3 #define MAX_COMPARE_NAME 100
5 #define MAX_CMDLINE 1024
7 // the rates we are going to send to netdata
8 // will have this detail
10 // 1 will send just integer parts to netdata
11 // 100 will send 2 decimal points
12 // 1000 will send 3 decimal points
14 #define RATES_DETAIL 10000ULL
17 pid_t pid_max = 32768;
21 unsigned long long global_iterations_counter = 1;
22 unsigned long long file_counter = 0;
23 int proc_pid_cmdline_is_needed = 0;
24 int include_exited_childs = 1;
25 char *host_prefix = "";
26 char *config_dir = CONFIG_DIR;
28 pid_t *all_pids_sortlist = NULL;
30 // will be automatically set to 1, if guest values are collected
31 int show_guest_time = 0;
32 int show_guest_time_old = 0;
34 int enable_guest_charts = 0;
35 int enable_file_charts = 1;
37 // ----------------------------------------------------------------------------
39 void netdata_cleanup_and_exit(int ret) {
44 // ----------------------------------------------------------------------------
46 // to retrieve settings of the system
48 long get_system_cpus(void) {
53 char filename[FILENAME_MAX + 1];
54 snprintfz(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
56 ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
59 ff = procfile_readall(ff);
64 for(i = 0; i < procfile_lines(ff); i++) {
65 if(!procfile_linewords(ff, i)) continue;
67 if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
70 if(processors < 1) processors = 1;
76 pid_t get_system_pid_max(void) {
80 char filename[FILENAME_MAX + 1];
81 snprintfz(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
82 ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
85 ff = procfile_readall(ff);
89 mpid = (pid_t)atoi(procfile_lineword(ff, 0, 0));
90 if(!mpid) mpid = 32768;
96 // ----------------------------------------------------------------------------
98 // target is the structure that process data are aggregated
101 char compare[MAX_COMPARE_NAME + 1];
102 uint32_t comparehash;
105 char id[MAX_NAME + 1];
108 char name[MAX_NAME + 1];
113 unsigned long long minflt;
114 unsigned long long cminflt;
115 unsigned long long majflt;
116 unsigned long long cmajflt;
117 unsigned long long utime;
118 unsigned long long stime;
119 unsigned long long gtime;
120 unsigned long long cutime;
121 unsigned long long cstime;
122 unsigned long long cgtime;
123 unsigned long long num_threads;
124 unsigned long long rss;
126 unsigned long long statm_size;
127 unsigned long long statm_resident;
128 unsigned long long statm_share;
129 unsigned long long statm_text;
130 unsigned long long statm_lib;
131 unsigned long long statm_data;
132 unsigned long long statm_dirty;
134 unsigned long long io_logical_bytes_read;
135 unsigned long long io_logical_bytes_written;
136 unsigned long long io_read_calls;
137 unsigned long long io_write_calls;
138 unsigned long long io_storage_bytes_read;
139 unsigned long long io_storage_bytes_written;
140 unsigned long long io_cancelled_write_bytes;
143 unsigned long long openfiles;
144 unsigned long long openpipes;
145 unsigned long long opensockets;
146 unsigned long long openinotifies;
147 unsigned long long openeventfds;
148 unsigned long long opentimerfds;
149 unsigned long long opensignalfds;
150 unsigned long long openeventpolls;
151 unsigned long long openother;
153 unsigned long processes; // how many processes have been merged to this
154 int exposed; // if set, we have sent this to netdata
155 int hidden; // if set, we set the hidden flag on the dimension
158 int starts_with; // if set, the compare string matches only the
159 // beginning of the command
161 struct target *target; // the one that will be reported to netdata
166 // ----------------------------------------------------------------------------
168 // aggregate all processes in groups, to have a limited number of dimensions
170 struct target *apps_groups_root_target = NULL;
171 struct target *apps_groups_default_target = NULL;
172 long apps_groups_targets = 0;
174 struct target *users_root_target = NULL;
175 struct target *groups_root_target = NULL;
177 struct target *get_users_target(uid_t uid)
180 for(w = users_root_target ; w ; w = w->next)
181 if(w->uid == uid) return w;
183 w = callocz(sizeof(struct target), 1);
184 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", uid);
185 w->comparehash = simple_hash(w->compare);
186 w->comparelen = strlen(w->compare);
188 snprintfz(w->id, MAX_NAME, "%u", uid);
189 w->idhash = simple_hash(w->id);
191 struct passwd *pw = getpwuid(uid);
193 snprintfz(w->name, MAX_NAME, "%u", uid);
195 snprintfz(w->name, MAX_NAME, "%s", pw->pw_name);
197 netdata_fix_chart_name(w->name);
201 w->next = users_root_target;
202 users_root_target = w;
205 fprintf(stderr, "apps.plugin: added uid %u ('%s') target\n", w->uid, w->name);
210 struct target *get_groups_target(gid_t gid)
213 for(w = groups_root_target ; w ; w = w->next)
214 if(w->gid == gid) return w;
216 w = callocz(sizeof(struct target), 1);
217 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", gid);
218 w->comparehash = simple_hash(w->compare);
219 w->comparelen = strlen(w->compare);
221 snprintfz(w->id, MAX_NAME, "%u", gid);
222 w->idhash = simple_hash(w->id);
224 struct group *gr = getgrgid(gid);
226 snprintfz(w->name, MAX_NAME, "%u", gid);
228 snprintfz(w->name, MAX_NAME, "%s", gr->gr_name);
230 netdata_fix_chart_name(w->name);
234 w->next = groups_root_target;
235 groups_root_target = w;
238 fprintf(stderr, "apps.plugin: added gid %u ('%s') target\n", w->gid, w->name);
243 // find or create a new target
244 // there are targets that are just aggregated to other target (the second argument)
245 struct target *get_apps_groups_target(const char *id, struct target *target) {
246 int tdebug = 0, thidden = 0, ends_with = 0;
247 const char *nid = id;
249 while(nid[0] == '-' || nid[0] == '+' || nid[0] == '*') {
250 if(nid[0] == '-') thidden = 1;
251 if(nid[0] == '+') tdebug = 1;
252 if(nid[0] == '*') ends_with = 1;
255 uint32_t hash = simple_hash(id);
257 struct target *w, *last = apps_groups_root_target;
258 for(w = apps_groups_root_target ; w ; w = w->next) {
259 if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
265 w = callocz(sizeof(struct target), 1);
266 strncpyz(w->id, nid, MAX_NAME);
267 w->idhash = simple_hash(w->id);
269 strncpyz(w->name, nid, MAX_NAME);
271 strncpyz(w->compare, nid, MAX_COMPARE_NAME);
272 size_t len = strlen(w->compare);
273 if(w->compare[len - 1] == '*') {
274 w->compare[len - 1] = '\0';
277 w->ends_with = ends_with;
279 if(w->starts_with && w->ends_with)
280 proc_pid_cmdline_is_needed = 1;
282 w->comparehash = simple_hash(w->compare);
283 w->comparelen = strlen(w->compare);
289 // append it, to maintain the order in apps_groups.conf
290 if(last) last->next = w;
291 else apps_groups_root_target = w;
294 fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
296 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
297 , w->target?w->target->id:w->id
298 , (w->hidden)?"hidden":"-"
299 , (w->debug)?"debug":"-"
305 // read the apps_groups.conf file
306 int read_apps_groups_conf(const char *name)
308 char filename[FILENAME_MAX + 1];
310 snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
313 fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
315 // ----------------------------------------
317 procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT);
320 procfile_set_quotes(ff, "'\"");
322 ff = procfile_readall(ff);
326 unsigned long line, lines = procfile_lines(ff);
328 for(line = 0; line < lines ;line++) {
329 unsigned long word, words = procfile_linewords(ff, line);
330 struct target *w = NULL;
332 char *t = procfile_lineword(ff, line, 0);
333 if(!t || !*t) continue;
335 for(word = 0; word < words ;word++) {
336 char *s = procfile_lineword(ff, line, word);
337 if(!s || !*s) continue;
342 struct target *n = get_apps_groups_target(s, w);
344 error("Cannot create target '%s' (line %lu, word %lu)", s, line, word);
352 int tdebug = 0, thidden = 0;
354 while(t[0] == '-' || t[0] == '+') {
355 if(t[0] == '-') thidden = 1;
356 if(t[0] == '+') tdebug = 1;
360 strncpyz(w->name, t, MAX_NAME);
365 fprintf(stderr, "apps.plugin: AGGREGATION TARGET NAME '%s' on ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
368 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
369 , w->target?w->target->id:w->id
370 , (w->hidden)?"hidden":"-"
371 , (w->debug)?"debug":"-"
378 apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
379 if(!apps_groups_default_target)
380 error("Cannot create default target");
382 strncpyz(apps_groups_default_target->name, "other", MAX_NAME);
388 // ----------------------------------------------------------------------------
389 // data to store for each pid
392 #define PID_LOG_IO 0x00000001
393 #define PID_LOG_STATM 0x00000002
394 #define PID_LOG_CMDLINE 0x00000004
395 #define PID_LOG_FDS 0x00000008
396 #define PID_LOG_STAT 0x00000010
400 char comm[MAX_COMPARE_NAME + 1];
401 char cmdline[MAX_CMDLINE + 1];
413 // these are raw values collected
414 unsigned long long minflt_raw;
415 unsigned long long cminflt_raw;
416 unsigned long long majflt_raw;
417 unsigned long long cmajflt_raw;
418 unsigned long long utime_raw;
419 unsigned long long stime_raw;
420 unsigned long long gtime_raw; // guest_time
421 unsigned long long cutime_raw;
422 unsigned long long cstime_raw;
423 unsigned long long cgtime_raw; // cguest_time
426 unsigned long long minflt;
427 unsigned long long cminflt;
428 unsigned long long majflt;
429 unsigned long long cmajflt;
430 unsigned long long utime;
431 unsigned long long stime;
432 unsigned long long gtime;
433 unsigned long long cutime;
434 unsigned long long cstime;
435 unsigned long long cgtime;
440 // int64_t itrealvalue;
441 // unsigned long long starttime;
442 // unsigned long long vsize;
443 unsigned long long rss;
444 // unsigned long long rsslim;
445 // unsigned long long starcode;
446 // unsigned long long endcode;
447 // unsigned long long startstack;
448 // unsigned long long kstkesp;
449 // unsigned long long kstkeip;
452 // uint64_t sigignore;
453 // uint64_t sigcatch;
457 // int32_t exit_signal;
458 // int32_t processor;
459 // uint32_t rt_priority;
461 // unsigned long long delayacct_blkio_ticks;
466 unsigned long long statm_size;
467 unsigned long long statm_resident;
468 unsigned long long statm_share;
469 unsigned long long statm_text;
470 unsigned long long statm_lib;
471 unsigned long long statm_data;
472 unsigned long long statm_dirty;
474 unsigned long long io_logical_bytes_read_raw;
475 unsigned long long io_logical_bytes_written_raw;
476 unsigned long long io_read_calls_raw;
477 unsigned long long io_write_calls_raw;
478 unsigned long long io_storage_bytes_read_raw;
479 unsigned long long io_storage_bytes_written_raw;
480 unsigned long long io_cancelled_write_bytes_raw;
482 unsigned long long io_logical_bytes_read;
483 unsigned long long io_logical_bytes_written;
484 unsigned long long io_read_calls;
485 unsigned long long io_write_calls;
486 unsigned long long io_storage_bytes_read;
487 unsigned long long io_storage_bytes_written;
488 unsigned long long io_cancelled_write_bytes;
490 int *fds; // array of fds it uses
491 int fds_size; // the size of the fds array
493 int children_count; // number of processes directly referencing this
494 int keep; // 1 when we need to keep this process in memory even after it exited
495 int keeploops; // increases by 1 every time keep is 1 and updated 0
496 int updated; // 1 when the process is currently running
497 int merged; // 1 when it has been merged to its parent
498 int new_entry; // 1 when this is a new process, just saw for the first time
499 int read; // 1 when we have already read this process for this iteration
500 int sortlist; // higher numbers = top on the process tree
501 // each process gets a unique number
503 struct target *target; // app_groups.conf targets
504 struct target *user_target; // uid based targets
505 struct target *group_target; // gid based targets
507 unsigned long long stat_collected_usec;
508 unsigned long long last_stat_collected_usec;
510 unsigned long long io_collected_usec;
511 unsigned long long last_io_collected_usec;
514 char *statm_filename;
516 char *cmdline_filename;
518 struct pid_stat *parent;
519 struct pid_stat *prev;
520 struct pid_stat *next;
521 } *root_of_pids = NULL, **all_pids;
523 long all_pids_count = 0;
525 struct pid_stat *get_pid_entry(pid_t pid) {
527 all_pids[pid]->new_entry = 0;
528 return all_pids[pid];
531 all_pids[pid] = callocz(sizeof(struct pid_stat), 1);
532 all_pids[pid]->fds = callocz(sizeof(int), 100);
533 all_pids[pid]->fds_size = 100;
535 if(root_of_pids) root_of_pids->prev = all_pids[pid];
536 all_pids[pid]->next = root_of_pids;
537 root_of_pids = all_pids[pid];
539 all_pids[pid]->pid = pid;
540 all_pids[pid]->new_entry = 1;
544 return all_pids[pid];
547 void del_pid_entry(pid_t pid) {
549 error("attempted to free pid %d that is not allocated.", pid);
554 fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
556 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
557 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
558 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
560 if(all_pids[pid]->fds) freez(all_pids[pid]->fds);
561 if(all_pids[pid]->stat_filename) freez(all_pids[pid]->stat_filename);
562 if(all_pids[pid]->statm_filename) freez(all_pids[pid]->statm_filename);
563 if(all_pids[pid]->io_filename) freez(all_pids[pid]->io_filename);
564 if(all_pids[pid]->cmdline_filename) freez(all_pids[pid]->cmdline_filename);
565 freez(all_pids[pid]);
567 all_pids[pid] = NULL;
572 // ----------------------------------------------------------------------------
573 // update pids from proc
575 int read_proc_pid_cmdline(struct pid_stat *p) {
577 if(unlikely(!p->cmdline_filename)) {
578 char filename[FILENAME_MAX + 1];
579 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", host_prefix, p->pid);
580 p->cmdline_filename = strdupz(filename);
583 int fd = open(p->cmdline_filename, O_RDONLY, 0666);
584 if(unlikely(fd == -1)) goto cleanup;
586 ssize_t i, bytes = read(fd, p->cmdline, MAX_CMDLINE);
589 if(unlikely(bytes < 0)) goto cleanup;
591 p->cmdline[bytes] = '\0';
592 for(i = 0; i < bytes ; i++)
593 if(unlikely(!p->cmdline[i])) p->cmdline[i] = ' ';
596 fprintf(stderr, "Read file '%s' contents: %s\n", p->cmdline_filename, p->cmdline);
601 // copy the command to the command line
602 strncpyz(p->cmdline, p->comm, MAX_CMDLINE);
606 int read_proc_pid_ownership(struct pid_stat *p) {
607 if(unlikely(!p->stat_filename)) {
608 error("pid %d does not have a stat_filename", p->pid);
612 // ----------------------------------------
616 if(stat(p->stat_filename, &st) != 0) {
617 error("Cannot stat file '%s'", p->stat_filename);
627 int read_proc_pid_stat(struct pid_stat *p) {
628 static procfile *ff = NULL;
630 if(unlikely(!p->stat_filename)) {
631 char filename[FILENAME_MAX + 1];
632 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
633 p->stat_filename = strdupz(filename);
636 int set_quotes = (!ff)?1:0;
638 ff = procfile_reopen(ff, p->stat_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
639 if(unlikely(!ff)) goto cleanup;
641 // if(set_quotes) procfile_set_quotes(ff, "()");
642 if(set_quotes) procfile_set_open_close(ff, "(", ")");
644 ff = procfile_readall(ff);
645 if(unlikely(!ff)) goto cleanup;
647 p->last_stat_collected_usec = p->stat_collected_usec;
648 p->stat_collected_usec = time_usec();
651 // p->pid = atol(procfile_lineword(ff, 0, 0+i));
653 strncpyz(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
655 // p->state = *(procfile_lineword(ff, 0, 2));
656 p->ppid = (int32_t) atol(procfile_lineword(ff, 0, 3));
657 // p->pgrp = atol(procfile_lineword(ff, 0, 4));
658 // p->session = atol(procfile_lineword(ff, 0, 5));
659 // p->tty_nr = atol(procfile_lineword(ff, 0, 6));
660 // p->tpgid = atol(procfile_lineword(ff, 0, 7));
661 // p->flags = strtoull(procfile_lineword(ff, 0, 8), NULL, 10);
663 unsigned long long last;
665 last = p->minflt_raw;
666 p->minflt_raw = strtoull(procfile_lineword(ff, 0, 9), NULL, 10);
667 p->minflt = (p->minflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
669 last = p->cminflt_raw;
670 p->cminflt_raw = strtoull(procfile_lineword(ff, 0, 10), NULL, 10);
671 p->cminflt = (p->cminflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
673 last = p->majflt_raw;
674 p->majflt_raw = strtoull(procfile_lineword(ff, 0, 11), NULL, 10);
675 p->majflt = (p->majflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
677 last = p->cmajflt_raw;
678 p->cmajflt_raw = strtoull(procfile_lineword(ff, 0, 12), NULL, 10);
679 p->cmajflt = (p->cmajflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
682 p->utime_raw = strtoull(procfile_lineword(ff, 0, 13), NULL, 10);
683 p->utime = (p->utime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
686 p->stime_raw = strtoull(procfile_lineword(ff, 0, 14), NULL, 10);
687 p->stime = (p->stime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
689 last = p->cutime_raw;
690 p->cutime_raw = strtoull(procfile_lineword(ff, 0, 15), NULL, 10);
691 p->cutime = (p->cutime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
693 last = p->cstime_raw;
694 p->cstime_raw = strtoull(procfile_lineword(ff, 0, 16), NULL, 10);
695 p->cstime = (p->cstime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
697 // p->priority = strtoull(procfile_lineword(ff, 0, 17), NULL, 10);
698 // p->nice = strtoull(procfile_lineword(ff, 0, 18), NULL, 10);
699 p->num_threads = (int32_t) atol(procfile_lineword(ff, 0, 19));
700 // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20), NULL, 10);
701 // p->starttime = strtoull(procfile_lineword(ff, 0, 21), NULL, 10);
702 // p->vsize = strtoull(procfile_lineword(ff, 0, 22), NULL, 10);
703 p->rss = strtoull(procfile_lineword(ff, 0, 23), NULL, 10);
704 // p->rsslim = strtoull(procfile_lineword(ff, 0, 24), NULL, 10);
705 // p->starcode = strtoull(procfile_lineword(ff, 0, 25), NULL, 10);
706 // p->endcode = strtoull(procfile_lineword(ff, 0, 26), NULL, 10);
707 // p->startstack = strtoull(procfile_lineword(ff, 0, 27), NULL, 10);
708 // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28), NULL, 10);
709 // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29), NULL, 10);
710 // p->signal = strtoull(procfile_lineword(ff, 0, 30), NULL, 10);
711 // p->blocked = strtoull(procfile_lineword(ff, 0, 31), NULL, 10);
712 // p->sigignore = strtoull(procfile_lineword(ff, 0, 32), NULL, 10);
713 // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33), NULL, 10);
714 // p->wchan = strtoull(procfile_lineword(ff, 0, 34), NULL, 10);
715 // p->nswap = strtoull(procfile_lineword(ff, 0, 35), NULL, 10);
716 // p->cnswap = strtoull(procfile_lineword(ff, 0, 36), NULL, 10);
717 // p->exit_signal = atol(procfile_lineword(ff, 0, 37));
718 // p->processor = atol(procfile_lineword(ff, 0, 38));
719 // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39), NULL, 10);
720 // p->policy = strtoul(procfile_lineword(ff, 0, 40), NULL, 10);
721 // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41), NULL, 10);
723 if(enable_guest_charts) {
725 p->gtime_raw = strtoull(procfile_lineword(ff, 0, 42), NULL, 10);
726 p->gtime = (p->gtime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
728 last = p->cgtime_raw;
729 p->cgtime_raw = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
730 p->cgtime = (p->cgtime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
732 if (show_guest_time || p->gtime || p->cgtime) {
733 p->utime -= (p->utime >= p->gtime) ? p->gtime : p->utime;
734 p->cutime -= (p->cutime >= p->cgtime) ? p->cgtime : p->cutime;
739 if(unlikely(debug || (p->target && p->target->debug)))
740 fprintf(stderr, "apps.plugin: READ PROC/PID/STAT: %s/proc/%d/stat, process: '%s' on target '%s' (dt=%llu) VALUES: utime=%llu, stime=%llu, cutime=%llu, cstime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu, threads=%d\n", host_prefix, p->pid, p->comm, (p->target)?p->target->name:"UNSET", p->stat_collected_usec - p->last_stat_collected_usec, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt, p->num_threads);
742 if(unlikely(global_iterations_counter == 1)) {
773 int read_proc_pid_statm(struct pid_stat *p) {
774 static procfile *ff = NULL;
776 if(unlikely(!p->statm_filename)) {
777 char filename[FILENAME_MAX + 1];
778 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
779 p->statm_filename = strdupz(filename);
782 ff = procfile_reopen(ff, p->statm_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
783 if(unlikely(!ff)) goto cleanup;
785 ff = procfile_readall(ff);
786 if(unlikely(!ff)) goto cleanup;
790 p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
791 p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
792 p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
793 p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
794 p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
795 p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
796 p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
802 p->statm_resident = 0;
811 int read_proc_pid_io(struct pid_stat *p) {
812 static procfile *ff = NULL;
814 if(unlikely(!p->io_filename)) {
815 char filename[FILENAME_MAX + 1];
816 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
817 p->io_filename = strdupz(filename);
821 ff = procfile_reopen(ff, p->io_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
822 if(unlikely(!ff)) goto cleanup;
824 ff = procfile_readall(ff);
825 if(unlikely(!ff)) goto cleanup;
829 p->last_io_collected_usec = p->io_collected_usec;
830 p->io_collected_usec = time_usec();
832 unsigned long long last;
834 last = p->io_logical_bytes_read_raw;
835 p->io_logical_bytes_read_raw = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
836 p->io_logical_bytes_read = (p->io_logical_bytes_read_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
838 last = p->io_logical_bytes_written_raw;
839 p->io_logical_bytes_written_raw = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
840 p->io_logical_bytes_written = (p->io_logical_bytes_written_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
842 last = p->io_read_calls_raw;
843 p->io_read_calls_raw = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
844 p->io_read_calls = (p->io_read_calls_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
846 last = p->io_write_calls_raw;
847 p->io_write_calls_raw = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
848 p->io_write_calls = (p->io_write_calls_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
850 last = p->io_storage_bytes_read_raw;
851 p->io_storage_bytes_read_raw = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
852 p->io_storage_bytes_read = (p->io_storage_bytes_read_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
854 last = p->io_storage_bytes_written_raw;
855 p->io_storage_bytes_written_raw = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
856 p->io_storage_bytes_written = (p->io_storage_bytes_written_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
858 last = p->io_cancelled_write_bytes_raw;
859 p->io_cancelled_write_bytes_raw = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
860 p->io_cancelled_write_bytes = (p->io_cancelled_write_bytes_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
862 if(unlikely(global_iterations_counter == 1)) {
863 p->io_logical_bytes_read = 0;
864 p->io_logical_bytes_written = 0;
865 p->io_read_calls = 0;
866 p->io_write_calls = 0;
867 p->io_storage_bytes_read = 0;
868 p->io_storage_bytes_written = 0;
869 p->io_cancelled_write_bytes = 0;
875 p->io_logical_bytes_read = 0;
876 p->io_logical_bytes_written = 0;
877 p->io_read_calls = 0;
878 p->io_write_calls = 0;
879 p->io_storage_bytes_read = 0;
880 p->io_storage_bytes_written = 0;
881 p->io_cancelled_write_bytes = 0;
885 unsigned long long global_utime = 0;
886 unsigned long long global_stime = 0;
887 unsigned long long global_gtime = 0;
889 int read_proc_stat() {
890 static char filename[FILENAME_MAX + 1] = "";
891 static procfile *ff = NULL;
892 static unsigned long long utime_raw = 0, stime_raw = 0, gtime_raw = 0, gntime_raw = 0, ntime_raw = 0, collected_usec = 0, last_collected_usec = 0;
895 snprintfz(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
896 ff = procfile_open(filename, " \t:", PROCFILE_FLAG_DEFAULT);
897 if(unlikely(!ff)) goto cleanup;
900 ff = procfile_readall(ff);
901 if(unlikely(!ff)) goto cleanup;
903 last_collected_usec = collected_usec;
904 collected_usec = time_usec();
908 unsigned long long last;
911 utime_raw = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
912 global_utime = (utime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
914 // nice time, on user time
916 ntime_raw = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
917 global_utime += (ntime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
920 stime_raw = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
921 global_stime = (stime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
924 gtime_raw = strtoull(procfile_lineword(ff, 0, 10), NULL, 10);
925 global_gtime = (gtime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
927 if(enable_guest_charts) {
928 // guest nice time, on guest time
930 gntime_raw = strtoull(procfile_lineword(ff, 0, 11), NULL, 10);
931 global_gtime += (gntime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
933 // remove guest time from user time
934 global_utime -= (global_utime > global_gtime) ? global_gtime : global_utime;
937 if(unlikely(global_iterations_counter == 1)) {
953 // ----------------------------------------------------------------------------
955 // this is used to keep a global list of all open files of the system
956 // it is needed in order to calculate the unique files processes have open
958 #define FILE_DESCRIPTORS_INCREASE_STEP 100
960 struct file_descriptor {
962 #ifdef NETDATA_INTERNAL_CHECKS
964 #endif /* NETDATA_INTERNAL_CHECKS */
972 int all_files_len = 0;
973 int all_files_size = 0;
975 int file_descriptor_compare(void* a, void* b) {
976 #ifdef NETDATA_INTERNAL_CHECKS
977 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
978 error("Corrupted index data detected. Please report this.");
979 #endif /* NETDATA_INTERNAL_CHECKS */
981 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
984 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
988 return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
991 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
993 avl_tree all_files_index = {
995 file_descriptor_compare
998 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
999 struct file_descriptor tmp;
1000 tmp.hash = (hash)?hash:simple_hash(name);
1004 #ifdef NETDATA_INTERNAL_CHECKS
1005 tmp.magic = 0x0BADCAFE;
1006 #endif /* NETDATA_INTERNAL_CHECKS */
1008 return (struct file_descriptor *)avl_search(&all_files_index, (avl *) &tmp);
1011 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
1012 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
1014 #define FILETYPE_OTHER 0
1015 #define FILETYPE_FILE 1
1016 #define FILETYPE_PIPE 2
1017 #define FILETYPE_SOCKET 3
1018 #define FILETYPE_INOTIFY 4
1019 #define FILETYPE_EVENTFD 5
1020 #define FILETYPE_EVENTPOLL 6
1021 #define FILETYPE_TIMERFD 7
1022 #define FILETYPE_SIGNALFD 8
1024 void file_descriptor_not_used(int id)
1026 if(id > 0 && id < all_files_size) {
1028 #ifdef NETDATA_INTERNAL_CHECKS
1029 if(all_files[id].magic != 0x0BADCAFE) {
1030 error("Ignoring request to remove empty file id %d.", id);
1033 #endif /* NETDATA_INTERNAL_CHECKS */
1036 fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
1038 if(all_files[id].count > 0) {
1039 all_files[id].count--;
1041 if(!all_files[id].count) {
1043 fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
1045 file_descriptor_remove(&all_files[id]);
1046 #ifdef NETDATA_INTERNAL_CHECKS
1047 all_files[id].magic = 0x00000000;
1048 #endif /* NETDATA_INTERNAL_CHECKS */
1053 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
1055 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
1058 int file_descriptor_find_or_add(const char *name)
1060 static int last_pos = 0;
1061 uint32_t hash = simple_hash(name);
1064 fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
1066 struct file_descriptor *fd = file_descriptor_find(name, hash);
1070 fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
1077 // check we have enough memory to add it
1078 if(!all_files || all_files_len == all_files_size) {
1079 void *old = all_files;
1082 // there is no empty slot
1084 fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
1086 all_files = reallocz(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
1088 // if the address changed, we have to rebuild the index
1089 // since all pointers are now invalid
1090 if(old && old != (void *)all_files) {
1092 fprintf(stderr, "apps.plugin: >> re-indexing.\n");
1094 all_files_index.root = NULL;
1095 for(i = 0; i < all_files_size; i++) {
1096 if(!all_files[i].count) continue;
1097 file_descriptor_add(&all_files[i]);
1101 fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
1104 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
1105 all_files[i].count = 0;
1106 all_files[i].name = NULL;
1107 #ifdef NETDATA_INTERNAL_CHECKS
1108 all_files[i].magic = 0x00000000;
1109 #endif /* NETDATA_INTERNAL_CHECKS */
1110 all_files[i].pos = i;
1113 if(!all_files_size) all_files_len = 1;
1114 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
1118 fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
1120 // search for an empty slot
1122 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
1123 if(c >= all_files_size) c = 0;
1124 if(c == 0) continue;
1126 if(!all_files[c].count) {
1128 fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
1130 #ifdef NETDATA_INTERNAL_CHECKS
1131 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
1132 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
1133 #endif /* NETDATA_INTERNAL_CHECKS */
1136 fprintf(stderr, "apps.plugin: >> %s fd position %d for %s (last name: %s)\n", all_files[c].name?"re-using":"using", c, name, all_files[c].name);
1138 if(all_files[c].name) freez((void *)all_files[c].name);
1139 all_files[c].name = NULL;
1144 if(i == all_files_size) {
1145 fatal("We should find an empty slot, but there isn't any");
1150 fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
1154 // else we have an empty slot in 'c'
1157 if(name[0] == '/') type = FILETYPE_FILE;
1158 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
1159 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
1160 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
1161 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
1162 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
1163 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
1164 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
1165 else if(strncmp(name, "anon_inode:", 11) == 0) {
1167 fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
1169 type = FILETYPE_OTHER;
1173 fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
1175 type = FILETYPE_OTHER;
1178 all_files[c].name = strdupz(name);
1179 all_files[c].hash = hash;
1180 all_files[c].type = type;
1181 all_files[c].pos = c;
1182 all_files[c].count = 1;
1183 #ifdef NETDATA_INTERNAL_CHECKS
1184 all_files[c].magic = 0x0BADCAFE;
1185 #endif /* NETDATA_INTERNAL_CHECKS */
1186 file_descriptor_add(&all_files[c]);
1189 fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1194 int read_pid_file_descriptors(struct pid_stat *p) {
1195 char dirname[FILENAME_MAX+1];
1197 snprintfz(dirname, FILENAME_MAX, "%s/proc/%d/fd", host_prefix, p->pid);
1198 DIR *fds = opendir(dirname);
1202 char fdname[FILENAME_MAX + 1];
1203 char linkname[FILENAME_MAX + 1];
1205 // make the array negative
1206 for(c = 0 ; c < p->fds_size ; c++)
1207 p->fds[c] = -p->fds[c];
1209 while((de = readdir(fds))) {
1210 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
1213 // check if the fds array is small
1214 int fdid = atoi(de->d_name);
1215 if(fdid < 0) continue;
1216 if(fdid >= p->fds_size) {
1217 // it is small, extend it
1219 fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
1221 p->fds = reallocz(p->fds, (fdid + 100) * sizeof(int));
1223 fatal("Cannot re-allocate fds for %s", p->comm);
1227 // and initialize it
1228 for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
1229 p->fds_size = fdid + 100;
1232 if(p->fds[fdid] == 0) {
1233 // we don't know this fd, get it
1235 sprintf(fdname, "%s/proc/%d/fd/%s", host_prefix, p->pid, de->d_name);
1236 ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
1238 if(debug || (p->target && p->target->debug)) {
1239 if(debug || (p->target && p->target->debug))
1240 error("Cannot read link %s", fdname);
1247 // if another process already has this, we will get
1249 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1252 // else make it positive again, we need it
1253 // of course, the actual file may have changed, but we don't care so much
1254 // FIXME: we could compare the inode as returned by readdir dirent structure
1255 else p->fds[fdid] = -p->fds[fdid];
1259 // remove all the negative file descriptors
1260 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1261 file_descriptor_not_used(-p->fds[c]);
1270 // ----------------------------------------------------------------------------
1272 int print_process_and_parents(struct pid_stat *p, unsigned long long time) {
1273 char *prefix = "\\_ ";
1277 indent = print_process_and_parents(p->parent, p->stat_collected_usec);
1281 char buffer[indent + 1];
1284 for(i = 0; i < indent ;i++) buffer[i] = ' ';
1287 fprintf(stderr, " %s %s%s (%d %s %lld"
1292 , p->updated?"running":"exited"
1293 , (long long)p->stat_collected_usec - (long long)time
1296 if(p->utime) fprintf(stderr, " utime=%llu", p->utime);
1297 if(p->stime) fprintf(stderr, " stime=%llu", p->stime);
1298 if(p->gtime) fprintf(stderr, " gtime=%llu", p->gtime);
1299 if(p->cutime) fprintf(stderr, " cutime=%llu", p->cutime);
1300 if(p->cstime) fprintf(stderr, " cstime=%llu", p->cstime);
1301 if(p->cgtime) fprintf(stderr, " cgtime=%llu", p->cgtime);
1302 if(p->minflt) fprintf(stderr, " minflt=%llu", p->minflt);
1303 if(p->cminflt) fprintf(stderr, " cminflt=%llu", p->cminflt);
1304 if(p->majflt) fprintf(stderr, " majflt=%llu", p->majflt);
1305 if(p->cmajflt) fprintf(stderr, " cmajflt=%llu", p->cmajflt);
1306 fprintf(stderr, ")\n");
1311 void print_process_tree(struct pid_stat *p, char *msg) {
1313 fprintf(stderr, "%s: process %s (%d, %s) with parents:\n", msg, p->comm, p->pid, p->updated?"running":"exited");
1314 print_process_and_parents(p, p->stat_collected_usec);
1317 void find_lost_child_debug(struct pid_stat *pe, unsigned long long lost, int type) {
1319 struct pid_stat *p = NULL;
1321 for(p = root_of_pids; p ; p = p->next) {
1322 if(p == pe) continue;
1326 if(p->cminflt > lost) {
1327 fprintf(stderr, " > process %d (%s) could use the lost exited child minflt %llu of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
1333 if(p->cmajflt > lost) {
1334 fprintf(stderr, " > process %d (%s) could use the lost exited child majflt %llu of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
1340 if(p->cutime > lost) {
1341 fprintf(stderr, " > process %d (%s) could use the lost exited child utime %llu of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
1347 if(p->cstime > lost) {
1348 fprintf(stderr, " > process %d (%s) could use the lost exited child stime %llu of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
1354 if(p->cgtime > lost) {
1355 fprintf(stderr, " > process %d (%s) could use the lost exited child gtime %llu of process %d (%s)\n", p->pid, p->comm, lost, pe->pid, pe->comm);
1365 fprintf(stderr, " > cannot find any process to use the lost exited child minflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1369 fprintf(stderr, " > cannot find any process to use the lost exited child majflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1373 fprintf(stderr, " > cannot find any process to use the lost exited child utime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1377 fprintf(stderr, " > cannot find any process to use the lost exited child stime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1381 fprintf(stderr, " > cannot find any process to use the lost exited child gtime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1387 unsigned long long remove_exited_child_from_parent(unsigned long long *field, unsigned long long *pfield) {
1388 unsigned long long absorbed = 0;
1390 if(*field > *pfield) {
1391 absorbed += *pfield;
1404 void process_exited_processes() {
1407 for(p = root_of_pids; p ; p = p->next) {
1408 if(p->updated || !p->stat_collected_usec)
1411 struct pid_stat *pp = p->parent;
1413 unsigned long long utime = (p->utime_raw + p->cutime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1414 unsigned long long stime = (p->stime_raw + p->cstime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1415 unsigned long long gtime = (p->gtime_raw + p->cgtime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1416 unsigned long long minflt = (p->minflt_raw + p->cminflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1417 unsigned long long majflt = (p->majflt_raw + p->cmajflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1419 if(utime + stime + gtime + minflt + majflt == 0)
1422 if(unlikely(debug)) {
1424 fprintf(stderr, "Absorb %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1427 , p->updated?"running":"exited"
1434 print_process_tree(p, "Searching parents");
1437 for(pp = p->parent; pp ; pp = pp->parent) {
1438 if(!pp->updated) continue;
1440 unsigned long long absorbed;
1441 absorbed = remove_exited_child_from_parent(&utime, &pp->cutime);
1442 if(unlikely(debug && absorbed))
1443 fprintf(stderr, " > process %s (%d %s) absorbed %llu utime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, utime);
1445 absorbed = remove_exited_child_from_parent(&stime, &pp->cstime);
1446 if(unlikely(debug && absorbed))
1447 fprintf(stderr, " > process %s (%d %s) absorbed %llu stime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, stime);
1449 absorbed = remove_exited_child_from_parent(>ime, &pp->cgtime);
1450 if(unlikely(debug && absorbed))
1451 fprintf(stderr, " > process %s (%d %s) absorbed %llu gtime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, gtime);
1453 absorbed = remove_exited_child_from_parent(&minflt, &pp->cminflt);
1454 if(unlikely(debug && absorbed))
1455 fprintf(stderr, " > process %s (%d %s) absorbed %llu minflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, minflt);
1457 absorbed = remove_exited_child_from_parent(&majflt, &pp->cmajflt);
1458 if(unlikely(debug && absorbed))
1459 fprintf(stderr, " > process %s (%d %s) absorbed %llu majflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, majflt);
1462 if(unlikely(utime + stime + gtime + minflt + majflt > 0)) {
1463 if(unlikely(debug)) {
1464 if(utime) find_lost_child_debug(p, utime, 3);
1465 if(stime) find_lost_child_debug(p, stime, 4);
1466 if(gtime) find_lost_child_debug(p, gtime, 5);
1467 if(minflt) find_lost_child_debug(p, minflt, 1);
1468 if(majflt) find_lost_child_debug(p, majflt, 2);
1474 fprintf(stderr, " > remaining resources - KEEP - for another loop: %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1477 , p->updated?"running":"exited"
1485 for(pp = p->parent; pp ; pp = pp->parent) {
1486 if(pp->updated) break;
1490 fprintf(stderr, " > - KEEP - parent for another loop: %s (%d %s)\n"
1493 , pp->updated?"running":"exited"
1497 p->utime_raw = utime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1498 p->stime_raw = stime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1499 p->gtime_raw = gtime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1500 p->minflt_raw = minflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1501 p->majflt_raw = majflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1502 p->cutime_raw = p->cstime_raw = p->cgtime_raw = p->cminflt_raw = p->cmajflt_raw = 0;
1505 fprintf(stderr, "\n");
1507 else if(unlikely(debug)) {
1508 fprintf(stderr, " > totally absorbed - DONE - %s (%d %s)\n"
1511 , p->updated?"running":"exited"
1517 void link_all_processes_to_their_parents(void) {
1518 struct pid_stat *p, *pp;
1520 // link all children to their parents
1521 // and update children count on parents
1522 for(p = root_of_pids; p ; p = p->next) {
1523 // for each process found
1528 if(unlikely(!p->ppid)) {
1533 pp = all_pids[p->ppid];
1536 pp->children_count++;
1538 if(unlikely(debug || (p->target && p->target->debug)))
1539 fprintf(stderr, "apps.plugin: \tchild %d (%s, %s) on target '%s' has parent %d (%s, %s). Parent: utime=%llu, stime=%llu, gtime=%llu, minflt=%llu, majflt=%llu, cutime=%llu, cstime=%llu, cgtime=%llu, cminflt=%llu, cmajflt=%llu\n", p->pid, p->comm, p->updated?"running":"exited", (p->target)?p->target->name:"UNSET", pp->pid, pp->comm, pp->updated?"running":"exited", pp->utime, pp->stime, pp->gtime, pp->minflt, pp->majflt, pp->cutime, pp->cstime, pp->cgtime, pp->cminflt, pp->cmajflt);
1543 error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1548 // ----------------------------------------------------------------------------
1550 // 1. read all files in /proc
1551 // 2. for each numeric directory:
1552 // i. read /proc/pid/stat
1553 // ii. read /proc/pid/statm
1554 // iii. read /proc/pid/io (requires root access)
1555 // iii. read the entries in directory /proc/pid/fd (requires root access)
1557 // a. find or create a struct file_descriptor
1558 // b. cleanup any old/unused file_descriptors
1560 // after all these, some pids may be linked to targets, while others may not
1562 // in case of errors, only 1 every 1000 errors is printed
1563 // to avoid filling up all disk space
1564 // if debug is enabled, all errors are printed
1566 static int compar_pid(const void *pid1, const void *pid2) {
1568 struct pid_stat *p1 = all_pids[*((pid_t *)pid1)];
1569 struct pid_stat *p2 = all_pids[*((pid_t *)pid2)];
1571 if(p1->sortlist > p2->sortlist)
1577 static inline int managed_log(struct pid_stat *p, uint32_t log, int status) {
1578 if(unlikely(!status)) {
1579 // error("command failed log %u, errno %d", log, errno);
1581 if(unlikely(debug || errno != ENOENT)) {
1582 if(unlikely(debug || !(p->log_thrown & log))) {
1583 p->log_thrown |= log;
1586 error("Cannot process %s/proc/%d/io (command '%s')", host_prefix, p->pid, p->comm);
1590 error("Cannot process %s/proc/%d/statm (command '%s')", host_prefix, p->pid, p->comm);
1593 case PID_LOG_CMDLINE:
1594 error("Cannot process %s/proc/%d/cmdline (command '%s')", host_prefix, p->pid, p->comm);
1598 error("Cannot process entries in %s/proc/%d/fd (command '%s')", host_prefix, p->pid, p->comm);
1605 error("unhandled error for pid %d, command '%s'", p->pid, p->comm);
1612 else if(unlikely(p->log_thrown & log)) {
1613 // error("unsetting log %u on pid %d", log, p->pid);
1614 p->log_thrown &= ~log;
1620 void collect_data_for_pid(pid_t pid) {
1621 if(unlikely(pid <= 0 || pid > pid_max)) {
1622 error("Invalid pid %d read (expected 1 to %d). Ignoring process.", pid, pid_max);
1626 struct pid_stat *p = get_pid_entry(pid);
1627 if(unlikely(!p || p->read)) return;
1630 // fprintf(stderr, "Reading process %d (%s), sortlist %d\n", p->pid, p->comm, p->sortlist);
1632 // --------------------------------------------------------------------
1635 if(unlikely(!managed_log(p, PID_LOG_STAT, read_proc_pid_stat(p))))
1636 // there is no reason to proceed if we cannot get its status
1639 read_proc_pid_ownership(p);
1641 // check its parent pid
1642 if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
1643 error("Pid %d (command '%s') states invalid parent pid %d. Using 0.", pid, p->comm, p->ppid);
1647 // --------------------------------------------------------------------
1650 managed_log(p, PID_LOG_IO, read_proc_pid_io(p));
1652 // --------------------------------------------------------------------
1653 // /proc/<pid>/statm
1655 if(unlikely(!managed_log(p, PID_LOG_STATM, read_proc_pid_statm(p))))
1656 // there is no reason to proceed if we cannot get its memory status
1659 // --------------------------------------------------------------------
1662 // check if it is target
1663 // we do this only once, the first time this pid is loaded
1664 if(unlikely(p->new_entry)) {
1665 // /proc/<pid>/cmdline
1666 if(likely(proc_pid_cmdline_is_needed))
1667 managed_log(p, PID_LOG_CMDLINE, read_proc_pid_cmdline(p));
1670 fprintf(stderr, "apps.plugin: \tJust added %d (%s)\n", pid, p->comm);
1672 uint32_t hash = simple_hash(p->comm);
1673 size_t pclen = strlen(p->comm);
1676 for(w = apps_groups_root_target; w ; w = w->next) {
1677 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1679 // find it - 4 cases:
1680 // 1. the target is not a pattern
1681 // 2. the target has the prefix
1682 // 3. the target has the suffix
1683 // 4. the target is something inside cmdline
1684 if( (!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm))
1685 || (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen))
1686 || (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen]))
1687 || (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && strstr(p->cmdline, w->compare))
1689 if(w->target) p->target = w->target;
1692 if(debug || (p->target && p->target->debug))
1693 fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1700 // --------------------------------------------------------------------
1703 if(enable_file_charts)
1704 managed_log(p, PID_LOG_FDS, read_pid_file_descriptors(p));
1706 // --------------------------------------------------------------------
1709 if(unlikely(debug && include_exited_childs && all_pids_count && p->ppid && all_pids[p->ppid] && !all_pids[p->ppid]->read))
1710 fprintf(stderr, "Read process %d (%s) sortlisted %d, but its parent %d (%s) sortlisted %d, is not read\n", p->pid, p->comm, p->sortlist, all_pids[p->ppid]->pid, all_pids[p->ppid]->comm, all_pids[p->ppid]->sortlist);
1712 // mark it as updated
1718 int collect_data_for_all_processes_from_proc(void) {
1719 struct pid_stat *p = NULL;
1721 if(all_pids_count) {
1722 // read parents before childs
1723 // this is needed to prevent a situation where
1724 // a child is found running, but until we read
1725 // its parent, it has exited and its parent
1726 // has accumulated its resources
1729 for(p = root_of_pids; p ; p = p->next) {
1734 p->children_count = 0;
1737 all_pids_sortlist[slc++] = p->pid;
1740 if(unlikely(slc != all_pids_count)) {
1741 error("Internal error: I was thinking I had %ld processes in my arrays, but it seems there are more.", all_pids_count);
1742 all_pids_count = slc;
1745 if(include_exited_childs) {
1746 qsort((void *)all_pids_sortlist, all_pids_count, sizeof(pid_t), compar_pid);
1747 for(slc = 0; slc < all_pids_count; slc++)
1748 collect_data_for_pid(all_pids_sortlist[slc]);
1752 char dirname[FILENAME_MAX + 1];
1754 snprintfz(dirname, FILENAME_MAX, "%s/proc", host_prefix);
1755 DIR *dir = opendir(dirname);
1758 struct dirent *file = NULL;
1760 while((file = readdir(dir))) {
1761 char *endptr = file->d_name;
1762 pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
1764 // make sure we read a valid number
1765 if(unlikely(endptr == file->d_name || *endptr != '\0'))
1768 collect_data_for_pid(pid);
1772 // normally this is done
1773 // however we may have processes exited while we collected values
1774 // so let's find the exited ones
1775 // we do this by collecting the ownership of process
1776 // if we manage to get the ownership, the process still runs
1779 link_all_processes_to_their_parents();
1780 process_exited_processes();
1785 // ----------------------------------------------------------------------------
1786 // update statistics on the targets
1788 // 1. link all childs to their parents
1789 // 2. go from bottom to top, marking as merged all childs to their parents
1790 // this step links all parents without a target to the child target, if any
1791 // 3. link all top level processes (the ones not merged) to the default target
1792 // 4. go from top to bottom, linking all childs without a target, to their parent target
1793 // after this step, all processes have a target
1794 // [5. for each killed pid (updated = 0), remove its usage from its target]
1795 // 6. zero all apps_groups_targets
1796 // 7. concentrate all values on the apps_groups_targets
1797 // 8. remove all killed processes
1798 // 9. find the unique file count for each target
1799 // check: update_apps_groups_statistics()
1801 void cleanup_exited_pids(void) {
1803 struct pid_stat *p = NULL;
1805 for(p = root_of_pids; p ;) {
1806 if(!p->updated && (!p->keep || p->keeploops > 0)) {
1807 // fprintf(stderr, "\tEXITED %d %s [parent %d %s, target %s] utime=%llu, stime=%llu, gtime=%llu, cutime=%llu, cstime=%llu, cgtime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu\n", p->pid, p->comm, p->parent->pid, p->parent->comm, p->target->name, p->utime, p->stime, p->gtime, p->cutime, p->cstime, p->cgtime, p->minflt, p->majflt, p->cminflt, p->cmajflt);
1809 if(unlikely(debug && (p->keep || p->keeploops)))
1810 fprintf(stderr, " > CLEANUP cannot keep exited process %d (%s) anymore - removing it.\n", p->pid, p->comm);
1812 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1813 file_descriptor_not_used(p->fds[c]);
1822 if(unlikely(p->keep)) p->keeploops++;
1829 void apply_apps_groups_targets_inheritance(void) {
1830 struct pid_stat *p = NULL;
1832 // children that do not have a target
1833 // inherit their target from their parent
1834 int found = 1, loops = 0;
1836 if(unlikely(debug)) loops++;
1838 for(p = root_of_pids; p ; p = p->next) {
1839 // if this process does not have a target
1840 // and it has a parent
1841 // and its parent has a target
1842 // then, set the parent's target to this process
1843 if(unlikely(!p->target && p->parent && p->parent->target)) {
1844 p->target = p->parent->target;
1847 if(debug || (p->target && p->target->debug))
1848 fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s).\n", p->target->name, p->pid, p->comm, p->parent->pid, p->parent->comm);
1853 // find all the procs with 0 childs and merge them to their parents
1854 // repeat, until nothing more can be done.
1858 if(unlikely(debug)) loops++;
1861 for(p = root_of_pids; p ; p = p->next) {
1862 if(unlikely(!p->sortlist && !p->children_count))
1863 p->sortlist = sortlist++;
1865 // if this process does not have any children
1866 // and is not already merged
1868 // and its parent has children
1869 // and the target of this process and its parent is the same, or the parent does not have a target
1870 // and its parent is not init
1871 // then, mark them as merged.
1876 && p->parent->children_count
1877 && (p->target == p->parent->target || !p->parent->target)
1880 p->parent->children_count--;
1883 // the parent inherits the child's target, if it does not have a target itself
1884 if(unlikely(p->target && !p->parent->target)) {
1885 p->parent->target = p->target;
1887 if(debug || (p->target && p->target->debug))
1888 fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its child %d (%s).\n", p->target->name, p->parent->pid, p->parent->comm, p->pid, p->comm);
1896 fprintf(stderr, "apps.plugin: TARGET INHERITANCE: merged %d processes\n", found);
1899 // init goes always to default target
1901 all_pids[1]->target = apps_groups_default_target;
1903 // give a default target on all top level processes
1904 if(unlikely(debug)) loops++;
1905 for(p = root_of_pids; p ; p = p->next) {
1906 // if the process is not merged itself
1907 // then is is a top level process
1908 if(unlikely(!p->merged && !p->target))
1909 p->target = apps_groups_default_target;
1911 // make sure all processes have a sortlist
1912 if(unlikely(!p->sortlist))
1913 p->sortlist = sortlist++;
1917 all_pids[1]->sortlist = sortlist++;
1919 // give a target to all merged child processes
1922 if(unlikely(debug)) loops++;
1924 for(p = root_of_pids; p ; p = p->next) {
1925 if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
1926 p->target = p->parent->target;
1929 if(debug || (p->target && p->target->debug))
1930 fprintf(stderr, "apps.plugin: \t\tTARGET INHERITANCE: %s is inherited by %d (%s) from its parent %d (%s) at phase 2.\n", p->target->name, p->pid, p->comm, p->parent->pid, p->parent->comm);
1936 fprintf(stderr, "apps.plugin: apply_apps_groups_targets_inheritance() made %d loops on the process tree\n", loops);
1939 long zero_all_targets(struct target *root) {
1943 for (w = root; w ; w = w->next) {
1946 if(w->fds) freez(w->fds);
1964 w->statm_resident = 0;
1971 w->io_logical_bytes_read = 0;
1972 w->io_logical_bytes_written = 0;
1973 w->io_read_calls = 0;
1974 w->io_write_calls = 0;
1975 w->io_storage_bytes_read = 0;
1976 w->io_storage_bytes_written = 0;
1977 w->io_cancelled_write_bytes = 0;
1983 void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) {
1986 if(unlikely(!w->fds))
1987 w->fds = callocz(sizeof(int), (size_t) all_files_size);
1989 if(likely(p->updated)) {
1990 w->cutime += p->cutime;
1991 w->cstime += p->cstime;
1992 w->cgtime += p->cgtime;
1993 w->cminflt += p->cminflt;
1994 w->cmajflt += p->cmajflt;
1996 w->utime += p->utime;
1997 w->stime += p->stime;
1998 w->gtime += p->gtime;
1999 w->minflt += p->minflt;
2000 w->majflt += p->majflt;
2004 w->statm_size += p->statm_size;
2005 w->statm_resident += p->statm_resident;
2006 w->statm_share += p->statm_share;
2007 w->statm_text += p->statm_text;
2008 w->statm_lib += p->statm_lib;
2009 w->statm_data += p->statm_data;
2010 w->statm_dirty += p->statm_dirty;
2012 w->io_logical_bytes_read += p->io_logical_bytes_read;
2013 w->io_logical_bytes_written += p->io_logical_bytes_written;
2014 w->io_read_calls += p->io_read_calls;
2015 w->io_write_calls += p->io_write_calls;
2016 w->io_storage_bytes_read += p->io_storage_bytes_read;
2017 w->io_storage_bytes_written += p->io_storage_bytes_written;
2018 w->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
2021 w->num_threads += p->num_threads;
2023 if(likely(w->fds)) {
2025 for(c = 0; c < p->fds_size ;c++) {
2026 if(p->fds[c] == 0) continue;
2028 if(likely(p->fds[c] < all_files_size)) {
2029 if(w->fds) w->fds[p->fds[c]]++;
2032 error("Invalid fd number %d", p->fds[c]);
2036 if(unlikely(debug || w->debug))
2037 fprintf(stderr, "apps.plugin: \taggregating '%s' pid %d on target '%s' utime=%llu, stime=%llu, gtime=%llu, cutime=%llu, cstime=%llu, cgtime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu\n", p->comm, p->pid, w->name, p->utime, p->stime, p->gtime, p->cutime, p->cstime, p->cgtime, p->minflt, p->majflt, p->cminflt, p->cmajflt);
2041 void count_targets_fds(struct target *root) {
2045 for (w = root; w ; w = w->next) {
2046 if(!w->fds) continue;
2051 w->openinotifies = 0;
2052 w->openeventfds = 0;
2053 w->opentimerfds = 0;
2054 w->opensignalfds = 0;
2055 w->openeventpolls = 0;
2058 for(c = 1; c < all_files_size ;c++) {
2060 switch(all_files[c].type) {
2069 case FILETYPE_SOCKET:
2073 case FILETYPE_INOTIFY:
2077 case FILETYPE_EVENTFD:
2081 case FILETYPE_TIMERFD:
2085 case FILETYPE_SIGNALFD:
2089 case FILETYPE_EVENTPOLL:
2090 w->openeventpolls++;
2103 void calculate_netdata_statistics(void) {
2104 apply_apps_groups_targets_inheritance();
2106 zero_all_targets(users_root_target);
2107 zero_all_targets(groups_root_target);
2108 apps_groups_targets = zero_all_targets(apps_groups_root_target);
2110 // this has to be done, before the cleanup
2111 struct pid_stat *p = NULL;
2112 struct target *w = NULL, *o = NULL;
2114 // concentrate everything on the apps_groups_targets
2115 for(p = root_of_pids; p ; p = p->next) {
2117 // --------------------------------------------------------------------
2118 // apps_groups targets
2119 if(likely(p->target))
2120 aggregate_pid_on_target(p->target, p, NULL);
2122 error("pid %d %s was left without a target!", p->pid, p->comm);
2125 // --------------------------------------------------------------------
2128 if(likely(p->user_target && p->user_target->uid == p->uid))
2131 if(unlikely(debug && p->user_target))
2132 fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched user from %u (%s) to %u.\n", p->pid, p->comm, p->user_target->uid, p->user_target->name, p->uid);
2134 w = p->user_target = get_users_target(p->uid);
2138 aggregate_pid_on_target(w, p, o);
2140 error("pid %d %s was left without a user target!", p->pid, p->comm);
2143 // --------------------------------------------------------------------
2145 o = p->group_target;
2146 if(likely(p->group_target && p->group_target->gid == p->gid))
2147 w = p->group_target;
2149 if(unlikely(debug && p->group_target))
2150 fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched group from %u (%s) to %u.\n", p->pid, p->comm, p->group_target->gid, p->group_target->name, p->gid);
2152 w = p->group_target = get_groups_target(p->gid);
2156 aggregate_pid_on_target(w, p, o);
2158 error("pid %d %s was left without a group target!", p->pid, p->comm);
2162 count_targets_fds(apps_groups_root_target);
2163 count_targets_fds(users_root_target);
2164 count_targets_fds(groups_root_target);
2166 cleanup_exited_pids();
2169 // ----------------------------------------------------------------------------
2170 // update chart dimensions
2172 BUFFER *output = NULL;
2173 int print_calculated_number(char *str, calculated_number value) { (void)str; (void)value; return 0; }
2175 static inline void send_BEGIN(const char *type, const char *id, unsigned long long usec) {
2176 // fprintf(stdout, "BEGIN %s.%s %llu\n", type, id, usec);
2177 buffer_strcat(output, "BEGIN ");
2178 buffer_strcat(output, type);
2179 buffer_strcat(output, ".");
2180 buffer_strcat(output, id);
2181 buffer_strcat(output, " ");
2182 buffer_print_llu(output, usec);
2183 buffer_strcat(output, "\n");
2186 static inline void send_SET(const char *name, unsigned long long value) {
2187 // fprintf(stdout, "SET %s = %llu\n", name, value);
2188 buffer_strcat(output, "SET ");
2189 buffer_strcat(output, name);
2190 buffer_strcat(output, " = ");
2191 buffer_print_llu(output, value);
2192 buffer_strcat(output, "\n");
2195 static inline void send_END(void) {
2196 // fprintf(stdout, "END\n");
2197 buffer_strcat(output, "END\n");
2200 double utime_fix_ratio = 1.0, stime_fix_ratio = 1.0, gtime_fix_ratio = 1.0, cutime_fix_ratio = 1.0, cstime_fix_ratio = 1.0, cgtime_fix_ratio = 1.0;
2201 double minflt_fix_ratio = 1.0, majflt_fix_ratio = 1.0, cminflt_fix_ratio = 1.0, cmajflt_fix_ratio = 1.0;
2203 unsigned long long send_resource_usage_to_netdata() {
2204 static struct timeval last = { 0, 0 };
2205 static struct rusage me_last;
2210 unsigned long long usec;
2211 unsigned long long cpuuser;
2212 unsigned long long cpusyst;
2215 gettimeofday(&last, NULL);
2216 getrusage(RUSAGE_SELF, &me_last);
2218 // the first time, give a zero to allow
2219 // netdata calibrate to the current time
2220 // usec = update_every * 1000000ULL;
2226 gettimeofday(&now, NULL);
2227 getrusage(RUSAGE_SELF, &me);
2229 usec = usec_dt(&now, &last);
2230 cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
2231 cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
2233 bcopy(&now, &last, sizeof(struct timeval));
2234 bcopy(&me, &me_last, sizeof(struct rusage));
2237 buffer_sprintf(output,
2238 "BEGIN netdata.apps_cpu %llu\n"
2240 "SET system = %llu\n"
2242 "BEGIN netdata.apps_files %llu\n"
2243 "SET files = %llu\n"
2246 "SET targets = %ld\n"
2248 "BEGIN netdata.apps_fix %llu\n"
2249 "SET utime = %llu\n"
2250 "SET stime = %llu\n"
2251 "SET gtime = %llu\n"
2252 "SET minflt = %llu\n"
2253 "SET majflt = %llu\n"
2262 , apps_groups_targets
2264 , (unsigned long long)(utime_fix_ratio * 100 * RATES_DETAIL)
2265 , (unsigned long long)(stime_fix_ratio * 100 * RATES_DETAIL)
2266 , (unsigned long long)(gtime_fix_ratio * 100 * RATES_DETAIL)
2267 , (unsigned long long)(minflt_fix_ratio * 100 * RATES_DETAIL)
2268 , (unsigned long long)(majflt_fix_ratio * 100 * RATES_DETAIL)
2271 if(include_exited_childs)
2272 buffer_sprintf(output,
2273 "BEGIN netdata.apps_children_fix %llu\n"
2274 "SET cutime = %llu\n"
2275 "SET cstime = %llu\n"
2276 "SET cgtime = %llu\n"
2277 "SET cminflt = %llu\n"
2278 "SET cmajflt = %llu\n"
2281 , (unsigned long long)(cutime_fix_ratio * 100 * RATES_DETAIL)
2282 , (unsigned long long)(cstime_fix_ratio * 100 * RATES_DETAIL)
2283 , (unsigned long long)(cgtime_fix_ratio * 100 * RATES_DETAIL)
2284 , (unsigned long long)(cminflt_fix_ratio * 100 * RATES_DETAIL)
2285 , (unsigned long long)(cmajflt_fix_ratio * 100 * RATES_DETAIL)
2291 void normalize_data(struct target *root) {
2294 // childs processing introduces spikes
2295 // here we try to eliminate them by disabling childs processing either for specific dimensions
2296 // or entirely. Of course, either way, we disable it just a single iteration.
2298 unsigned long long max = processors * hz * RATES_DETAIL;
2299 unsigned long long utime = 0, cutime = 0, stime = 0, cstime = 0, gtime = 0, cgtime = 0, minflt = 0, cminflt = 0, majflt = 0, cmajflt = 0;
2301 if(global_utime > max) global_utime = max;
2302 if(global_stime > max) global_stime = max;
2303 if(global_gtime > max) global_gtime = max;
2305 for(w = root; w ; w = w->next) {
2306 if(w->target || (!w->processes && !w->exposed)) continue;
2311 cutime += w->cutime;
2312 cstime += w->cstime;
2313 cgtime += w->cgtime;
2315 minflt += w->minflt;
2316 majflt += w->majflt;
2317 cminflt += w->cminflt;
2318 cmajflt += w->cmajflt;
2321 if((global_utime || global_stime || global_gtime) && (utime || stime || gtime)) {
2322 if(global_utime + global_stime + global_gtime > utime + cutime + stime + cstime + gtime + cgtime) {
2323 // everything we collected fits
2329 cgtime_fix_ratio = 1.0; //(double)(global_utime + global_stime) / (double)(utime + cutime + stime + cstime);
2331 else if(global_utime + global_stime > utime + stime) {
2332 // childrens resources are too high
2333 // lower only the children resources
2336 gtime_fix_ratio = 1.0;
2339 cgtime_fix_ratio = (double)((global_utime + global_stime) - (utime + stime)) / (double)(cutime + cstime);
2342 // even running processes are unrealistic
2343 // zero the children resources
2344 // lower the running processes resources
2347 gtime_fix_ratio = (double)(global_utime + global_stime) / (double)(utime + stime);
2350 cgtime_fix_ratio = 0.0;
2359 cgtime_fix_ratio = 0.0;
2362 if(utime_fix_ratio > 1.0) utime_fix_ratio = 1.0;
2363 if(cutime_fix_ratio > 1.0) cutime_fix_ratio = 1.0;
2364 if(stime_fix_ratio > 1.0) stime_fix_ratio = 1.0;
2365 if(cstime_fix_ratio > 1.0) cstime_fix_ratio = 1.0;
2366 if(gtime_fix_ratio > 1.0) gtime_fix_ratio = 1.0;
2367 if(cgtime_fix_ratio > 1.0) cgtime_fix_ratio = 1.0;
2369 // if(utime_fix_ratio < 0.0) utime_fix_ratio = 0.0;
2370 // if(cutime_fix_ratio < 0.0) cutime_fix_ratio = 0.0;
2371 // if(stime_fix_ratio < 0.0) stime_fix_ratio = 0.0;
2372 // if(cstime_fix_ratio < 0.0) cstime_fix_ratio = 0.0;
2373 // if(gtime_fix_ratio < 0.0) gtime_fix_ratio = 0.0;
2374 // if(cgtime_fix_ratio < 0.0) cgtime_fix_ratio = 0.0;
2377 // we use cpu time to normalize page faults
2378 // the problem is that to find the proper max values
2379 // for page faults we have to parse /proc/vmstat
2380 // which is quite big to do it again (netdata does it already)
2382 // a better solution could be to somehow have netdata
2383 // do this normalization for us
2385 if(utime || stime || gtime)
2387 minflt_fix_ratio = (double)(utime * utime_fix_ratio + stime * stime_fix_ratio + gtime * gtime_fix_ratio) / (double)(utime + stime + gtime);
2390 majflt_fix_ratio = 1.0;
2392 if(cutime || cstime || cgtime)
2394 cminflt_fix_ratio = (double)(cutime * cutime_fix_ratio + cstime * cstime_fix_ratio + cgtime * cgtime_fix_ratio) / (double)(cutime + cstime + cgtime);
2397 cmajflt_fix_ratio = 1.0;
2401 if(unlikely(debug)) {
2403 "SYSTEM: u=%llu s=%llu g=%llu "
2404 "COLLECTED: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2405 "DELTA: u=%lld s=%lld g=%lld "
2406 "FIX: u=%0.2f s=%0.2f g=%0.2f cu=%0.2f cs=%0.2f cg=%0.2f "
2407 "FINALLY: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2418 , (long long)utime + (long long)cutime - (long long)global_utime
2419 , (long long)stime + (long long)cstime - (long long)global_stime
2420 , (long long)gtime + (long long)cgtime - (long long)global_gtime
2427 , (unsigned long long)(utime * utime_fix_ratio)
2428 , (unsigned long long)(stime * stime_fix_ratio)
2429 , (unsigned long long)(gtime * gtime_fix_ratio)
2430 , (unsigned long long)(cutime * cutime_fix_ratio)
2431 , (unsigned long long)(cstime * cstime_fix_ratio)
2432 , (unsigned long long)(cgtime * cgtime_fix_ratio)
2437 void send_collected_data_to_netdata(struct target *root, const char *type, unsigned long long usec) {
2440 send_BEGIN(type, "cpu", usec);
2441 for (w = root; w ; w = w->next) {
2442 if(unlikely(w->exposed))
2443 send_SET(w->name, (unsigned long long)(w->utime * utime_fix_ratio) + (unsigned long long)(w->stime * stime_fix_ratio) + (unsigned long long)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cutime * cutime_fix_ratio) + (unsigned long long)(w->cstime * cstime_fix_ratio) + (unsigned long long)(w->cgtime * cgtime_fix_ratio)):0ULL));
2447 send_BEGIN(type, "cpu_user", usec);
2448 for (w = root; w ; w = w->next) {
2449 if(unlikely(w->exposed))
2450 send_SET(w->name, (unsigned long long)(w->utime * utime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cutime * cutime_fix_ratio)):0ULL));
2454 send_BEGIN(type, "cpu_system", usec);
2455 for (w = root; w ; w = w->next) {
2456 if(unlikely(w->exposed))
2457 send_SET(w->name, (unsigned long long)(w->stime * stime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cstime * cstime_fix_ratio)):0ULL));
2461 if(show_guest_time) {
2462 send_BEGIN(type, "cpu_guest", usec);
2463 for (w = root; w ; w = w->next) {
2464 if(unlikely(w->exposed))
2465 send_SET(w->name, (unsigned long long)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cgtime * cgtime_fix_ratio)):0ULL));
2470 send_BEGIN(type, "threads", usec);
2471 for (w = root; w ; w = w->next) {
2472 if(unlikely(w->exposed))
2473 send_SET(w->name, w->num_threads);
2477 send_BEGIN(type, "processes", usec);
2478 for (w = root; w ; w = w->next) {
2479 if(unlikely(w->exposed))
2480 send_SET(w->name, w->processes);
2484 send_BEGIN(type, "mem", usec);
2485 for (w = root; w ; w = w->next) {
2486 if(unlikely(w->exposed))
2487 send_SET(w->name, (w->statm_resident > w->statm_share)?(w->statm_resident - w->statm_share):0ULL);
2491 send_BEGIN(type, "minor_faults", usec);
2492 for (w = root; w ; w = w->next) {
2493 if(unlikely(w->exposed))
2494 send_SET(w->name, (unsigned long long)(w->minflt * minflt_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cminflt * cminflt_fix_ratio)):0ULL));
2498 send_BEGIN(type, "major_faults", usec);
2499 for (w = root; w ; w = w->next) {
2500 if(unlikely(w->exposed))
2501 send_SET(w->name, (unsigned long long)(w->majflt * majflt_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cmajflt * cmajflt_fix_ratio)):0ULL));
2505 send_BEGIN(type, "lreads", usec);
2506 for (w = root; w ; w = w->next) {
2507 if(unlikely(w->exposed))
2508 send_SET(w->name, w->io_logical_bytes_read);
2512 send_BEGIN(type, "lwrites", usec);
2513 for (w = root; w ; w = w->next) {
2514 if(unlikely(w->exposed))
2515 send_SET(w->name, w->io_logical_bytes_written);
2519 send_BEGIN(type, "preads", usec);
2520 for (w = root; w ; w = w->next) {
2521 if(unlikely(w->exposed))
2522 send_SET(w->name, w->io_storage_bytes_read);
2526 send_BEGIN(type, "pwrites", usec);
2527 for (w = root; w ; w = w->next) {
2528 if(unlikely(w->exposed))
2529 send_SET(w->name, w->io_storage_bytes_written);
2533 if(enable_file_charts) {
2534 send_BEGIN(type, "files", usec);
2535 for (w = root; w; w = w->next) {
2536 if (unlikely(w->exposed))
2537 send_SET(w->name, w->openfiles);
2541 send_BEGIN(type, "sockets", usec);
2542 for (w = root; w; w = w->next) {
2543 if (unlikely(w->exposed))
2544 send_SET(w->name, w->opensockets);
2548 send_BEGIN(type, "pipes", usec);
2549 for (w = root; w; w = w->next) {
2550 if (unlikely(w->exposed))
2551 send_SET(w->name, w->openpipes);
2558 // ----------------------------------------------------------------------------
2559 // generate the charts
2561 void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title)
2564 int newly_added = 0;
2566 for(w = root ; w ; w = w->next) {
2567 if (w->target) continue;
2569 if (!w->exposed && w->processes) {
2572 if (debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
2576 // nothing more to show
2577 if(!newly_added && show_guest_time == show_guest_time_old) return;
2579 // we have something new to show
2580 // update the charts
2581 buffer_sprintf(output, "CHART %s.cpu '' '%s CPU Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu stacked 20001 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
2582 for (w = root; w ; w = w->next) {
2583 if(unlikely(w->exposed))
2584 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu %s\n", w->name, hz * RATES_DETAIL / 100, w->hidden ? "hidden" : "");
2587 buffer_sprintf(output, "CHART %s.mem '' '%s Dedicated Memory (w/o shared)' 'MB' mem %s.mem stacked 20003 %d\n", type, title, type, update_every);
2588 for (w = root; w ; w = w->next) {
2589 if(unlikely(w->exposed))
2590 buffer_sprintf(output, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
2593 buffer_sprintf(output, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20005 %d\n", type, title, type, update_every);
2594 for (w = root; w ; w = w->next) {
2595 if(unlikely(w->exposed))
2596 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2599 buffer_sprintf(output, "CHART %s.processes '' '%s Processes' 'processes' processes %s.processes stacked 20004 %d\n", type, title, type, update_every);
2600 for (w = root; w ; w = w->next) {
2601 if(unlikely(w->exposed))
2602 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2605 buffer_sprintf(output, "CHART %s.cpu_user '' '%s CPU User Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu_user stacked 20020 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
2606 for (w = root; w ; w = w->next) {
2607 if(unlikely(w->exposed))
2608 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2611 buffer_sprintf(output, "CHART %s.cpu_system '' '%s CPU System Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu_system stacked 20021 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
2612 for (w = root; w ; w = w->next) {
2613 if(unlikely(w->exposed))
2614 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2617 if(show_guest_time) {
2618 buffer_sprintf(output, "CHART %s.cpu_guest '' '%s CPU Guest Time (%d%% = %d core%s)' 'cpu time %%' cpu %s.cpu_system stacked 20022 %d\n", type, title, (processors * 100), processors, (processors > 1) ? "s" : "", type, update_every);
2619 for (w = root; w; w = w->next) {
2620 if(unlikely(w->exposed))
2621 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2625 buffer_sprintf(output, "CHART %s.major_faults '' '%s Major Page Faults (swap read)' 'page faults/s' swap %s.major_faults stacked 20010 %d\n", type, title, type, update_every);
2626 for (w = root; w ; w = w->next) {
2627 if(unlikely(w->exposed))
2628 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
2631 buffer_sprintf(output, "CHART %s.minor_faults '' '%s Minor Page Faults' 'page faults/s' mem %s.minor_faults stacked 20011 %d\n", type, title, type, update_every);
2632 for (w = root; w ; w = w->next) {
2633 if(unlikely(w->exposed))
2634 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
2637 buffer_sprintf(output, "CHART %s.lreads '' '%s Disk Logical Reads' 'kilobytes/s' disk %s.lreads stacked 20042 %d\n", type, title, type, update_every);
2638 for (w = root; w ; w = w->next) {
2639 if(unlikely(w->exposed))
2640 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2643 buffer_sprintf(output, "CHART %s.lwrites '' '%s I/O Logical Writes' 'kilobytes/s' disk %s.lwrites stacked 20042 %d\n", type, title, type, update_every);
2644 for (w = root; w ; w = w->next) {
2645 if(unlikely(w->exposed))
2646 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2649 buffer_sprintf(output, "CHART %s.preads '' '%s Disk Reads' 'kilobytes/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every);
2650 for (w = root; w ; w = w->next) {
2651 if(unlikely(w->exposed))
2652 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2655 buffer_sprintf(output, "CHART %s.pwrites '' '%s Disk Writes' 'kilobytes/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
2656 for (w = root; w ; w = w->next) {
2657 if(unlikely(w->exposed))
2658 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2661 if(enable_file_charts) {
2662 buffer_sprintf(output, "CHART %s.files '' '%s Open Files' 'open files' disk %s.files stacked 20050 %d\n", type,
2663 title, type, update_every);
2664 for (w = root; w; w = w->next) {
2665 if (unlikely(w->exposed))
2666 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2669 buffer_sprintf(output, "CHART %s.sockets '' '%s Open Sockets' 'open sockets' net %s.sockets stacked 20051 %d\n",
2670 type, title, type, update_every);
2671 for (w = root; w; w = w->next) {
2672 if (unlikely(w->exposed))
2673 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2676 buffer_sprintf(output, "CHART %s.pipes '' '%s Pipes' 'open pipes' processes %s.pipes stacked 20053 %d\n", type,
2677 title, type, update_every);
2678 for (w = root; w; w = w->next) {
2679 if (unlikely(w->exposed))
2680 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2686 // ----------------------------------------------------------------------------
2687 // parse command line arguments
2689 void parse_args(int argc, char **argv)
2694 for(i = 1; i < argc; i++) {
2696 int n = atoi(argv[i]);
2703 if(strcmp("debug", argv[i]) == 0) {
2705 // debug_flags = 0xffffffff;
2709 if(strcmp("no-childs", argv[i]) == 0 || strcmp("without-childs", argv[i]) == 0) {
2710 include_exited_childs = 0;
2714 if(strcmp("with-childs", argv[i]) == 0) {
2715 include_exited_childs = 1;
2719 if(strcmp("with-guest", argv[i]) == 0) {
2720 enable_guest_charts = 1;
2724 if(strcmp("no-guest", argv[i]) == 0 || strcmp("without-guest", argv[i]) == 0) {
2725 enable_guest_charts = 0;
2729 if(strcmp("with-files", argv[i]) == 0) {
2730 enable_file_charts = 1;
2734 if(strcmp("no-files", argv[i]) == 0 || strcmp("without-files", argv[i]) == 0) {
2735 enable_file_charts = 0;
2739 if(strcmp("-h", argv[i]) == 0 || strcmp("--help", argv[i]) == 0) {
2742 "(C) 2016 Costa Tsaousis"
2744 "This program is a data collector plugin for netdata.\n"
2746 "Valid command line options:\n"
2748 "SECONDS set the data collection frequency\n"
2750 "debug enable debugging (lot of output)\n"
2753 "without-childs enable / disable aggregating exited\n"
2754 " children resources into parents\n"
2755 " (default is enabled)\n"
2758 "without-guest enable / disable reporting guest charts\n"
2759 " (default is disabled)\n"
2762 "without-files enable / disable reporting files, sockets, pipes\n"
2763 " (default is enabled)\n"
2765 "NAME read apps_NAME.conf instead of\n"
2766 " apps_groups.conf\n"
2767 " (default NAME=groups)\n"
2777 error("Cannot understand option %s", argv[i]);
2781 if(freq > 0) update_every = freq;
2782 if(!name) name = "groups";
2784 if(read_apps_groups_conf(name)) {
2785 error("Cannot read process groups %s", name);
2790 int main(int argc, char **argv)
2792 // debug_flags = D_PROCFILE;
2794 // set the name for logging
2795 program_name = "apps.plugin";
2797 info("started on pid %d", getpid());
2799 // disable syslog for apps.plugin
2800 error_log_syslog = 0;
2802 // set errors flood protection to 100 logs per hour
2803 error_log_errors_per_period = 100;
2804 error_log_throttle_period = 3600;
2806 host_prefix = getenv("NETDATA_HOST_PREFIX");
2807 if(host_prefix == NULL) {
2808 // info("NETDATA_HOST_PREFIX is not passed from netdata");
2811 // else info("Found NETDATA_HOST_PREFIX='%s'", host_prefix);
2813 config_dir = getenv("NETDATA_CONFIG_DIR");
2814 if(config_dir == NULL) {
2815 // info("NETDATA_CONFIG_DIR is not passed from netdata");
2816 config_dir = CONFIG_DIR;
2818 // else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
2820 #ifdef NETDATA_INTERNAL_CHECKS
2821 if(debug_flags != 0) {
2822 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
2823 if(setrlimit(RLIMIT_CORE, &rl) != 0)
2824 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
2825 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
2827 #endif /* NETDATA_INTERNAL_CHECKS */
2829 procfile_adaptive_initial_allocation = 1;
2831 time_t started_t = time(NULL);
2834 pid_max = get_system_pid_max();
2835 processors = get_system_cpus();
2837 parse_args(argc, argv);
2839 all_pids_sortlist = callocz(sizeof(pid_t), (size_t)pid_max);
2840 all_pids = callocz(sizeof(struct pid_stat *), (size_t) pid_max);
2842 output = buffer_create(1024);
2843 buffer_sprintf(output,
2844 "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %1$d\n"
2845 "DIMENSION user '' incremental 1 1000\n"
2846 "DIMENSION system '' incremental 1 1000\n"
2847 "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %1$d\n"
2848 "DIMENSION files '' incremental 1 1\n"
2849 "DIMENSION pids '' absolute 1 1\n"
2850 "DIMENSION fds '' absolute 1 1\n"
2851 "DIMENSION targets '' absolute 1 1\n"
2852 "CHART netdata.apps_fix '' 'Apps Plugin Normalization Ratios' 'percentage' apps.plugin netdata.apps_fix line 140002 %1$d\n"
2853 "DIMENSION utime '' absolute 1 %2$llu\n"
2854 "DIMENSION stime '' absolute 1 %2$llu\n"
2855 "DIMENSION gtime '' absolute 1 %2$llu\n"
2856 "DIMENSION minflt '' absolute 1 %2$llu\n"
2857 "DIMENSION majflt '' absolute 1 %2$llu\n"
2862 if(include_exited_childs)
2863 buffer_sprintf(output,
2864 "CHART netdata.apps_children_fix '' 'Apps Plugin Exited Children Normalization Ratios' 'percentage' apps.plugin netdata.apps_children_fix line 140003 %1$d\n"
2865 "DIMENSION cutime '' absolute 1 %2$llu\n"
2866 "DIMENSION cstime '' absolute 1 %2$llu\n"
2867 "DIMENSION cgtime '' absolute 1 %2$llu\n"
2868 "DIMENSION cminflt '' absolute 1 %2$llu\n"
2869 "DIMENSION cmajflt '' absolute 1 %2$llu\n"
2874 #ifndef PROFILING_MODE
2875 unsigned long long sunext = (time(NULL) - (time(NULL) % update_every) + update_every) * 1000000ULL;
2876 unsigned long long sunow;
2877 #endif /* PROFILING_MODE */
2879 global_iterations_counter = 1;
2880 for(;1; global_iterations_counter++) {
2881 #ifndef PROFILING_MODE
2882 // delay until it is our time to run
2883 while((sunow = time_usec()) < sunext)
2884 sleep_usec(sunext - sunow);
2886 // find the next time we need to run
2887 while(time_usec() > sunext)
2888 sunext += update_every * 1000000ULL;
2889 #endif /* PROFILING_MODE */
2891 if(!collect_data_for_all_processes_from_proc()) {
2892 error("Cannot collect /proc data for running processes. Disabling apps.plugin...");
2893 printf("DISABLE\n");
2897 calculate_netdata_statistics();
2898 normalize_data(apps_groups_root_target);
2900 unsigned long long dt = send_resource_usage_to_netdata();
2902 // this is smart enough to show only newly added apps, when needed
2903 send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps");
2904 send_charts_updates_to_netdata(users_root_target, "users", "Users");
2905 send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups");
2907 send_collected_data_to_netdata(apps_groups_root_target, "apps", dt);
2908 send_collected_data_to_netdata(users_root_target, "users", dt);
2909 send_collected_data_to_netdata(groups_root_target, "groups", dt);
2911 show_guest_time_old = show_guest_time;
2913 //if(puts(buffer_tostring(output)) == EOF)
2914 if(write(STDOUT_FILENO, buffer_tostring(output), buffer_strlen(output)) == -1)
2915 fatal("Cannot send chart values to netdata.");
2918 buffer_flush(output);
2921 fprintf(stderr, "apps.plugin: done Loop No %llu\n", global_iterations_counter);
2923 current_t = time(NULL);
2925 #ifndef PROFILING_MODE
2926 // restart check (14400 seconds)
2927 if(current_t - started_t > 14400) exit(0);
2929 if(current_t - started_t > 10) exit(0);
2930 #endif /* PROFILING_MODE */