342  // We might consider reducing their capacity by half if, say, their usage is below 25%.
343 
344  return n_records;
345 }
346 
347 //------------------------------------------------------------------------------
348 // Heart beat
349 //------------------------------------------------------------------------------
350 
352 {
353  static const char *tpfx = "heart_beat() ";
354 
355  const Configuration &conf = Cache::Conf();
356 
357  const int s_queue_proc_interval = 10;
358  const int s_sshot_report_interval = conf.m_dirStatsInterval; // 1, 5, 10, 15, 30 or 60 minutes
359  const int s_purge_check_interval = 60;
360  const int s_purge_report_interval = conf.m_purgeInterval;
361  const int s_purge_cold_files_interval = conf.m_purgeInterval * conf.m_purgeAgeBasedPeriod;
362 
363  // initial scan performed as part of config
364 
365  time_t now = time(0);
366  time_t next_queue_proc_time = now + s_queue_proc_interval;
367  time_t next_sshot_report_time = (now / 60) * 60 + 60; // at next full minute
368  time_t next_purge_check_time = now + s_purge_check_interval;
369  time_t next_purge_report_time = now + s_purge_report_interval;
370  time_t next_purge_cold_files_time = now + s_purge_cold_files_interval;
371 
372  while (true)
373  {
374  time_t start = time(0);
375  time_t next_event = std::min({ next_queue_proc_time, next_sshot_report_time,
376  next_purge_check_time, next_purge_report_time, next_purge_cold_files_time });
377 
378  if (next_event > start)
379  {
380  unsigned int t_sleep = next_event - start;
381  TRACE(Debug, tpfx << "sleeping for " << t_sleep << " seconds until the next beat.");
382  sleep(t_sleep);
383  }
384 
385  // Check if purge has been running and has completed yet.
386  // For now this is only used to prevent removal of empty leaf directories
387  // during stat propagation so we do not need to wait for the condition in
388  // the above sleep.
389  if (m_purge_task_active) {
391  if (m_purge_task_complete) {
393  }
394  }
395 
396  time_t queue_swap_time = time(0);
397 
398  // Always process the queues.
399  int n_processed = process_queues();
400  next_queue_proc_time = queue_swap_time + s_queue_proc_interval;
401  TRACE(Debug, tpfx << "process_queues -- n_records=" << n_processed);
402 
403  // Always update basic info on m_fs_state (space, usage, file_usage).
405 
406  now = time(0);
407 
408  // Make planning for fs_state_update, sshot dump and purge task.
409  // Second two require the first, so figure out what is going to happen.
410  bool do_sshot_report = next_sshot_report_time <= now;
411  bool do_purge_check = next_purge_check_time <= now;
412  bool do_purge_report = next_purge_report_time <= now;
413  bool do_purge_cold_files = next_purge_cold_files_time <= now;
414 
415  // Update stats in usages if any secondary activity will happen.
416  if (do_sshot_report || do_purge_check || do_purge_report || do_purge_cold_files)
417  {
418  unlink_func unlink_foo = [&](const std::string &dp)->int {
419  int ret = m_oss.Unlink(dp.c_str());
420  if (ret != 0) {
421  TRACE(Info, tpfx << "Empty dir unlink error: " << ret << " at " << dp);
422  } else {
423  TRACE(Debug, tpfx << "Empty dir unlink success: " << dp);
424  }
425  return ret;
426  };
427 
428  // Potentially prune the empty leaf dirs even less frequently, once per hour, maybe?
429  bool purge_leaf_dirs = do_sshot_report && ! m_purge_task_active;
430  m_fs_state.update_stats_and_usages(queue_swap_time, purge_leaf_dirs, unlink_foo);
431 
432  // This reporting into log/stdout is to be removed.
433  // Meaning of conf.is_dir_stat_reporting_on() etc is to be clarified / improved.
434  if (do_sshot_report && conf.is_dir_stat_reporting_on())
435  {
436  const int store_depth = conf.m_dirStatsStoreDepth;
437  #ifdef RM_DEBUG
438  const DirState &root_ds = *m_fs_state.get_root();
439  dprintf("Snapshot n_dirs=%d, total n_dirs=%d\n", root_ds.count_dirs_to_level(store_depth),
441  #endif
442  m_fs_state.dump_recursively(store_depth);
443  }
444 
445  m_fs_state.reset_stats(queue_swap_time);
446  }
447 
448  if (do_sshot_report)
449  {
450  // Sshot reports are equidistant, at "full" reporting interval.
451  next_sshot_report_time = ((now + 1) / s_sshot_report_interval) * s_sshot_report_interval + s_sshot_report_interval;
452 
453  // This should dump out binary snapshot into /pfc-stats/, if so configured.
454 
455  // json dump to std::out for debug purpose
456  DataFsSnapshot ss(m_fs_state, m_fs_state.m_sshot_stats_reset_time);
457  const DirState &root_ds = *m_fs_state.get_root();
458  const int store_depth = conf.m_dirStatsStoreDepth;
459  const int n_sshot_dirs = root_ds.count_dirs_to_level(store_depth);
460  ss.m_dir_states.reserve(n_sshot_dirs);
461  ss.m_dir_states.emplace_back( DirStateElement(root_ds, -1) );
462  fill_sshot_vec_children(root_ds, 0, ss.m_dir_states, store_depth);
463 
464  // This should really be export to a file (preferably binary, but then bin->json command is needed, too).
465  // ss.dump();
466 
467  const char* dumpfile = "/pfc-stats/DirStat.json";
468  ss.write_json_file(dumpfile, m_oss, false);
469  m_fs_state.reset_sshot_stats(queue_swap_time);
470  }
471 
472  if (do_purge_check || do_purge_report || do_purge_cold_files)
473  {
474  perform_purge_check(do_purge_cold_files, do_purge_report ? TRACE_Info : TRACE_Debug);
475 
476  next_purge_check_time = now + s_purge_check_interval;
477  if (do_purge_report) next_purge_report_time = now + s_purge_report_interval;
478  if (do_purge_cold_files) next_purge_cold_files_time = now + s_purge_cold_files_interval;
479  }
480 
481  } // end while forever
482 }
483 
484 //------------------------------------------------------------------------------
485 // DirState export helpers
486 //------------------------------------------------------------------------------
487 
489  int parent_idx,
490  std::vector<DirStateElement> &vec,
491  int max_depth)
492 {
493  int pos = vec.size();
494  int n_children = parent_ds.m_subdirs.size();
495 
496  DirStateElement &parent_dse = vec[parent_idx];
497  parent_dse.m_daughters_begin = pos;
498  parent_dse.m_daughters_end = pos + n_children;
499 
500  if (n_children == 0) return;
501 
502  for (auto const & [name, child] : parent_ds.m_subdirs)
503  {
504  vec.emplace_back( DirStateElement(child, parent_idx) );
505  }
506 
507  if (parent_ds.m_depth < max_depth)
508  {
509  for (auto const & [name, child] : parent_ds.m_subdirs)
510  {
511  fill_sshot_vec_children(child, pos, vec, max_depth);
512  ++pos;
513  }
514  }
515 }
516 
518  int parent_idx,
519  std::vector<DirPurgeElement> &vec,
520  int max_depth)
521 {
522  int pos = vec.size();
523  int n_children = parent_ds.m_subdirs.size();
524 
525  DirPurgeElement &parent_dpe = vec[parent_idx];
526  parent_dpe.m_daughters_begin = pos;
527  parent_dpe.m_daughters_end = pos + n_children;
528 
529  if (n_children == 0) return;
530 
531  for (auto const & [name, child] : parent_ds.m_subdirs)
532  {
533  vec.emplace_back( DirPurgeElement(child, child.m_here_usage, child.m_recursive_subdir_usage, parent_idx) );
534  }
535 
536  if (parent_ds.m_depth < max_depth)
537  {
538  for (auto const & [name, child] : parent_ds.m_subdirs)
539  {
540  fill_pshot_vec_children(child, pos, vec, max_depth);
541  ++pos;
542  }
543  }
544 }
545 
546 //------------------------------------------------------------------------------
547 // Purge helpers, drivers, etc.
548 //------------------------------------------------------------------------------
549 
551 {
552  static const char *trc_pfx = "update_vs_and_file_usage_info() ";
553 
554  const auto &conf = Cache::Conf();
555  XrdOssVSInfo vsi;
556 
557  // StatVS error (after it succeeded in config) implies a memory corruption (according to Mr. H).
558  if (m_oss.StatVS(&vsi, conf.m_data_space.c_str(), 1) < 0) {
559  TRACE(Error, trc_pfx << "can't get StatVS for oss space '" << conf.m_data_space << "'. This is a fatal error.");
560  _exit(1);
561  }
562  m_fs_state.m_disk_total = vsi.Total;
563  m_fs_state.m_disk_used = vsi.Total - vsi.Free;
564  m_fs_state.m_file_usage = 512ll * m_current_usage_in_st_blocks;
565  if (m_oss.StatVS(&vsi, conf.m_meta_space.c_str(), 1) < 0) {
566  TRACE(Error, trc_pfx << "can't get StatVS for oss space '" << conf.m_meta_space << "'. This is a fatal error.");
567  _exit(1);
568  }
569  m_fs_state.m_meta_total = vsi.Total;
570  m_fs_state.m_meta_used = vsi.Total - vsi.Free;
571 }
572 
573 long long ResourceMonitor::get_file_usage_bytes_to_remove(const DataFsPurgeshot &ps, long long write_estimate, int tl)
574 {
575  // short names from config values
576  const Configuration &conf = Cache::Conf();
577  long long f0 = conf.m_fileUsageBaseline;
578  long long f1 = conf.m_fileUsageNominal;
579  long long f2 = conf.m_fileUsageMax;
580  long long w1 = conf.m_diskUsageLWM;
581  long long w2 = conf.m_diskUsageHWM;
582 
583  // get usage from purge snapshot
584  long long T = ps.m_disk_total;
585  long long x = ps.m_file_usage;
586  long long u = ps.m_disk_used;
587 
588  // get file usage increase from the previous time interval check
589  long long delta = write_estimate;
590  TRACE_INT(tl, "file usage increased since the previous purge interval in bytes: " << delta );
591 
592  long long bytes_to_remove = 0;
593 
594  // helper lambda function
595  auto clamp = [&x, &bytes_to_remove](long long lowval, long long highval)
596  {
597  long long val = x;
598  long long newval = val - bytes_to_remove;
599 
600  // removed too much
601  if (newval < lowval)
602  {
603  return lowval - val;
604  }
605 
606  // removed too little
607  if (newval > highval)
608  {
609  return val - highval;
610  }
611  // keep the original value
612  return bytes_to_remove;
613  };
614 
615  // under file quota, nothing to do
616  if (x < f0)
617  return 0;
618 
619  // total disk usage exceeds highWatermark
620  if (u >= w2)
621  {
622  TRACE_INT(tl, "Disk usage: " << ps.m_disk_used << " exceed highWatermark " << conf.m_diskUsageHWM);
623  float frac_u = static_cast<float>(u - w2) / (T - w2);
624  float frac_x = static_cast<float>(x - f0) / (f1 - f0);
625 
626  if (w2 == T)
627  {
628  bytes_to_remove = u -w1;
629  }
630  else
631  {
632  if (frac_x > frac_u)
633  {
634  // the cache is the reason for going out of w2 range
635  bytes_to_remove = (frac_x - frac_u) * (f1 - f0);
636  bytes_to_remove += delta;
637  bytes_to_remove = clamp(f0, f1);
638  }
639  else
640  {
641  // someone else is filling disk space, go to f1
642  bytes_to_remove = clamp(f0, f2);
643  }
644  return bytes_to_remove;
645  }
646  }
647 
648  // file quota and total disk usage is within normal range, check if this space usage is
649  // proportinal to disk usage and correct it
650  if (u > w1 && x > f1)
651  {
652  float frac_u = static_cast<float>(u - w1) / (w2 - w1);
653  float frac_x = static_cast<float>(x - f1) / (f2 - f1);
654  if (frac_x > frac_u)
655  {
656  TRACE_INT(tl, "Disproportional file quota usage comapared to disc usage (frac_x/frac_u) = " << frac_x << "/"<< frac_u);
657  bytes_to_remove = (frac_x - frac_u) * (f2 - f1);
658  bytes_to_remove += delta;
659  }
660 
661  // check the new x val will not be below f0
662  bytes_to_remove = clamp(f0, f2);
663  return bytes_to_remove;
664  }
665 
666  // final check: disk useage is lower that w1, check if exceed the max file usage f2
667  if (x > f2)
668  {
669  // drop usage to f2
670  // compare with global disk usage in the previous purge cycle (default 300s)
671  // check delta is not overflowing f2, else set numver of bytes to remove according remove to f0
672 
673  TRACE_INT(tl, "File usage exceeds maxim file usage. Total disk usage is under lowWatermark. Clearing to low file usage.");
674  long long f2delta = std::max(f2 - delta, f0);
675  bytes_to_remove = clamp(f0, f2delta);
676  return bytes_to_remove;
677  }
678 
679  return bytes_to_remove;
680 }
681 
682 void ResourceMonitor::perform_purge_check(bool purge_cold_files, int tl)
683 {
684  static const char *trc_pfx = "perform_purge_check() ";
685  const Configuration &conf = Cache::Conf();
686 
687  std::unique_ptr<DataFsPurgeshot> psp( new DataFsPurgeshot(m_fs_state) );
688  DataFsPurgeshot &ps = *psp;
689 
690  ps.m_file_usage = 512ll * m_current_usage_in_st_blocks;
691  // These are potentially wrong as cache might be writing over preallocated byte ranges.
693  // Can have another estimate based on eiter writes or st-blocks from purge-stats, once we have them.
694 
695  TRACE_INT(tl, trc_pfx << "Purge check:");
696 
697  ps.m_bytes_to_remove = 0;
698  if (conf.are_file_usage_limits_set())
699  {
700  ps.m_bytes_to_remove = get_file_usage_bytes_to_remove(ps, ps.m_estimated_writes_from_writeq, tl);
701  }
702  else
703  {
704  if (ps.m_disk_used > conf.m_diskUsageHWM)
705  {
706  TRACE_INT(tl, "Disk usage: " << ps.m_disk_used << " exceed highWatermark.");
708  }
709  }
710 
711  ps.m_space_based_purge = ps.m_bytes_to_remove ? 1 : 0;
712 
713  // Purge precheck -- check if age-based purge is required
714  // We ignore uvkeep time, it requires reading of cinfo files and it is enforced in File::Open() anyway.
715 
716  if (purge_cold_files && conf.is_age_based_purge_in_effect()) // || conf.is_uvkeep_purge_in_effect())
717  {
718  ps.m_age_based_purge = true;
719  }
720 
721  TRACE_INT(tl, "\tbytes_to_remove = " << ps.m_bytes_to_remove << " B");
722  TRACE_INT(tl, "\tspace_based_purge = " << ps.m_space_based_purge);
723  TRACE_INT(tl, "\tage_based_purge = " << ps.m_age_based_purge);
724 
725  bool periodic = Cache::GetInstance().GetPurgePin() ?
727 
728  if ( ! ps.m_space_based_purge && ! ps.m_age_based_purge && !periodic ) {
729  TRACE(Info, trc_pfx << "purge not required.");
731  return;
732  }
733  if (m_purge_task_active) {
734  TRACE(Warning, trc_pfx << "purge required but previous purge task is still active!");
735  return;
736  }
737 
738  TRACE(Info, trc_pfx << "scheduling purge task.");
739 
740  // At this point we have all the information: report, decide on action.
741  // There is still some missing infrastructure, especially as regards to purge-plugin:
742  // - at what point do we start bugging the pu-pin to start coughing up purge lists?
743  // - have a new parameter or just do it "one cycle before full"?
744  // - what if it doesn't -- when do we do the old-stlye scan & purge?
745  // - how do we do age-based purge and uvkeep purge?
746  // - they are really quite different -- and could run separately, registering
747  // files into a purge-candidate list. This has to be rechecked before the actual
748  // deletion -- eg, by comparing stat time of cinfo + doing the is-active / is-purge-protected.
749 
750  const DirState &root_ds = *m_fs_state.get_root();
751  const int n_calc_dirs = 1 + root_ds.m_here_usage.m_NDirectories + root_ds.m_recursive_subdir_usage.m_NDirectories;
752 #ifdef RM_DEBUG
753  const int n_pshot_dirs = root_ds.count_dirs_to_level(9999);
754  dprintf("purge dir count recursive=%d vs from_usage=%d\n", n_pshot_dirs, n_calc_dirs);
755 #endif
756  ps.m_dir_vec.reserve(n_calc_dirs);
757  ps.m_dir_vec.emplace_back( DirPurgeElement(root_ds, root_ds.m_here_usage, root_ds.m_recursive_subdir_usage, -1) );
758  fill_pshot_vec_children(root_ds, 0, ps.m_dir_vec, 9999);
759 
760  m_purge_task_active = true;
761 
762  struct PurgeDriverJob : public XrdJob
763  {
764  DataFsPurgeshot *m_purge_shot_ptr;
765 
766  PurgeDriverJob(DataFsPurgeshot *psp) :
767  XrdJob("XrdPfc::ResourceMonitor::PurgeDriver"),
768  m_purge_shot_ptr(psp)
769  {}
770 
771  void DoIt() override
772  {
773  Cache::ResMon().perform_purge_task(*m_purge_shot_ptr);
775 
776  delete m_purge_shot_ptr;
777  delete this;
778  }
779  };
780 
781  Cache::schedP->Schedule( new PurgeDriverJob(psp.release()) );
782 }
783 
784 namespace XrdPfc
785 {
787 }
788 
790 {
791  // BEWARE: Runs in a dedicated thread - is only to communicate back to the
792  // hear_beat() / data structs via the purge queues and condition variable.
793 
794  // const char *tpfx = "perform_purge_task ";
795 
796  {
798  m_purge_task_start = time(0);
799  }
800 
801  // For now, fall back to the old purge ... to be improved with:
802  // - new scan, following the DataFsPurgeshot;
803  // - usage of cinfo stat mtime for time of last access (touch already done at output);
804  // - use DirState* to report back purged files.
805  // Already changed to report back purged files --- but using the string / path variant.
806  OldStylePurgeDriver(ps); // In XrdPfcPurge.cc
807 }
808 
810 {
811  // Separated out so the purge_task can exit without post-checks.
812 
813  {
815  m_purge_task_end = time(0);
816  m_purge_task_complete = true;
818  }
820 }
821 
822 //==============================================================================
823 // Main thread function, do initial test, then enter heart_beat().
824 //==============================================================================
825 
827 {
828  // setup for in-scan -- this is called from initial setup.
829  MutexHolder _lck(m_dir_scan_mutex);
830  m_dir_scan_in_progress = true;
831 }
832 
834 {
835  const char *tpfx = "main_thread_function ";
836  {
837  time_t is_start = time(0);
838  m_fs_state.init_stat_reset_times(is_start);
839  TRACE(Info, tpfx << "Stating initial directory scan.");
840 
841  if ( ! perform_initial_scan()) {
842  TRACE(Error, tpfx << "Initial directory scan has failed. This is a terminal error, aborting.")
843  _exit(1);
844  }
845  // Reset of m_dir_scan_in_progress is done in perform_initial_scan()
846 
847  time_t is_duration = time(0) - is_start;
848  TRACE(Info, tpfx << "Initial directory scan complete, duration=" << is_duration <<"s");
849 
850  // run first process queues
851  int n_proc_is = process_queues();
852  TRACE(Info, tpfx << "First process_queues finished, n_records=" << n_proc_is);
853 
854  // shrink queues if scan time was longer than 30s.
855  if (is_duration > 30 || n_proc_is > 3000)
856  {
857  m_file_open_q.shrink_read_queue();
858  m_file_update_stats_q.shrink_read_queue();
859  m_file_close_q.shrink_read_queue();
860  m_file_purge_q1.shrink_read_queue();
861  m_file_purge_q2.shrink_read_queue();
862  m_file_purge_q3.shrink_read_queue();
863  }
864  }
865  heart_beat();
866 }
867 
868 //==============================================================================
869 // Old prototype from Cache / Purge, now to go into heart_beat() here, above.
870 //==============================================================================
871 
873 {
874  // static const char *trc_pfx = "ResourceMonitorHeartBeat() ";
875 
876  // Pause before initial run
877  sleep(1);
878 
879  // XXXX Setup initial / constant stats (total RAM, total disk, ???)
880 
883 
884  S.Lock();
885 
887 
889 
890  S.UnLock();
891 
892  // XXXX Schedule initial disk scan, time it!
893  //
894  // TRACE(Info, trc_pfx << "scheduling intial disk scan.");
895  // schedP->Schedule( new ScanAndPurgeJob("XrdPfc::ScanAndPurge") );
896  //
897  // bool scan_and_purge_running = true;
898 
899  // XXXX Could we really hold last-usage for all files in memory?
900 
901  // XXXX Think how to handle disk-full, scan/purge not finishing:
902  // - start dropping things out of write queue, but only when RAM gets near full;
903  // - monitoring this then becomes a high-priority job, inner loop with sleep of,
904  // say, 5 or 10 seconds.
905 
906  while (true)
907  {
908  time_t heartbeat_start = time(0);
909 
910  // TRACE(Info, trc_pfx << "HeartBeat starting ...");
911 
912  // if sumary monitoring configured, pupulate OucCacheStats:
913  S.Lock();
914 
915  // - available / used disk space (files usage calculated elsewhere (maybe))
916 
917  // - RAM usage
918  /* XXXX From Cache
919  { XrdSysMutexHelper lck(&m_RAM_mutex);
920  X.MemUsed = m_RAM_used;
921  X.MemWriteQ = m_RAM_write_queue;
922  }
923  */
924 
925  // - files opened / closed etc
926 
927  // do estimate of available space
928  S.UnLock();
929 
930  // if needed, schedule purge in a different thread.
931  // purge is:
932  // - deep scan + gather FSPurgeState
933  // - actual purge
934  //
935  // this thread can continue running and, if needed, stop writing to disk
936  // if purge is taking too long.
937 
938  // think how data is passed / synchronized between this and purge thread
939 
940  // !!!! think how stat collection is done and propgated upwards;
941  // until now it was done once per purge-interval.
942  // now stats will be added up more often, but purge will be done
943  // only occasionally.
944  // also, do we report cumulative values or deltas? cumulative should
945  // be easier and consistent with summary data.
946  // still, some are state - like disk usage, num of files.
947 
948  // Do we take care of directories that need to be newly added into DirState hierarchy?
949  // I.e., when user creates new directories and these are covered by either full
950  // spec or by root + depth declaration.
951 
952  int heartbeat_duration = time(0) - heartbeat_start;
953 
954  // TRACE(Info, trc_pfx << "HeartBeat finished, heartbeat_duration " << heartbeat_duration);
955 
956  // int sleep_time = m_fs_state..m_purgeInterval - heartbeat_duration;
957  int sleep_time = 60 - heartbeat_duration;
958  if (sleep_time > 0)
959  {
960  sleep(sleep_time);
961  }
962  }
963 }
int DoIt(int argpnt, int argc, char **argv, bool singleshot)
Definition: XrdAccTest.cc:262
static void child()
#define TRACE_Debug
Definition: XrdCmsTrace.hh:37
#define TRACE_Info
#define XrdOssOK
Definition: XrdOss.hh:50
#define dprintf(...)
void Proto_ResourceMonitorHeartBeat()
#define TRACE_INT(act, x)
Definition: XrdPfcTrace.hh:52
bool Debug
@ Error
#define TRACE(act, x)
Definition: XrdTrace.hh:63
Definition: XrdJob.hh:43
virtual int Opendir(const char *path, XrdOucEnv &env)
Definition: XrdOss.hh:79
long long Total
Definition: XrdOssVS.hh:90
long long Free
Definition: XrdOssVS.hh:91
virtual XrdOssDF * newDir(const char *tident)=0
virtual int StatVS(XrdOssVSInfo *vsP, const char *sname=0, int updt=0)
Definition: XrdOss.cc:117
virtual int Unlink(const char *path, int Opts=0, XrdOucEnv *envP=0)=0
struct XrdOucCacheStats::CacheStats X
XrdOucCacheStats Statistics
Definition: XrdOucCache.hh:686
PurgePin * GetPurgePin() const
Definition: XrdPfc.hh:272
static const Configuration & Conf()
Definition: XrdPfc.cc:134
XrdSysTrace * GetTrace()
Definition: XrdPfc.hh:283
static ResourceMonitor & ResMon()
Definition: XrdPfc.cc:135
void ClearPurgeProtectedSet()
Definition: XrdPfc.cc:684
static Cache & GetInstance()
Singleton access.
Definition: XrdPfc.cc:132
static XrdScheduler * schedP
Definition: XrdPfc.hh:290
long long WritesSinceLastCall()
Definition: XrdPfc.cc:320
void AddUp(const DirStats &s)
Definition: XrdPfcStats.hh:192
long long m_StBlocksRemoved
Definition: XrdPfcStats.hh:149
std::vector< std::string > m_current_dirs
XrdOucEnv & default_env()
void slurp_dir_ll(XrdOssDF &dh, int dir_level, const char *path, const char *trc_pfx)
bool begin_traversal(DirState *root, const char *root_path)
std::set< std::string > m_protected_top_dirs
bool cd_down(const std::string &dir_name)
std::map< std::string, FilePairStat > m_current_files
Status of cached file. Can be read from and written into a binary file.
Definition: XrdPfcInfo.hh:41
virtual bool CallPeriodically()
void CrossCheckIfScanIsInProgress(const std::string &lfn, XrdSysCondVar &cond)
void perform_purge_check(bool purge_cold_files, int tl)
void fill_sshot_vec_children(const DirState &parent_ds, int parent_idx, std::vector< DirStateElement > &vec, int max_depth)
void perform_purge_task(DataFsPurgeshot &ps)
void scan_dir_and_recurse(FsTraversal &fst)
void fill_pshot_vec_children(const DirState &parent_ds, int parent_idx, std::vector< DirPurgeElement > &vec, int max_depth)
void Schedule(XrdJob *jp)
@ Warning
Definition: XrdPfc.hh:41
void OldStylePurgeDriver(DataFsPurgeshot &ps)
Definition: XrdPfcPurge.cc:99
std::function< int(const std::string &)> unlink_func
Contains parameters configurable from the xrootd config file.
Definition: XrdPfc.hh:64
long long m_RamAbsAvailable
available from configuration
Definition: XrdPfc.hh:109
long long m_diskTotalSpace
total disk space on configured partition or oss space
Definition: XrdPfc.hh:91
long long m_fileUsageMax
cache purge - files usage maximum
Definition: XrdPfc.hh:96
long long m_fileUsageBaseline
cache purge - files usage baseline
Definition: XrdPfc.hh:94
int m_dirStatsStoreDepth
depth to which statistics should be collected
Definition: XrdPfc.hh:106
long long m_diskUsageHWM
cache purge - disk usage high water mark
Definition: XrdPfc.hh:93
bool are_file_usage_limits_set() const
Definition: XrdPfc.hh:67
long long m_fileUsageNominal
cache purge - files usage nominal
Definition: XrdPfc.hh:95
int m_purgeAgeBasedPeriod
peform cold file / uvkeep purge every this many purge cycles
Definition: XrdPfc.hh:99
long long m_diskUsageLWM
cache purge - disk usage low water mark
Definition: XrdPfc.hh:92
int m_dirStatsInterval
time between resource monitor statistics dump in seconds
Definition: XrdPfc.hh:104
bool is_age_based_purge_in_effect() const
Definition: XrdPfc.hh:68
int m_purgeInterval
sleep interval between cache purges
Definition: XrdPfc.hh:97
bool is_dir_stat_reporting_on() const
Definition: XrdPfc.hh:70
std::vector< DirPurgeElement > m_dir_vec
void write_json_file(const std::string &fname, XrdOss &oss, bool include_preamble)
std::vector< DirStateElement > m_dir_states
void reset_stats(time_t last_update)
void dump_recursively(int max_depth) const
void init_stat_reset_times(time_t t)
void update_stats_and_usages(time_t last_update, bool purge_empty_dirs, unlink_func unlink_foo)
DirState * find_dirstate_for_lfn(const std::string &lfn, DirState **last_existing_dir=nullptr)
void reset_sshot_stats(time_t last_update)
DirUsage m_recursive_subdir_usage
DirState * get_parent()
int count_dirs_to_level(int max_depth) const
DirState * find_path(const std::string &path, int max_depth, bool parse_as_lfn, bool create_subdirs, DirState **last_existing_dir=nullptr)
void upward_propagate_initial_scan_usages()