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);
66 for(i = 0; i < procfile_lines(ff); i++) {
67 if(!procfile_linewords(ff, i)) continue;
69 if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
72 if(processors < 1) processors = 1;
78 pid_t get_system_pid_max(void) {
82 char filename[FILENAME_MAX + 1];
83 snprintfz(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
84 ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
87 ff = procfile_readall(ff);
93 mpid = (pid_t)atoi(procfile_lineword(ff, 0, 0));
94 if(!mpid) mpid = 32768;
100 // ----------------------------------------------------------------------------
102 // target is the structure that process data are aggregated
105 char compare[MAX_COMPARE_NAME + 1];
106 uint32_t comparehash;
109 char id[MAX_NAME + 1];
112 char name[MAX_NAME + 1];
117 unsigned long long minflt;
118 unsigned long long cminflt;
119 unsigned long long majflt;
120 unsigned long long cmajflt;
121 unsigned long long utime;
122 unsigned long long stime;
123 unsigned long long gtime;
124 unsigned long long cutime;
125 unsigned long long cstime;
126 unsigned long long cgtime;
127 unsigned long long num_threads;
128 unsigned long long rss;
130 unsigned long long statm_size;
131 unsigned long long statm_resident;
132 unsigned long long statm_share;
133 unsigned long long statm_text;
134 unsigned long long statm_lib;
135 unsigned long long statm_data;
136 unsigned long long statm_dirty;
138 unsigned long long io_logical_bytes_read;
139 unsigned long long io_logical_bytes_written;
140 unsigned long long io_read_calls;
141 unsigned long long io_write_calls;
142 unsigned long long io_storage_bytes_read;
143 unsigned long long io_storage_bytes_written;
144 unsigned long long io_cancelled_write_bytes;
147 unsigned long long openfiles;
148 unsigned long long openpipes;
149 unsigned long long opensockets;
150 unsigned long long openinotifies;
151 unsigned long long openeventfds;
152 unsigned long long opentimerfds;
153 unsigned long long opensignalfds;
154 unsigned long long openeventpolls;
155 unsigned long long openother;
157 unsigned long processes; // how many processes have been merged to this
158 int exposed; // if set, we have sent this to netdata
159 int hidden; // if set, we set the hidden flag on the dimension
162 int starts_with; // if set, the compare string matches only the
163 // beginning of the command
165 struct target *target; // the one that will be reported to netdata
170 // ----------------------------------------------------------------------------
172 // aggregate all processes in groups, to have a limited number of dimensions
174 struct target *apps_groups_root_target = NULL;
175 struct target *apps_groups_default_target = NULL;
176 long apps_groups_targets = 0;
178 struct target *users_root_target = NULL;
179 struct target *groups_root_target = NULL;
181 struct target *get_users_target(uid_t uid)
184 for(w = users_root_target ; w ; w = w->next)
185 if(w->uid == uid) return w;
187 w = callocz(sizeof(struct target), 1);
188 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", uid);
189 w->comparehash = simple_hash(w->compare);
190 w->comparelen = strlen(w->compare);
192 snprintfz(w->id, MAX_NAME, "%u", uid);
193 w->idhash = simple_hash(w->id);
195 struct passwd *pw = getpwuid(uid);
197 snprintfz(w->name, MAX_NAME, "%u", uid);
199 snprintfz(w->name, MAX_NAME, "%s", pw->pw_name);
201 netdata_fix_chart_name(w->name);
205 w->next = users_root_target;
206 users_root_target = w;
209 fprintf(stderr, "apps.plugin: added uid %u ('%s') target\n", w->uid, w->name);
214 struct target *get_groups_target(gid_t gid)
217 for(w = groups_root_target ; w ; w = w->next)
218 if(w->gid == gid) return w;
220 w = callocz(sizeof(struct target), 1);
221 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", gid);
222 w->comparehash = simple_hash(w->compare);
223 w->comparelen = strlen(w->compare);
225 snprintfz(w->id, MAX_NAME, "%u", gid);
226 w->idhash = simple_hash(w->id);
228 struct group *gr = getgrgid(gid);
230 snprintfz(w->name, MAX_NAME, "%u", gid);
232 snprintfz(w->name, MAX_NAME, "%s", gr->gr_name);
234 netdata_fix_chart_name(w->name);
238 w->next = groups_root_target;
239 groups_root_target = w;
242 fprintf(stderr, "apps.plugin: added gid %u ('%s') target\n", w->gid, w->name);
247 // find or create a new target
248 // there are targets that are just aggregated to other target (the second argument)
249 struct target *get_apps_groups_target(const char *id, struct target *target) {
250 int tdebug = 0, thidden = 0, ends_with = 0;
251 const char *nid = id;
253 while(nid[0] == '-' || nid[0] == '+' || nid[0] == '*') {
254 if(nid[0] == '-') thidden = 1;
255 if(nid[0] == '+') tdebug = 1;
256 if(nid[0] == '*') ends_with = 1;
259 uint32_t hash = simple_hash(id);
261 struct target *w, *last = apps_groups_root_target;
262 for(w = apps_groups_root_target ; w ; w = w->next) {
263 if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
269 w = callocz(sizeof(struct target), 1);
270 strncpyz(w->id, nid, MAX_NAME);
271 w->idhash = simple_hash(w->id);
273 strncpyz(w->name, nid, MAX_NAME);
275 strncpyz(w->compare, nid, MAX_COMPARE_NAME);
276 size_t len = strlen(w->compare);
277 if(w->compare[len - 1] == '*') {
278 w->compare[len - 1] = '\0';
281 w->ends_with = ends_with;
283 if(w->starts_with && w->ends_with)
284 proc_pid_cmdline_is_needed = 1;
286 w->comparehash = simple_hash(w->compare);
287 w->comparelen = strlen(w->compare);
293 // append it, to maintain the order in apps_groups.conf
294 if(last) last->next = w;
295 else apps_groups_root_target = w;
298 fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
300 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
301 , w->target?w->target->id:w->id
302 , (w->hidden)?"hidden":"-"
303 , (w->debug)?"debug":"-"
309 // read the apps_groups.conf file
310 int read_apps_groups_conf(const char *name)
312 char filename[FILENAME_MAX + 1];
314 snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
317 fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
319 // ----------------------------------------
321 procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT);
324 procfile_set_quotes(ff, "'\"");
326 ff = procfile_readall(ff);
332 unsigned long line, lines = procfile_lines(ff);
334 for(line = 0; line < lines ;line++) {
335 unsigned long word, words = procfile_linewords(ff, line);
336 struct target *w = NULL;
338 char *t = procfile_lineword(ff, line, 0);
339 if(!t || !*t) continue;
341 for(word = 0; word < words ;word++) {
342 char *s = procfile_lineword(ff, line, word);
343 if(!s || !*s) continue;
348 struct target *n = get_apps_groups_target(s, w);
350 error("Cannot create target '%s' (line %lu, word %lu)", s, line, word);
358 int tdebug = 0, thidden = 0;
360 while(t[0] == '-' || t[0] == '+') {
361 if(t[0] == '-') thidden = 1;
362 if(t[0] == '+') tdebug = 1;
366 strncpyz(w->name, t, MAX_NAME);
371 fprintf(stderr, "apps.plugin: AGGREGATION TARGET NAME '%s' on ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
374 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
375 , w->target?w->target->id:w->id
376 , (w->hidden)?"hidden":"-"
377 , (w->debug)?"debug":"-"
384 apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
385 if(!apps_groups_default_target)
386 error("Cannot create default target");
388 strncpyz(apps_groups_default_target->name, "other", MAX_NAME);
394 // ----------------------------------------------------------------------------
395 // data to store for each pid
398 #define PID_LOG_IO 0x00000001
399 #define PID_LOG_STATM 0x00000002
400 #define PID_LOG_CMDLINE 0x00000004
401 #define PID_LOG_FDS 0x00000008
402 #define PID_LOG_STAT 0x00000010
406 char comm[MAX_COMPARE_NAME + 1];
407 char cmdline[MAX_CMDLINE + 1];
419 // these are raw values collected
420 unsigned long long minflt_raw;
421 unsigned long long cminflt_raw;
422 unsigned long long majflt_raw;
423 unsigned long long cmajflt_raw;
424 unsigned long long utime_raw;
425 unsigned long long stime_raw;
426 unsigned long long gtime_raw; // guest_time
427 unsigned long long cutime_raw;
428 unsigned long long cstime_raw;
429 unsigned long long cgtime_raw; // cguest_time
432 unsigned long long minflt;
433 unsigned long long cminflt;
434 unsigned long long majflt;
435 unsigned long long cmajflt;
436 unsigned long long utime;
437 unsigned long long stime;
438 unsigned long long gtime;
439 unsigned long long cutime;
440 unsigned long long cstime;
441 unsigned long long cgtime;
446 // int64_t itrealvalue;
447 // unsigned long long starttime;
448 // unsigned long long vsize;
449 unsigned long long rss;
450 // unsigned long long rsslim;
451 // unsigned long long starcode;
452 // unsigned long long endcode;
453 // unsigned long long startstack;
454 // unsigned long long kstkesp;
455 // unsigned long long kstkeip;
458 // uint64_t sigignore;
459 // uint64_t sigcatch;
463 // int32_t exit_signal;
464 // int32_t processor;
465 // uint32_t rt_priority;
467 // unsigned long long delayacct_blkio_ticks;
472 unsigned long long statm_size;
473 unsigned long long statm_resident;
474 unsigned long long statm_share;
475 unsigned long long statm_text;
476 unsigned long long statm_lib;
477 unsigned long long statm_data;
478 unsigned long long statm_dirty;
480 unsigned long long io_logical_bytes_read_raw;
481 unsigned long long io_logical_bytes_written_raw;
482 unsigned long long io_read_calls_raw;
483 unsigned long long io_write_calls_raw;
484 unsigned long long io_storage_bytes_read_raw;
485 unsigned long long io_storage_bytes_written_raw;
486 unsigned long long io_cancelled_write_bytes_raw;
488 unsigned long long io_logical_bytes_read;
489 unsigned long long io_logical_bytes_written;
490 unsigned long long io_read_calls;
491 unsigned long long io_write_calls;
492 unsigned long long io_storage_bytes_read;
493 unsigned long long io_storage_bytes_written;
494 unsigned long long io_cancelled_write_bytes;
496 int *fds; // array of fds it uses
497 int fds_size; // the size of the fds array
499 int children_count; // number of processes directly referencing this
500 int keep; // 1 when we need to keep this process in memory even after it exited
501 int keeploops; // increases by 1 every time keep is 1 and updated 0
502 int updated; // 1 when the process is currently running
503 int merged; // 1 when it has been merged to its parent
504 int new_entry; // 1 when this is a new process, just saw for the first time
505 int read; // 1 when we have already read this process for this iteration
506 int sortlist; // higher numbers = top on the process tree
507 // each process gets a unique number
509 struct target *target; // app_groups.conf targets
510 struct target *user_target; // uid based targets
511 struct target *group_target; // gid based targets
513 unsigned long long stat_collected_usec;
514 unsigned long long last_stat_collected_usec;
516 unsigned long long io_collected_usec;
517 unsigned long long last_io_collected_usec;
520 char *statm_filename;
522 char *cmdline_filename;
524 struct pid_stat *parent;
525 struct pid_stat *prev;
526 struct pid_stat *next;
527 } *root_of_pids = NULL, **all_pids;
529 long all_pids_count = 0;
531 struct pid_stat *get_pid_entry(pid_t pid) {
533 all_pids[pid]->new_entry = 0;
534 return all_pids[pid];
537 all_pids[pid] = callocz(sizeof(struct pid_stat), 1);
538 all_pids[pid]->fds = callocz(sizeof(int), 100);
539 all_pids[pid]->fds_size = 100;
541 if(root_of_pids) root_of_pids->prev = all_pids[pid];
542 all_pids[pid]->next = root_of_pids;
543 root_of_pids = all_pids[pid];
545 all_pids[pid]->pid = pid;
546 all_pids[pid]->new_entry = 1;
550 return all_pids[pid];
553 void del_pid_entry(pid_t pid) {
555 error("attempted to free pid %d that is not allocated.", pid);
560 fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
562 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
563 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
564 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
566 if(all_pids[pid]->fds) freez(all_pids[pid]->fds);
567 if(all_pids[pid]->stat_filename) freez(all_pids[pid]->stat_filename);
568 if(all_pids[pid]->statm_filename) freez(all_pids[pid]->statm_filename);
569 if(all_pids[pid]->io_filename) freez(all_pids[pid]->io_filename);
570 if(all_pids[pid]->cmdline_filename) freez(all_pids[pid]->cmdline_filename);
571 freez(all_pids[pid]);
573 all_pids[pid] = NULL;
578 // ----------------------------------------------------------------------------
579 // update pids from proc
581 int read_proc_pid_cmdline(struct pid_stat *p) {
583 if(unlikely(!p->cmdline_filename)) {
584 char filename[FILENAME_MAX + 1];
585 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", host_prefix, p->pid);
586 p->cmdline_filename = strdupz(filename);
589 int fd = open(p->cmdline_filename, O_RDONLY, 0666);
590 if(unlikely(fd == -1)) goto cleanup;
592 ssize_t i, bytes = read(fd, p->cmdline, MAX_CMDLINE);
595 if(unlikely(bytes < 0)) goto cleanup;
597 p->cmdline[bytes] = '\0';
598 for(i = 0; i < bytes ; i++)
599 if(unlikely(!p->cmdline[i])) p->cmdline[i] = ' ';
602 fprintf(stderr, "Read file '%s' contents: %s\n", p->cmdline_filename, p->cmdline);
607 // copy the command to the command line
608 strncpyz(p->cmdline, p->comm, MAX_CMDLINE);
612 int read_proc_pid_ownership(struct pid_stat *p) {
613 if(unlikely(!p->stat_filename)) {
614 error("pid %d does not have a stat_filename", p->pid);
618 // ----------------------------------------
622 if(stat(p->stat_filename, &st) != 0) {
623 error("Cannot stat file '%s'", p->stat_filename);
633 int read_proc_pid_stat(struct pid_stat *p) {
634 static procfile *ff = NULL;
636 if(unlikely(!p->stat_filename)) {
637 char filename[FILENAME_MAX + 1];
638 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
639 p->stat_filename = strdupz(filename);
642 int set_quotes = (!ff)?1:0;
644 ff = procfile_reopen(ff, p->stat_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
645 if(unlikely(!ff)) goto cleanup;
647 // if(set_quotes) procfile_set_quotes(ff, "()");
648 if(set_quotes) procfile_set_open_close(ff, "(", ")");
650 ff = procfile_readall(ff);
651 if(unlikely(!ff)) goto cleanup;
653 p->last_stat_collected_usec = p->stat_collected_usec;
654 p->stat_collected_usec = time_usec();
657 // p->pid = atol(procfile_lineword(ff, 0, 0+i));
659 strncpyz(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
661 // p->state = *(procfile_lineword(ff, 0, 2));
662 p->ppid = (int32_t) atol(procfile_lineword(ff, 0, 3));
663 // p->pgrp = atol(procfile_lineword(ff, 0, 4));
664 // p->session = atol(procfile_lineword(ff, 0, 5));
665 // p->tty_nr = atol(procfile_lineword(ff, 0, 6));
666 // p->tpgid = atol(procfile_lineword(ff, 0, 7));
667 // p->flags = strtoull(procfile_lineword(ff, 0, 8), NULL, 10);
669 unsigned long long last;
671 last = p->minflt_raw;
672 p->minflt_raw = strtoull(procfile_lineword(ff, 0, 9), NULL, 10);
673 p->minflt = (p->minflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
675 last = p->cminflt_raw;
676 p->cminflt_raw = strtoull(procfile_lineword(ff, 0, 10), NULL, 10);
677 p->cminflt = (p->cminflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
679 last = p->majflt_raw;
680 p->majflt_raw = strtoull(procfile_lineword(ff, 0, 11), NULL, 10);
681 p->majflt = (p->majflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
683 last = p->cmajflt_raw;
684 p->cmajflt_raw = strtoull(procfile_lineword(ff, 0, 12), NULL, 10);
685 p->cmajflt = (p->cmajflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
688 p->utime_raw = strtoull(procfile_lineword(ff, 0, 13), NULL, 10);
689 p->utime = (p->utime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
692 p->stime_raw = strtoull(procfile_lineword(ff, 0, 14), NULL, 10);
693 p->stime = (p->stime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
695 last = p->cutime_raw;
696 p->cutime_raw = strtoull(procfile_lineword(ff, 0, 15), NULL, 10);
697 p->cutime = (p->cutime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
699 last = p->cstime_raw;
700 p->cstime_raw = strtoull(procfile_lineword(ff, 0, 16), NULL, 10);
701 p->cstime = (p->cstime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
703 // p->priority = strtoull(procfile_lineword(ff, 0, 17), NULL, 10);
704 // p->nice = strtoull(procfile_lineword(ff, 0, 18), NULL, 10);
705 p->num_threads = (int32_t) atol(procfile_lineword(ff, 0, 19));
706 // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20), NULL, 10);
707 // p->starttime = strtoull(procfile_lineword(ff, 0, 21), NULL, 10);
708 // p->vsize = strtoull(procfile_lineword(ff, 0, 22), NULL, 10);
709 p->rss = strtoull(procfile_lineword(ff, 0, 23), NULL, 10);
710 // p->rsslim = strtoull(procfile_lineword(ff, 0, 24), NULL, 10);
711 // p->starcode = strtoull(procfile_lineword(ff, 0, 25), NULL, 10);
712 // p->endcode = strtoull(procfile_lineword(ff, 0, 26), NULL, 10);
713 // p->startstack = strtoull(procfile_lineword(ff, 0, 27), NULL, 10);
714 // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28), NULL, 10);
715 // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29), NULL, 10);
716 // p->signal = strtoull(procfile_lineword(ff, 0, 30), NULL, 10);
717 // p->blocked = strtoull(procfile_lineword(ff, 0, 31), NULL, 10);
718 // p->sigignore = strtoull(procfile_lineword(ff, 0, 32), NULL, 10);
719 // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33), NULL, 10);
720 // p->wchan = strtoull(procfile_lineword(ff, 0, 34), NULL, 10);
721 // p->nswap = strtoull(procfile_lineword(ff, 0, 35), NULL, 10);
722 // p->cnswap = strtoull(procfile_lineword(ff, 0, 36), NULL, 10);
723 // p->exit_signal = atol(procfile_lineword(ff, 0, 37));
724 // p->processor = atol(procfile_lineword(ff, 0, 38));
725 // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39), NULL, 10);
726 // p->policy = strtoul(procfile_lineword(ff, 0, 40), NULL, 10);
727 // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41), NULL, 10);
729 if(enable_guest_charts) {
731 p->gtime_raw = strtoull(procfile_lineword(ff, 0, 42), NULL, 10);
732 p->gtime = (p->gtime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
734 last = p->cgtime_raw;
735 p->cgtime_raw = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
736 p->cgtime = (p->cgtime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
738 if (show_guest_time || p->gtime || p->cgtime) {
739 p->utime -= (p->utime >= p->gtime) ? p->gtime : p->utime;
740 p->cutime -= (p->cutime >= p->cgtime) ? p->cgtime : p->cutime;
745 if(unlikely(debug || (p->target && p->target->debug)))
746 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);
748 if(unlikely(global_iterations_counter == 1)) {
779 int read_proc_pid_statm(struct pid_stat *p) {
780 static procfile *ff = NULL;
782 if(unlikely(!p->statm_filename)) {
783 char filename[FILENAME_MAX + 1];
784 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
785 p->statm_filename = strdupz(filename);
788 ff = procfile_reopen(ff, p->statm_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
789 if(unlikely(!ff)) goto cleanup;
791 ff = procfile_readall(ff);
792 if(unlikely(!ff)) goto cleanup;
796 p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
797 p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
798 p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
799 p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
800 p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
801 p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
802 p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
808 p->statm_resident = 0;
817 int read_proc_pid_io(struct pid_stat *p) {
818 static procfile *ff = NULL;
820 if(unlikely(!p->io_filename)) {
821 char filename[FILENAME_MAX + 1];
822 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
823 p->io_filename = strdupz(filename);
827 ff = procfile_reopen(ff, p->io_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
828 if(unlikely(!ff)) goto cleanup;
830 ff = procfile_readall(ff);
831 if(unlikely(!ff)) goto cleanup;
835 p->last_io_collected_usec = p->io_collected_usec;
836 p->io_collected_usec = time_usec();
838 unsigned long long last;
840 last = p->io_logical_bytes_read_raw;
841 p->io_logical_bytes_read_raw = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
842 p->io_logical_bytes_read = (p->io_logical_bytes_read_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
844 last = p->io_logical_bytes_written_raw;
845 p->io_logical_bytes_written_raw = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
846 p->io_logical_bytes_written = (p->io_logical_bytes_written_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
848 last = p->io_read_calls_raw;
849 p->io_read_calls_raw = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
850 p->io_read_calls = (p->io_read_calls_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
852 last = p->io_write_calls_raw;
853 p->io_write_calls_raw = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
854 p->io_write_calls = (p->io_write_calls_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
856 last = p->io_storage_bytes_read_raw;
857 p->io_storage_bytes_read_raw = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
858 p->io_storage_bytes_read = (p->io_storage_bytes_read_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
860 last = p->io_storage_bytes_written_raw;
861 p->io_storage_bytes_written_raw = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
862 p->io_storage_bytes_written = (p->io_storage_bytes_written_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
864 last = p->io_cancelled_write_bytes_raw;
865 p->io_cancelled_write_bytes_raw = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
866 p->io_cancelled_write_bytes = (p->io_cancelled_write_bytes_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
868 if(unlikely(global_iterations_counter == 1)) {
869 p->io_logical_bytes_read = 0;
870 p->io_logical_bytes_written = 0;
871 p->io_read_calls = 0;
872 p->io_write_calls = 0;
873 p->io_storage_bytes_read = 0;
874 p->io_storage_bytes_written = 0;
875 p->io_cancelled_write_bytes = 0;
881 p->io_logical_bytes_read = 0;
882 p->io_logical_bytes_written = 0;
883 p->io_read_calls = 0;
884 p->io_write_calls = 0;
885 p->io_storage_bytes_read = 0;
886 p->io_storage_bytes_written = 0;
887 p->io_cancelled_write_bytes = 0;
891 unsigned long long global_utime = 0;
892 unsigned long long global_stime = 0;
893 unsigned long long global_gtime = 0;
895 int read_proc_stat() {
896 static char filename[FILENAME_MAX + 1] = "";
897 static procfile *ff = NULL;
898 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;
901 snprintfz(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
902 ff = procfile_open(filename, " \t:", PROCFILE_FLAG_DEFAULT);
903 if(unlikely(!ff)) goto cleanup;
906 ff = procfile_readall(ff);
907 if(unlikely(!ff)) goto cleanup;
909 last_collected_usec = collected_usec;
910 collected_usec = time_usec();
914 unsigned long long last;
917 utime_raw = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
918 global_utime = (utime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
920 // nice time, on user time
922 ntime_raw = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
923 global_utime += (ntime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
926 stime_raw = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
927 global_stime = (stime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
930 gtime_raw = strtoull(procfile_lineword(ff, 0, 10), NULL, 10);
931 global_gtime = (gtime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
933 if(enable_guest_charts) {
934 // guest nice time, on guest time
936 gntime_raw = strtoull(procfile_lineword(ff, 0, 11), NULL, 10);
937 global_gtime += (gntime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
939 // remove guest time from user time
940 global_utime -= (global_utime > global_gtime) ? global_gtime : global_utime;
943 if(unlikely(global_iterations_counter == 1)) {
959 // ----------------------------------------------------------------------------
961 // this is used to keep a global list of all open files of the system
962 // it is needed in order to calculate the unique files processes have open
964 #define FILE_DESCRIPTORS_INCREASE_STEP 100
966 struct file_descriptor {
968 #ifdef NETDATA_INTERNAL_CHECKS
970 #endif /* NETDATA_INTERNAL_CHECKS */
978 int all_files_len = 0;
979 int all_files_size = 0;
981 int file_descriptor_compare(void* a, void* b) {
982 #ifdef NETDATA_INTERNAL_CHECKS
983 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
984 error("Corrupted index data detected. Please report this.");
985 #endif /* NETDATA_INTERNAL_CHECKS */
987 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
990 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
994 return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
997 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
999 avl_tree all_files_index = {
1001 file_descriptor_compare
1004 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
1005 struct file_descriptor tmp;
1006 tmp.hash = (hash)?hash:simple_hash(name);
1010 #ifdef NETDATA_INTERNAL_CHECKS
1011 tmp.magic = 0x0BADCAFE;
1012 #endif /* NETDATA_INTERNAL_CHECKS */
1014 return (struct file_descriptor *)avl_search(&all_files_index, (avl *) &tmp);
1017 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
1018 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
1020 #define FILETYPE_OTHER 0
1021 #define FILETYPE_FILE 1
1022 #define FILETYPE_PIPE 2
1023 #define FILETYPE_SOCKET 3
1024 #define FILETYPE_INOTIFY 4
1025 #define FILETYPE_EVENTFD 5
1026 #define FILETYPE_EVENTPOLL 6
1027 #define FILETYPE_TIMERFD 7
1028 #define FILETYPE_SIGNALFD 8
1030 void file_descriptor_not_used(int id)
1032 if(id > 0 && id < all_files_size) {
1034 #ifdef NETDATA_INTERNAL_CHECKS
1035 if(all_files[id].magic != 0x0BADCAFE) {
1036 error("Ignoring request to remove empty file id %d.", id);
1039 #endif /* NETDATA_INTERNAL_CHECKS */
1042 fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
1044 if(all_files[id].count > 0) {
1045 all_files[id].count--;
1047 if(!all_files[id].count) {
1049 fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
1051 file_descriptor_remove(&all_files[id]);
1052 #ifdef NETDATA_INTERNAL_CHECKS
1053 all_files[id].magic = 0x00000000;
1054 #endif /* NETDATA_INTERNAL_CHECKS */
1059 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
1061 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
1064 int file_descriptor_find_or_add(const char *name)
1066 static int last_pos = 0;
1067 uint32_t hash = simple_hash(name);
1070 fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
1072 struct file_descriptor *fd = file_descriptor_find(name, hash);
1076 fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
1083 // check we have enough memory to add it
1084 if(!all_files || all_files_len == all_files_size) {
1085 void *old = all_files;
1088 // there is no empty slot
1090 fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
1092 all_files = reallocz(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
1094 // if the address changed, we have to rebuild the index
1095 // since all pointers are now invalid
1096 if(old && old != (void *)all_files) {
1098 fprintf(stderr, "apps.plugin: >> re-indexing.\n");
1100 all_files_index.root = NULL;
1101 for(i = 0; i < all_files_size; i++) {
1102 if(!all_files[i].count) continue;
1103 file_descriptor_add(&all_files[i]);
1107 fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
1110 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
1111 all_files[i].count = 0;
1112 all_files[i].name = NULL;
1113 #ifdef NETDATA_INTERNAL_CHECKS
1114 all_files[i].magic = 0x00000000;
1115 #endif /* NETDATA_INTERNAL_CHECKS */
1116 all_files[i].pos = i;
1119 if(!all_files_size) all_files_len = 1;
1120 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
1124 fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
1126 // search for an empty slot
1128 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
1129 if(c >= all_files_size) c = 0;
1130 if(c == 0) continue;
1132 if(!all_files[c].count) {
1134 fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
1136 #ifdef NETDATA_INTERNAL_CHECKS
1137 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
1138 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
1139 #endif /* NETDATA_INTERNAL_CHECKS */
1142 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);
1144 if(all_files[c].name) freez((void *)all_files[c].name);
1145 all_files[c].name = NULL;
1150 if(i == all_files_size) {
1151 fatal("We should find an empty slot, but there isn't any");
1156 fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
1160 // else we have an empty slot in 'c'
1163 if(name[0] == '/') type = FILETYPE_FILE;
1164 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
1165 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
1166 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
1167 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
1168 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
1169 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
1170 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
1171 else if(strncmp(name, "anon_inode:", 11) == 0) {
1173 fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
1175 type = FILETYPE_OTHER;
1179 fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
1181 type = FILETYPE_OTHER;
1184 all_files[c].name = strdupz(name);
1185 all_files[c].hash = hash;
1186 all_files[c].type = type;
1187 all_files[c].pos = c;
1188 all_files[c].count = 1;
1189 #ifdef NETDATA_INTERNAL_CHECKS
1190 all_files[c].magic = 0x0BADCAFE;
1191 #endif /* NETDATA_INTERNAL_CHECKS */
1192 file_descriptor_add(&all_files[c]);
1195 fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1200 int read_pid_file_descriptors(struct pid_stat *p) {
1201 char dirname[FILENAME_MAX+1];
1203 snprintfz(dirname, FILENAME_MAX, "%s/proc/%d/fd", host_prefix, p->pid);
1204 DIR *fds = opendir(dirname);
1208 char fdname[FILENAME_MAX + 1];
1209 char linkname[FILENAME_MAX + 1];
1211 // make the array negative
1212 for(c = 0 ; c < p->fds_size ; c++)
1213 p->fds[c] = -p->fds[c];
1215 while((de = readdir(fds))) {
1216 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
1219 // check if the fds array is small
1220 int fdid = atoi(de->d_name);
1221 if(fdid < 0) continue;
1222 if(fdid >= p->fds_size) {
1223 // it is small, extend it
1225 fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
1227 p->fds = reallocz(p->fds, (fdid + 100) * sizeof(int));
1229 fatal("Cannot re-allocate fds for %s", p->comm);
1233 // and initialize it
1234 for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
1235 p->fds_size = fdid + 100;
1238 if(p->fds[fdid] == 0) {
1239 // we don't know this fd, get it
1241 sprintf(fdname, "%s/proc/%d/fd/%s", host_prefix, p->pid, de->d_name);
1242 ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
1244 if(debug || (p->target && p->target->debug)) {
1245 if(debug || (p->target && p->target->debug))
1246 error("Cannot read link %s", fdname);
1253 // if another process already has this, we will get
1255 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1258 // else make it positive again, we need it
1259 // of course, the actual file may have changed, but we don't care so much
1260 // FIXME: we could compare the inode as returned by readdir dirent structure
1261 else p->fds[fdid] = -p->fds[fdid];
1265 // remove all the negative file descriptors
1266 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1267 file_descriptor_not_used(-p->fds[c]);
1276 // ----------------------------------------------------------------------------
1278 int print_process_and_parents(struct pid_stat *p, unsigned long long time) {
1279 char *prefix = "\\_ ";
1283 indent = print_process_and_parents(p->parent, p->stat_collected_usec);
1287 char buffer[indent + 1];
1290 for(i = 0; i < indent ;i++) buffer[i] = ' ';
1293 fprintf(stderr, " %s %s%s (%d %s %lld"
1298 , p->updated?"running":"exited"
1299 , (long long)p->stat_collected_usec - (long long)time
1302 if(p->utime) fprintf(stderr, " utime=%llu", p->utime);
1303 if(p->stime) fprintf(stderr, " stime=%llu", p->stime);
1304 if(p->gtime) fprintf(stderr, " gtime=%llu", p->gtime);
1305 if(p->cutime) fprintf(stderr, " cutime=%llu", p->cutime);
1306 if(p->cstime) fprintf(stderr, " cstime=%llu", p->cstime);
1307 if(p->cgtime) fprintf(stderr, " cgtime=%llu", p->cgtime);
1308 if(p->minflt) fprintf(stderr, " minflt=%llu", p->minflt);
1309 if(p->cminflt) fprintf(stderr, " cminflt=%llu", p->cminflt);
1310 if(p->majflt) fprintf(stderr, " majflt=%llu", p->majflt);
1311 if(p->cmajflt) fprintf(stderr, " cmajflt=%llu", p->cmajflt);
1312 fprintf(stderr, ")\n");
1317 void print_process_tree(struct pid_stat *p, char *msg) {
1319 fprintf(stderr, "%s: process %s (%d, %s) with parents:\n", msg, p->comm, p->pid, p->updated?"running":"exited");
1320 print_process_and_parents(p, p->stat_collected_usec);
1323 void find_lost_child_debug(struct pid_stat *pe, unsigned long long lost, int type) {
1325 struct pid_stat *p = NULL;
1327 for(p = root_of_pids; p ; p = p->next) {
1328 if(p == pe) continue;
1332 if(p->cminflt > lost) {
1333 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);
1339 if(p->cmajflt > lost) {
1340 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);
1346 if(p->cutime > lost) {
1347 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);
1353 if(p->cstime > lost) {
1354 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);
1360 if(p->cgtime > lost) {
1361 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);
1371 fprintf(stderr, " > cannot find any process to use the lost exited child minflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1375 fprintf(stderr, " > cannot find any process to use the lost exited child majflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1379 fprintf(stderr, " > cannot find any process to use the lost exited child utime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1383 fprintf(stderr, " > cannot find any process to use the lost exited child stime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1387 fprintf(stderr, " > cannot find any process to use the lost exited child gtime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1393 unsigned long long remove_exited_child_from_parent(unsigned long long *field, unsigned long long *pfield) {
1394 unsigned long long absorbed = 0;
1396 if(*field > *pfield) {
1397 absorbed += *pfield;
1410 void process_exited_processes() {
1413 for(p = root_of_pids; p ; p = p->next) {
1414 if(p->updated || !p->stat_collected_usec)
1417 struct pid_stat *pp = p->parent;
1419 unsigned long long utime = (p->utime_raw + p->cutime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1420 unsigned long long stime = (p->stime_raw + p->cstime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1421 unsigned long long gtime = (p->gtime_raw + p->cgtime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1422 unsigned long long minflt = (p->minflt_raw + p->cminflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1423 unsigned long long majflt = (p->majflt_raw + p->cmajflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1425 if(utime + stime + gtime + minflt + majflt == 0)
1428 if(unlikely(debug)) {
1430 fprintf(stderr, "Absorb %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1433 , p->updated?"running":"exited"
1440 print_process_tree(p, "Searching parents");
1443 for(pp = p->parent; pp ; pp = pp->parent) {
1444 if(!pp->updated) continue;
1446 unsigned long long absorbed;
1447 absorbed = remove_exited_child_from_parent(&utime, &pp->cutime);
1448 if(unlikely(debug && absorbed))
1449 fprintf(stderr, " > process %s (%d %s) absorbed %llu utime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, utime);
1451 absorbed = remove_exited_child_from_parent(&stime, &pp->cstime);
1452 if(unlikely(debug && absorbed))
1453 fprintf(stderr, " > process %s (%d %s) absorbed %llu stime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, stime);
1455 absorbed = remove_exited_child_from_parent(>ime, &pp->cgtime);
1456 if(unlikely(debug && absorbed))
1457 fprintf(stderr, " > process %s (%d %s) absorbed %llu gtime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, gtime);
1459 absorbed = remove_exited_child_from_parent(&minflt, &pp->cminflt);
1460 if(unlikely(debug && absorbed))
1461 fprintf(stderr, " > process %s (%d %s) absorbed %llu minflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, minflt);
1463 absorbed = remove_exited_child_from_parent(&majflt, &pp->cmajflt);
1464 if(unlikely(debug && absorbed))
1465 fprintf(stderr, " > process %s (%d %s) absorbed %llu majflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, majflt);
1468 if(unlikely(utime + stime + gtime + minflt + majflt > 0)) {
1469 if(unlikely(debug)) {
1470 if(utime) find_lost_child_debug(p, utime, 3);
1471 if(stime) find_lost_child_debug(p, stime, 4);
1472 if(gtime) find_lost_child_debug(p, gtime, 5);
1473 if(minflt) find_lost_child_debug(p, minflt, 1);
1474 if(majflt) find_lost_child_debug(p, majflt, 2);
1480 fprintf(stderr, " > remaining resources - KEEP - for another loop: %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1483 , p->updated?"running":"exited"
1491 for(pp = p->parent; pp ; pp = pp->parent) {
1492 if(pp->updated) break;
1496 fprintf(stderr, " > - KEEP - parent for another loop: %s (%d %s)\n"
1499 , pp->updated?"running":"exited"
1503 p->utime_raw = utime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1504 p->stime_raw = stime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1505 p->gtime_raw = gtime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1506 p->minflt_raw = minflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1507 p->majflt_raw = majflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1508 p->cutime_raw = p->cstime_raw = p->cgtime_raw = p->cminflt_raw = p->cmajflt_raw = 0;
1511 fprintf(stderr, "\n");
1513 else if(unlikely(debug)) {
1514 fprintf(stderr, " > totally absorbed - DONE - %s (%d %s)\n"
1517 , p->updated?"running":"exited"
1523 void link_all_processes_to_their_parents(void) {
1524 struct pid_stat *p, *pp;
1526 // link all children to their parents
1527 // and update children count on parents
1528 for(p = root_of_pids; p ; p = p->next) {
1529 // for each process found
1534 if(unlikely(!p->ppid)) {
1539 pp = all_pids[p->ppid];
1542 pp->children_count++;
1544 if(unlikely(debug || (p->target && p->target->debug)))
1545 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);
1549 error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1554 // ----------------------------------------------------------------------------
1556 // 1. read all files in /proc
1557 // 2. for each numeric directory:
1558 // i. read /proc/pid/stat
1559 // ii. read /proc/pid/statm
1560 // iii. read /proc/pid/io (requires root access)
1561 // iii. read the entries in directory /proc/pid/fd (requires root access)
1563 // a. find or create a struct file_descriptor
1564 // b. cleanup any old/unused file_descriptors
1566 // after all these, some pids may be linked to targets, while others may not
1568 // in case of errors, only 1 every 1000 errors is printed
1569 // to avoid filling up all disk space
1570 // if debug is enabled, all errors are printed
1572 static int compar_pid(const void *pid1, const void *pid2) {
1574 struct pid_stat *p1 = all_pids[*((pid_t *)pid1)];
1575 struct pid_stat *p2 = all_pids[*((pid_t *)pid2)];
1577 if(p1->sortlist > p2->sortlist)
1583 static inline int managed_log(struct pid_stat *p, uint32_t log, int status) {
1584 if(unlikely(!status)) {
1585 // error("command failed log %u, errno %d", log, errno);
1587 if(unlikely(debug || errno != ENOENT)) {
1588 if(unlikely(debug || !(p->log_thrown & log))) {
1589 p->log_thrown |= log;
1592 error("Cannot process %s/proc/%d/io (command '%s')", host_prefix, p->pid, p->comm);
1596 error("Cannot process %s/proc/%d/statm (command '%s')", host_prefix, p->pid, p->comm);
1599 case PID_LOG_CMDLINE:
1600 error("Cannot process %s/proc/%d/cmdline (command '%s')", host_prefix, p->pid, p->comm);
1604 error("Cannot process entries in %s/proc/%d/fd (command '%s')", host_prefix, p->pid, p->comm);
1611 error("unhandled error for pid %d, command '%s'", p->pid, p->comm);
1618 else if(unlikely(p->log_thrown & log)) {
1619 // error("unsetting log %u on pid %d", log, p->pid);
1620 p->log_thrown &= ~log;
1626 void collect_data_for_pid(pid_t pid) {
1627 if(unlikely(pid <= 0 || pid > pid_max)) {
1628 error("Invalid pid %d read (expected 1 to %d). Ignoring process.", pid, pid_max);
1632 struct pid_stat *p = get_pid_entry(pid);
1633 if(unlikely(!p || p->read)) return;
1636 // fprintf(stderr, "Reading process %d (%s), sortlist %d\n", p->pid, p->comm, p->sortlist);
1638 // --------------------------------------------------------------------
1641 if(unlikely(!managed_log(p, PID_LOG_STAT, read_proc_pid_stat(p))))
1642 // there is no reason to proceed if we cannot get its status
1645 read_proc_pid_ownership(p);
1647 // check its parent pid
1648 if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
1649 error("Pid %d (command '%s') states invalid parent pid %d. Using 0.", pid, p->comm, p->ppid);
1653 // --------------------------------------------------------------------
1656 managed_log(p, PID_LOG_IO, read_proc_pid_io(p));
1658 // --------------------------------------------------------------------
1659 // /proc/<pid>/statm
1661 if(unlikely(!managed_log(p, PID_LOG_STATM, read_proc_pid_statm(p))))
1662 // there is no reason to proceed if we cannot get its memory status
1665 // --------------------------------------------------------------------
1668 // check if it is target
1669 // we do this only once, the first time this pid is loaded
1670 if(unlikely(p->new_entry)) {
1671 // /proc/<pid>/cmdline
1672 if(likely(proc_pid_cmdline_is_needed))
1673 managed_log(p, PID_LOG_CMDLINE, read_proc_pid_cmdline(p));
1676 fprintf(stderr, "apps.plugin: \tJust added %d (%s)\n", pid, p->comm);
1678 uint32_t hash = simple_hash(p->comm);
1679 size_t pclen = strlen(p->comm);
1682 for(w = apps_groups_root_target; w ; w = w->next) {
1683 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1685 // find it - 4 cases:
1686 // 1. the target is not a pattern
1687 // 2. the target has the prefix
1688 // 3. the target has the suffix
1689 // 4. the target is something inside cmdline
1690 if( (!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm))
1691 || (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen))
1692 || (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen]))
1693 || (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && strstr(p->cmdline, w->compare))
1695 if(w->target) p->target = w->target;
1698 if(debug || (p->target && p->target->debug))
1699 fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1706 // --------------------------------------------------------------------
1709 if(enable_file_charts)
1710 managed_log(p, PID_LOG_FDS, read_pid_file_descriptors(p));
1712 // --------------------------------------------------------------------
1715 if(unlikely(debug && include_exited_childs && all_pids_count && p->ppid && all_pids[p->ppid] && !all_pids[p->ppid]->read))
1716 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);
1718 // mark it as updated
1724 int collect_data_for_all_processes_from_proc(void) {
1725 struct pid_stat *p = NULL;
1727 if(all_pids_count) {
1728 // read parents before childs
1729 // this is needed to prevent a situation where
1730 // a child is found running, but until we read
1731 // its parent, it has exited and its parent
1732 // has accumulated its resources
1735 for(p = root_of_pids; p ; p = p->next) {
1740 p->children_count = 0;
1743 all_pids_sortlist[slc++] = p->pid;
1746 if(unlikely(slc != all_pids_count)) {
1747 error("Internal error: I was thinking I had %ld processes in my arrays, but it seems there are more.", all_pids_count);
1748 all_pids_count = slc;
1751 if(include_exited_childs) {
1752 qsort((void *)all_pids_sortlist, all_pids_count, sizeof(pid_t), compar_pid);
1753 for(slc = 0; slc < all_pids_count; slc++)
1754 collect_data_for_pid(all_pids_sortlist[slc]);
1758 char dirname[FILENAME_MAX + 1];
1760 snprintfz(dirname, FILENAME_MAX, "%s/proc", host_prefix);
1761 DIR *dir = opendir(dirname);
1764 struct dirent *file = NULL;
1766 while((file = readdir(dir))) {
1767 char *endptr = file->d_name;
1768 pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
1770 // make sure we read a valid number
1771 if(unlikely(endptr == file->d_name || *endptr != '\0'))
1774 collect_data_for_pid(pid);
1778 // normally this is done
1779 // however we may have processes exited while we collected values
1780 // so let's find the exited ones
1781 // we do this by collecting the ownership of process
1782 // if we manage to get the ownership, the process still runs
1785 link_all_processes_to_their_parents();
1786 process_exited_processes();
1791 // ----------------------------------------------------------------------------
1792 // update statistics on the targets
1794 // 1. link all childs to their parents
1795 // 2. go from bottom to top, marking as merged all childs to their parents
1796 // this step links all parents without a target to the child target, if any
1797 // 3. link all top level processes (the ones not merged) to the default target
1798 // 4. go from top to bottom, linking all childs without a target, to their parent target
1799 // after this step, all processes have a target
1800 // [5. for each killed pid (updated = 0), remove its usage from its target]
1801 // 6. zero all apps_groups_targets
1802 // 7. concentrate all values on the apps_groups_targets
1803 // 8. remove all killed processes
1804 // 9. find the unique file count for each target
1805 // check: update_apps_groups_statistics()
1807 void cleanup_exited_pids(void) {
1809 struct pid_stat *p = NULL;
1811 for(p = root_of_pids; p ;) {
1812 if(!p->updated && (!p->keep || p->keeploops > 0)) {
1813 // 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);
1815 if(unlikely(debug && (p->keep || p->keeploops)))
1816 fprintf(stderr, " > CLEANUP cannot keep exited process %d (%s) anymore - removing it.\n", p->pid, p->comm);
1818 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1819 file_descriptor_not_used(p->fds[c]);
1828 if(unlikely(p->keep)) p->keeploops++;
1835 void apply_apps_groups_targets_inheritance(void) {
1836 struct pid_stat *p = NULL;
1838 // children that do not have a target
1839 // inherit their target from their parent
1840 int found = 1, loops = 0;
1842 if(unlikely(debug)) loops++;
1844 for(p = root_of_pids; p ; p = p->next) {
1845 // if this process does not have a target
1846 // and it has a parent
1847 // and its parent has a target
1848 // then, set the parent's target to this process
1849 if(unlikely(!p->target && p->parent && p->parent->target)) {
1850 p->target = p->parent->target;
1853 if(debug || (p->target && p->target->debug))
1854 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);
1859 // find all the procs with 0 childs and merge them to their parents
1860 // repeat, until nothing more can be done.
1864 if(unlikely(debug)) loops++;
1867 for(p = root_of_pids; p ; p = p->next) {
1868 if(unlikely(!p->sortlist && !p->children_count))
1869 p->sortlist = sortlist++;
1871 // if this process does not have any children
1872 // and is not already merged
1874 // and its parent has children
1875 // and the target of this process and its parent is the same, or the parent does not have a target
1876 // and its parent is not init
1877 // then, mark them as merged.
1882 && p->parent->children_count
1883 && (p->target == p->parent->target || !p->parent->target)
1886 p->parent->children_count--;
1889 // the parent inherits the child's target, if it does not have a target itself
1890 if(unlikely(p->target && !p->parent->target)) {
1891 p->parent->target = p->target;
1893 if(debug || (p->target && p->target->debug))
1894 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);
1902 fprintf(stderr, "apps.plugin: TARGET INHERITANCE: merged %d processes\n", found);
1905 // init goes always to default target
1907 all_pids[1]->target = apps_groups_default_target;
1909 // give a default target on all top level processes
1910 if(unlikely(debug)) loops++;
1911 for(p = root_of_pids; p ; p = p->next) {
1912 // if the process is not merged itself
1913 // then is is a top level process
1914 if(unlikely(!p->merged && !p->target))
1915 p->target = apps_groups_default_target;
1917 // make sure all processes have a sortlist
1918 if(unlikely(!p->sortlist))
1919 p->sortlist = sortlist++;
1923 all_pids[1]->sortlist = sortlist++;
1925 // give a target to all merged child processes
1928 if(unlikely(debug)) loops++;
1930 for(p = root_of_pids; p ; p = p->next) {
1931 if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
1932 p->target = p->parent->target;
1935 if(debug || (p->target && p->target->debug))
1936 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);
1942 fprintf(stderr, "apps.plugin: apply_apps_groups_targets_inheritance() made %d loops on the process tree\n", loops);
1945 long zero_all_targets(struct target *root) {
1949 for (w = root; w ; w = w->next) {
1952 if(w->fds) freez(w->fds);
1970 w->statm_resident = 0;
1977 w->io_logical_bytes_read = 0;
1978 w->io_logical_bytes_written = 0;
1979 w->io_read_calls = 0;
1980 w->io_write_calls = 0;
1981 w->io_storage_bytes_read = 0;
1982 w->io_storage_bytes_written = 0;
1983 w->io_cancelled_write_bytes = 0;
1989 void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) {
1992 if(unlikely(!w->fds))
1993 w->fds = callocz(sizeof(int), (size_t) all_files_size);
1995 if(likely(p->updated)) {
1996 w->cutime += p->cutime;
1997 w->cstime += p->cstime;
1998 w->cgtime += p->cgtime;
1999 w->cminflt += p->cminflt;
2000 w->cmajflt += p->cmajflt;
2002 w->utime += p->utime;
2003 w->stime += p->stime;
2004 w->gtime += p->gtime;
2005 w->minflt += p->minflt;
2006 w->majflt += p->majflt;
2010 w->statm_size += p->statm_size;
2011 w->statm_resident += p->statm_resident;
2012 w->statm_share += p->statm_share;
2013 w->statm_text += p->statm_text;
2014 w->statm_lib += p->statm_lib;
2015 w->statm_data += p->statm_data;
2016 w->statm_dirty += p->statm_dirty;
2018 w->io_logical_bytes_read += p->io_logical_bytes_read;
2019 w->io_logical_bytes_written += p->io_logical_bytes_written;
2020 w->io_read_calls += p->io_read_calls;
2021 w->io_write_calls += p->io_write_calls;
2022 w->io_storage_bytes_read += p->io_storage_bytes_read;
2023 w->io_storage_bytes_written += p->io_storage_bytes_written;
2024 w->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
2027 w->num_threads += p->num_threads;
2029 if(likely(w->fds)) {
2031 for(c = 0; c < p->fds_size ;c++) {
2032 if(p->fds[c] == 0) continue;
2034 if(likely(p->fds[c] < all_files_size)) {
2035 if(w->fds) w->fds[p->fds[c]]++;
2038 error("Invalid fd number %d", p->fds[c]);
2042 if(unlikely(debug || w->debug))
2043 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);
2047 void count_targets_fds(struct target *root) {
2051 for (w = root; w ; w = w->next) {
2052 if(!w->fds) continue;
2057 w->openinotifies = 0;
2058 w->openeventfds = 0;
2059 w->opentimerfds = 0;
2060 w->opensignalfds = 0;
2061 w->openeventpolls = 0;
2064 for(c = 1; c < all_files_size ;c++) {
2066 switch(all_files[c].type) {
2075 case FILETYPE_SOCKET:
2079 case FILETYPE_INOTIFY:
2083 case FILETYPE_EVENTFD:
2087 case FILETYPE_TIMERFD:
2091 case FILETYPE_SIGNALFD:
2095 case FILETYPE_EVENTPOLL:
2096 w->openeventpolls++;
2109 void calculate_netdata_statistics(void) {
2110 apply_apps_groups_targets_inheritance();
2112 zero_all_targets(users_root_target);
2113 zero_all_targets(groups_root_target);
2114 apps_groups_targets = zero_all_targets(apps_groups_root_target);
2116 // this has to be done, before the cleanup
2117 struct pid_stat *p = NULL;
2118 struct target *w = NULL, *o = NULL;
2120 // concentrate everything on the apps_groups_targets
2121 for(p = root_of_pids; p ; p = p->next) {
2123 // --------------------------------------------------------------------
2124 // apps_groups targets
2125 if(likely(p->target))
2126 aggregate_pid_on_target(p->target, p, NULL);
2128 error("pid %d %s was left without a target!", p->pid, p->comm);
2131 // --------------------------------------------------------------------
2134 if(likely(p->user_target && p->user_target->uid == p->uid))
2137 if(unlikely(debug && p->user_target))
2138 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);
2140 w = p->user_target = get_users_target(p->uid);
2144 aggregate_pid_on_target(w, p, o);
2146 error("pid %d %s was left without a user target!", p->pid, p->comm);
2149 // --------------------------------------------------------------------
2151 o = p->group_target;
2152 if(likely(p->group_target && p->group_target->gid == p->gid))
2153 w = p->group_target;
2155 if(unlikely(debug && p->group_target))
2156 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);
2158 w = p->group_target = get_groups_target(p->gid);
2162 aggregate_pid_on_target(w, p, o);
2164 error("pid %d %s was left without a group target!", p->pid, p->comm);
2168 count_targets_fds(apps_groups_root_target);
2169 count_targets_fds(users_root_target);
2170 count_targets_fds(groups_root_target);
2172 cleanup_exited_pids();
2175 // ----------------------------------------------------------------------------
2176 // update chart dimensions
2178 BUFFER *output = NULL;
2179 int print_calculated_number(char *str, calculated_number value) { (void)str; (void)value; return 0; }
2181 static inline void send_BEGIN(const char *type, const char *id, unsigned long long usec) {
2182 // fprintf(stdout, "BEGIN %s.%s %llu\n", type, id, usec);
2183 buffer_strcat(output, "BEGIN ");
2184 buffer_strcat(output, type);
2185 buffer_strcat(output, ".");
2186 buffer_strcat(output, id);
2187 buffer_strcat(output, " ");
2188 buffer_print_llu(output, usec);
2189 buffer_strcat(output, "\n");
2192 static inline void send_SET(const char *name, unsigned long long value) {
2193 // fprintf(stdout, "SET %s = %llu\n", name, value);
2194 buffer_strcat(output, "SET ");
2195 buffer_strcat(output, name);
2196 buffer_strcat(output, " = ");
2197 buffer_print_llu(output, value);
2198 buffer_strcat(output, "\n");
2201 static inline void send_END(void) {
2202 // fprintf(stdout, "END\n");
2203 buffer_strcat(output, "END\n");
2206 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;
2207 double minflt_fix_ratio = 1.0, majflt_fix_ratio = 1.0, cminflt_fix_ratio = 1.0, cmajflt_fix_ratio = 1.0;
2209 unsigned long long send_resource_usage_to_netdata() {
2210 static struct timeval last = { 0, 0 };
2211 static struct rusage me_last;
2216 unsigned long long usec;
2217 unsigned long long cpuuser;
2218 unsigned long long cpusyst;
2221 gettimeofday(&last, NULL);
2222 getrusage(RUSAGE_SELF, &me_last);
2224 // the first time, give a zero to allow
2225 // netdata calibrate to the current time
2226 // usec = update_every * 1000000ULL;
2232 gettimeofday(&now, NULL);
2233 getrusage(RUSAGE_SELF, &me);
2235 usec = usec_dt(&now, &last);
2236 cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
2237 cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
2239 bcopy(&now, &last, sizeof(struct timeval));
2240 bcopy(&me, &me_last, sizeof(struct rusage));
2243 buffer_sprintf(output,
2244 "BEGIN netdata.apps_cpu %llu\n"
2246 "SET system = %llu\n"
2248 "BEGIN netdata.apps_files %llu\n"
2249 "SET files = %llu\n"
2252 "SET targets = %ld\n"
2254 "BEGIN netdata.apps_fix %llu\n"
2255 "SET utime = %llu\n"
2256 "SET stime = %llu\n"
2257 "SET gtime = %llu\n"
2258 "SET minflt = %llu\n"
2259 "SET majflt = %llu\n"
2268 , apps_groups_targets
2270 , (unsigned long long)(utime_fix_ratio * 100 * RATES_DETAIL)
2271 , (unsigned long long)(stime_fix_ratio * 100 * RATES_DETAIL)
2272 , (unsigned long long)(gtime_fix_ratio * 100 * RATES_DETAIL)
2273 , (unsigned long long)(minflt_fix_ratio * 100 * RATES_DETAIL)
2274 , (unsigned long long)(majflt_fix_ratio * 100 * RATES_DETAIL)
2277 if(include_exited_childs)
2278 buffer_sprintf(output,
2279 "BEGIN netdata.apps_children_fix %llu\n"
2280 "SET cutime = %llu\n"
2281 "SET cstime = %llu\n"
2282 "SET cgtime = %llu\n"
2283 "SET cminflt = %llu\n"
2284 "SET cmajflt = %llu\n"
2287 , (unsigned long long)(cutime_fix_ratio * 100 * RATES_DETAIL)
2288 , (unsigned long long)(cstime_fix_ratio * 100 * RATES_DETAIL)
2289 , (unsigned long long)(cgtime_fix_ratio * 100 * RATES_DETAIL)
2290 , (unsigned long long)(cminflt_fix_ratio * 100 * RATES_DETAIL)
2291 , (unsigned long long)(cmajflt_fix_ratio * 100 * RATES_DETAIL)
2297 void normalize_data(struct target *root) {
2300 // childs processing introduces spikes
2301 // here we try to eliminate them by disabling childs processing either for specific dimensions
2302 // or entirely. Of course, either way, we disable it just a single iteration.
2304 unsigned long long max = processors * hz * RATES_DETAIL;
2305 unsigned long long utime = 0, cutime = 0, stime = 0, cstime = 0, gtime = 0, cgtime = 0, minflt = 0, cminflt = 0, majflt = 0, cmajflt = 0;
2307 if(global_utime > max) global_utime = max;
2308 if(global_stime > max) global_stime = max;
2309 if(global_gtime > max) global_gtime = max;
2311 for(w = root; w ; w = w->next) {
2312 if(w->target || (!w->processes && !w->exposed)) continue;
2317 cutime += w->cutime;
2318 cstime += w->cstime;
2319 cgtime += w->cgtime;
2321 minflt += w->minflt;
2322 majflt += w->majflt;
2323 cminflt += w->cminflt;
2324 cmajflt += w->cmajflt;
2327 if((global_utime || global_stime || global_gtime) && (utime || stime || gtime)) {
2328 if(global_utime + global_stime + global_gtime > utime + cutime + stime + cstime + gtime + cgtime) {
2329 // everything we collected fits
2335 cgtime_fix_ratio = 1.0; //(double)(global_utime + global_stime) / (double)(utime + cutime + stime + cstime);
2337 else if(global_utime + global_stime > utime + stime) {
2338 // childrens resources are too high
2339 // lower only the children resources
2342 gtime_fix_ratio = 1.0;
2345 cgtime_fix_ratio = (double)((global_utime + global_stime) - (utime + stime)) / (double)(cutime + cstime);
2348 // even running processes are unrealistic
2349 // zero the children resources
2350 // lower the running processes resources
2353 gtime_fix_ratio = (double)(global_utime + global_stime) / (double)(utime + stime);
2356 cgtime_fix_ratio = 0.0;
2365 cgtime_fix_ratio = 0.0;
2368 if(utime_fix_ratio > 1.0) utime_fix_ratio = 1.0;
2369 if(cutime_fix_ratio > 1.0) cutime_fix_ratio = 1.0;
2370 if(stime_fix_ratio > 1.0) stime_fix_ratio = 1.0;
2371 if(cstime_fix_ratio > 1.0) cstime_fix_ratio = 1.0;
2372 if(gtime_fix_ratio > 1.0) gtime_fix_ratio = 1.0;
2373 if(cgtime_fix_ratio > 1.0) cgtime_fix_ratio = 1.0;
2375 // if(utime_fix_ratio < 0.0) utime_fix_ratio = 0.0;
2376 // if(cutime_fix_ratio < 0.0) cutime_fix_ratio = 0.0;
2377 // if(stime_fix_ratio < 0.0) stime_fix_ratio = 0.0;
2378 // if(cstime_fix_ratio < 0.0) cstime_fix_ratio = 0.0;
2379 // if(gtime_fix_ratio < 0.0) gtime_fix_ratio = 0.0;
2380 // if(cgtime_fix_ratio < 0.0) cgtime_fix_ratio = 0.0;
2383 // we use cpu time to normalize page faults
2384 // the problem is that to find the proper max values
2385 // for page faults we have to parse /proc/vmstat
2386 // which is quite big to do it again (netdata does it already)
2388 // a better solution could be to somehow have netdata
2389 // do this normalization for us
2391 if(utime || stime || gtime)
2393 minflt_fix_ratio = (double)(utime * utime_fix_ratio + stime * stime_fix_ratio + gtime * gtime_fix_ratio) / (double)(utime + stime + gtime);
2396 majflt_fix_ratio = 1.0;
2398 if(cutime || cstime || cgtime)
2400 cminflt_fix_ratio = (double)(cutime * cutime_fix_ratio + cstime * cstime_fix_ratio + cgtime * cgtime_fix_ratio) / (double)(cutime + cstime + cgtime);
2403 cmajflt_fix_ratio = 1.0;
2407 if(unlikely(debug)) {
2409 "SYSTEM: u=%llu s=%llu g=%llu "
2410 "COLLECTED: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2411 "DELTA: u=%lld s=%lld g=%lld "
2412 "FIX: u=%0.2f s=%0.2f g=%0.2f cu=%0.2f cs=%0.2f cg=%0.2f "
2413 "FINALLY: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2424 , (long long)utime + (long long)cutime - (long long)global_utime
2425 , (long long)stime + (long long)cstime - (long long)global_stime
2426 , (long long)gtime + (long long)cgtime - (long long)global_gtime
2433 , (unsigned long long)(utime * utime_fix_ratio)
2434 , (unsigned long long)(stime * stime_fix_ratio)
2435 , (unsigned long long)(gtime * gtime_fix_ratio)
2436 , (unsigned long long)(cutime * cutime_fix_ratio)
2437 , (unsigned long long)(cstime * cstime_fix_ratio)
2438 , (unsigned long long)(cgtime * cgtime_fix_ratio)
2443 void send_collected_data_to_netdata(struct target *root, const char *type, unsigned long long usec) {
2446 send_BEGIN(type, "cpu", usec);
2447 for (w = root; w ; w = w->next) {
2448 if(unlikely(w->exposed))
2449 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));
2453 send_BEGIN(type, "cpu_user", usec);
2454 for (w = root; w ; w = w->next) {
2455 if(unlikely(w->exposed))
2456 send_SET(w->name, (unsigned long long)(w->utime * utime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cutime * cutime_fix_ratio)):0ULL));
2460 send_BEGIN(type, "cpu_system", usec);
2461 for (w = root; w ; w = w->next) {
2462 if(unlikely(w->exposed))
2463 send_SET(w->name, (unsigned long long)(w->stime * stime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cstime * cstime_fix_ratio)):0ULL));
2467 if(show_guest_time) {
2468 send_BEGIN(type, "cpu_guest", usec);
2469 for (w = root; w ; w = w->next) {
2470 if(unlikely(w->exposed))
2471 send_SET(w->name, (unsigned long long)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cgtime * cgtime_fix_ratio)):0ULL));
2476 send_BEGIN(type, "threads", usec);
2477 for (w = root; w ; w = w->next) {
2478 if(unlikely(w->exposed))
2479 send_SET(w->name, w->num_threads);
2483 send_BEGIN(type, "processes", usec);
2484 for (w = root; w ; w = w->next) {
2485 if(unlikely(w->exposed))
2486 send_SET(w->name, w->processes);
2490 send_BEGIN(type, "mem", usec);
2491 for (w = root; w ; w = w->next) {
2492 if(unlikely(w->exposed))
2493 send_SET(w->name, (w->statm_resident > w->statm_share)?(w->statm_resident - w->statm_share):0ULL);
2497 send_BEGIN(type, "minor_faults", usec);
2498 for (w = root; w ; w = w->next) {
2499 if(unlikely(w->exposed))
2500 send_SET(w->name, (unsigned long long)(w->minflt * minflt_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cminflt * cminflt_fix_ratio)):0ULL));
2504 send_BEGIN(type, "major_faults", usec);
2505 for (w = root; w ; w = w->next) {
2506 if(unlikely(w->exposed))
2507 send_SET(w->name, (unsigned long long)(w->majflt * majflt_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cmajflt * cmajflt_fix_ratio)):0ULL));
2511 send_BEGIN(type, "lreads", usec);
2512 for (w = root; w ; w = w->next) {
2513 if(unlikely(w->exposed))
2514 send_SET(w->name, w->io_logical_bytes_read);
2518 send_BEGIN(type, "lwrites", usec);
2519 for (w = root; w ; w = w->next) {
2520 if(unlikely(w->exposed))
2521 send_SET(w->name, w->io_logical_bytes_written);
2525 send_BEGIN(type, "preads", usec);
2526 for (w = root; w ; w = w->next) {
2527 if(unlikely(w->exposed))
2528 send_SET(w->name, w->io_storage_bytes_read);
2532 send_BEGIN(type, "pwrites", usec);
2533 for (w = root; w ; w = w->next) {
2534 if(unlikely(w->exposed))
2535 send_SET(w->name, w->io_storage_bytes_written);
2539 if(enable_file_charts) {
2540 send_BEGIN(type, "files", usec);
2541 for (w = root; w; w = w->next) {
2542 if (unlikely(w->exposed))
2543 send_SET(w->name, w->openfiles);
2547 send_BEGIN(type, "sockets", usec);
2548 for (w = root; w; w = w->next) {
2549 if (unlikely(w->exposed))
2550 send_SET(w->name, w->opensockets);
2554 send_BEGIN(type, "pipes", usec);
2555 for (w = root; w; w = w->next) {
2556 if (unlikely(w->exposed))
2557 send_SET(w->name, w->openpipes);
2564 // ----------------------------------------------------------------------------
2565 // generate the charts
2567 void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title)
2570 int newly_added = 0;
2572 for(w = root ; w ; w = w->next) {
2573 if (w->target) continue;
2575 if (!w->exposed && w->processes) {
2578 if (debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
2582 // nothing more to show
2583 if(!newly_added && show_guest_time == show_guest_time_old) return;
2585 // we have something new to show
2586 // update the charts
2587 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);
2588 for (w = root; w ; w = w->next) {
2589 if(unlikely(w->exposed))
2590 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu %s\n", w->name, hz * RATES_DETAIL / 100, w->hidden ? "hidden" : "");
2593 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);
2594 for (w = root; w ; w = w->next) {
2595 if(unlikely(w->exposed))
2596 buffer_sprintf(output, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
2599 buffer_sprintf(output, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20005 %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.processes '' '%s Processes' 'processes' processes %s.processes stacked 20004 %d\n", type, title, type, update_every);
2606 for (w = root; w ; w = w->next) {
2607 if(unlikely(w->exposed))
2608 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2611 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);
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 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);
2618 for (w = root; w ; w = w->next) {
2619 if(unlikely(w->exposed))
2620 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2623 if(show_guest_time) {
2624 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);
2625 for (w = root; w; w = w->next) {
2626 if(unlikely(w->exposed))
2627 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2631 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);
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.minor_faults '' '%s Minor Page Faults' 'page faults/s' mem %s.minor_faults stacked 20011 %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, RATES_DETAIL);
2643 buffer_sprintf(output, "CHART %s.lreads '' '%s Disk Logical Reads' 'kilobytes/s' disk %s.lreads 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.lwrites '' '%s I/O Logical Writes' 'kilobytes/s' disk %s.lwrites stacked 20042 %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.preads '' '%s Disk Reads' 'kilobytes/s' disk %s.preads 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 buffer_sprintf(output, "CHART %s.pwrites '' '%s Disk Writes' 'kilobytes/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
2662 for (w = root; w ; w = w->next) {
2663 if(unlikely(w->exposed))
2664 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2667 if(enable_file_charts) {
2668 buffer_sprintf(output, "CHART %s.files '' '%s Open Files' 'open files' disk %s.files stacked 20050 %d\n", type,
2669 title, type, update_every);
2670 for (w = root; w; w = w->next) {
2671 if (unlikely(w->exposed))
2672 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2675 buffer_sprintf(output, "CHART %s.sockets '' '%s Open Sockets' 'open sockets' net %s.sockets stacked 20051 %d\n",
2676 type, title, type, update_every);
2677 for (w = root; w; w = w->next) {
2678 if (unlikely(w->exposed))
2679 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2682 buffer_sprintf(output, "CHART %s.pipes '' '%s Pipes' 'open pipes' processes %s.pipes stacked 20053 %d\n", type,
2683 title, type, update_every);
2684 for (w = root; w; w = w->next) {
2685 if (unlikely(w->exposed))
2686 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2692 // ----------------------------------------------------------------------------
2693 // parse command line arguments
2695 void parse_args(int argc, char **argv)
2700 for(i = 1; i < argc; i++) {
2702 int n = atoi(argv[i]);
2709 if(strcmp("debug", argv[i]) == 0) {
2711 // debug_flags = 0xffffffff;
2715 if(strcmp("no-childs", argv[i]) == 0 || strcmp("without-childs", argv[i]) == 0) {
2716 include_exited_childs = 0;
2720 if(strcmp("with-childs", argv[i]) == 0) {
2721 include_exited_childs = 1;
2725 if(strcmp("with-guest", argv[i]) == 0) {
2726 enable_guest_charts = 1;
2730 if(strcmp("no-guest", argv[i]) == 0 || strcmp("without-guest", argv[i]) == 0) {
2731 enable_guest_charts = 0;
2735 if(strcmp("with-files", argv[i]) == 0) {
2736 enable_file_charts = 1;
2740 if(strcmp("no-files", argv[i]) == 0 || strcmp("without-files", argv[i]) == 0) {
2741 enable_file_charts = 0;
2745 if(strcmp("-h", argv[i]) == 0 || strcmp("--help", argv[i]) == 0) {
2748 "(C) 2016 Costa Tsaousis"
2750 "This program is a data collector plugin for netdata.\n"
2752 "Valid command line options:\n"
2754 "SECONDS set the data collection frequency\n"
2756 "debug enable debugging (lot of output)\n"
2759 "without-childs enable / disable aggregating exited\n"
2760 " children resources into parents\n"
2761 " (default is enabled)\n"
2764 "without-guest enable / disable reporting guest charts\n"
2765 " (default is disabled)\n"
2768 "without-files enable / disable reporting files, sockets, pipes\n"
2769 " (default is enabled)\n"
2771 "NAME read apps_NAME.conf instead of\n"
2772 " apps_groups.conf\n"
2773 " (default NAME=groups)\n"
2783 error("Cannot understand option %s", argv[i]);
2787 if(freq > 0) update_every = freq;
2788 if(!name) name = "groups";
2790 if(read_apps_groups_conf(name)) {
2791 error("Cannot read process groups %s", name);
2796 int main(int argc, char **argv)
2798 // debug_flags = D_PROCFILE;
2800 // set the name for logging
2801 program_name = "apps.plugin";
2803 info("started on pid %d", getpid());
2805 // disable syslog for apps.plugin
2806 error_log_syslog = 0;
2808 // set errors flood protection to 100 logs per hour
2809 error_log_errors_per_period = 100;
2810 error_log_throttle_period = 3600;
2812 host_prefix = getenv("NETDATA_HOST_PREFIX");
2813 if(host_prefix == NULL) {
2814 // info("NETDATA_HOST_PREFIX is not passed from netdata");
2817 // else info("Found NETDATA_HOST_PREFIX='%s'", host_prefix);
2819 config_dir = getenv("NETDATA_CONFIG_DIR");
2820 if(config_dir == NULL) {
2821 // info("NETDATA_CONFIG_DIR is not passed from netdata");
2822 config_dir = CONFIG_DIR;
2824 // else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
2826 #ifdef NETDATA_INTERNAL_CHECKS
2827 if(debug_flags != 0) {
2828 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
2829 if(setrlimit(RLIMIT_CORE, &rl) != 0)
2830 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
2831 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
2833 #endif /* NETDATA_INTERNAL_CHECKS */
2835 procfile_adaptive_initial_allocation = 1;
2837 time_t started_t = time(NULL);
2840 pid_max = get_system_pid_max();
2841 processors = get_system_cpus();
2843 parse_args(argc, argv);
2845 all_pids_sortlist = callocz(sizeof(pid_t), (size_t)pid_max);
2846 all_pids = callocz(sizeof(struct pid_stat *), (size_t) pid_max);
2848 output = buffer_create(1024);
2849 buffer_sprintf(output,
2850 "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %1$d\n"
2851 "DIMENSION user '' incremental 1 1000\n"
2852 "DIMENSION system '' incremental 1 1000\n"
2853 "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %1$d\n"
2854 "DIMENSION files '' incremental 1 1\n"
2855 "DIMENSION pids '' absolute 1 1\n"
2856 "DIMENSION fds '' absolute 1 1\n"
2857 "DIMENSION targets '' absolute 1 1\n"
2858 "CHART netdata.apps_fix '' 'Apps Plugin Normalization Ratios' 'percentage' apps.plugin netdata.apps_fix line 140002 %1$d\n"
2859 "DIMENSION utime '' absolute 1 %2$llu\n"
2860 "DIMENSION stime '' absolute 1 %2$llu\n"
2861 "DIMENSION gtime '' absolute 1 %2$llu\n"
2862 "DIMENSION minflt '' absolute 1 %2$llu\n"
2863 "DIMENSION majflt '' absolute 1 %2$llu\n"
2868 if(include_exited_childs)
2869 buffer_sprintf(output,
2870 "CHART netdata.apps_children_fix '' 'Apps Plugin Exited Children Normalization Ratios' 'percentage' apps.plugin netdata.apps_children_fix line 140003 %1$d\n"
2871 "DIMENSION cutime '' absolute 1 %2$llu\n"
2872 "DIMENSION cstime '' absolute 1 %2$llu\n"
2873 "DIMENSION cgtime '' absolute 1 %2$llu\n"
2874 "DIMENSION cminflt '' absolute 1 %2$llu\n"
2875 "DIMENSION cmajflt '' absolute 1 %2$llu\n"
2880 #ifndef PROFILING_MODE
2881 unsigned long long sunext = (time(NULL) - (time(NULL) % update_every) + update_every) * 1000000ULL;
2882 unsigned long long sunow;
2883 #endif /* PROFILING_MODE */
2885 global_iterations_counter = 1;
2886 for(;1; global_iterations_counter++) {
2887 #ifndef PROFILING_MODE
2888 // delay until it is our time to run
2889 while((sunow = time_usec()) < sunext)
2890 sleep_usec(sunext - sunow);
2892 // find the next time we need to run
2893 while(time_usec() > sunext)
2894 sunext += update_every * 1000000ULL;
2895 #endif /* PROFILING_MODE */
2897 if(!collect_data_for_all_processes_from_proc()) {
2898 error("Cannot collect /proc data for running processes. Disabling apps.plugin...");
2899 printf("DISABLE\n");
2903 calculate_netdata_statistics();
2904 normalize_data(apps_groups_root_target);
2906 unsigned long long dt = send_resource_usage_to_netdata();
2908 // this is smart enough to show only newly added apps, when needed
2909 send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps");
2910 send_charts_updates_to_netdata(users_root_target, "users", "Users");
2911 send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups");
2913 send_collected_data_to_netdata(apps_groups_root_target, "apps", dt);
2914 send_collected_data_to_netdata(users_root_target, "users", dt);
2915 send_collected_data_to_netdata(groups_root_target, "groups", dt);
2917 show_guest_time_old = show_guest_time;
2919 //if(puts(buffer_tostring(output)) == EOF)
2920 if(write(STDOUT_FILENO, buffer_tostring(output), buffer_strlen(output)) == -1)
2921 fatal("Cannot send chart values to netdata.");
2924 buffer_flush(output);
2927 fprintf(stderr, "apps.plugin: done Loop No %llu\n", global_iterations_counter);
2929 current_t = time(NULL);
2931 #ifndef PROFILING_MODE
2932 // restart check (14400 seconds)
2933 if(current_t - started_t > 14400) exit(0);
2935 if(current_t - started_t > 10) exit(0);
2936 #endif /* PROFILING_MODE */