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;
36 // ----------------------------------------------------------------------------
38 void netdata_cleanup_and_exit(int ret) {
43 // ----------------------------------------------------------------------------
45 // to retrieve settings of the system
47 long get_system_cpus(void) {
52 char filename[FILENAME_MAX + 1];
53 snprintfz(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
55 ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
58 ff = procfile_readall(ff);
65 for(i = 0; i < procfile_lines(ff); i++) {
66 if(!procfile_linewords(ff, i)) continue;
68 if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
71 if(processors < 1) processors = 1;
77 pid_t get_system_pid_max(void) {
81 char filename[FILENAME_MAX + 1];
82 snprintfz(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
83 ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
86 ff = procfile_readall(ff);
92 mpid = (pid_t)atoi(procfile_lineword(ff, 0, 0));
93 if(!mpid) mpid = 32768;
99 // ----------------------------------------------------------------------------
101 // target is the structure that process data are aggregated
104 char compare[MAX_COMPARE_NAME + 1];
105 uint32_t comparehash;
108 char id[MAX_NAME + 1];
111 char name[MAX_NAME + 1];
116 unsigned long long minflt;
117 unsigned long long cminflt;
118 unsigned long long majflt;
119 unsigned long long cmajflt;
120 unsigned long long utime;
121 unsigned long long stime;
122 unsigned long long gtime;
123 unsigned long long cutime;
124 unsigned long long cstime;
125 unsigned long long cgtime;
126 unsigned long long num_threads;
127 unsigned long long rss;
129 unsigned long long statm_size;
130 unsigned long long statm_resident;
131 unsigned long long statm_share;
132 unsigned long long statm_text;
133 unsigned long long statm_lib;
134 unsigned long long statm_data;
135 unsigned long long statm_dirty;
137 unsigned long long io_logical_bytes_read;
138 unsigned long long io_logical_bytes_written;
139 unsigned long long io_read_calls;
140 unsigned long long io_write_calls;
141 unsigned long long io_storage_bytes_read;
142 unsigned long long io_storage_bytes_written;
143 unsigned long long io_cancelled_write_bytes;
146 unsigned long long openfiles;
147 unsigned long long openpipes;
148 unsigned long long opensockets;
149 unsigned long long openinotifies;
150 unsigned long long openeventfds;
151 unsigned long long opentimerfds;
152 unsigned long long opensignalfds;
153 unsigned long long openeventpolls;
154 unsigned long long openother;
156 unsigned long processes; // how many processes have been merged to this
157 int exposed; // if set, we have sent this to netdata
158 int hidden; // if set, we set the hidden flag on the dimension
161 int starts_with; // if set, the compare string matches only the
162 // beginning of the command
164 struct target *target; // the one that will be reported to netdata
169 // ----------------------------------------------------------------------------
171 // aggregate all processes in groups, to have a limited number of dimensions
173 struct target *apps_groups_root_target = NULL;
174 struct target *apps_groups_default_target = NULL;
175 long apps_groups_targets = 0;
177 struct target *users_root_target = NULL;
178 struct target *groups_root_target = NULL;
180 struct target *get_users_target(uid_t uid)
183 for(w = users_root_target ; w ; w = w->next)
184 if(w->uid == uid) return w;
186 w = callocz(sizeof(struct target), 1);
187 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", uid);
188 w->comparehash = simple_hash(w->compare);
189 w->comparelen = strlen(w->compare);
191 snprintfz(w->id, MAX_NAME, "%u", uid);
192 w->idhash = simple_hash(w->id);
194 struct passwd *pw = getpwuid(uid);
196 snprintfz(w->name, MAX_NAME, "%u", uid);
198 snprintfz(w->name, MAX_NAME, "%s", pw->pw_name);
200 netdata_fix_chart_name(w->name);
204 w->next = users_root_target;
205 users_root_target = w;
208 fprintf(stderr, "apps.plugin: added uid %u ('%s') target\n", w->uid, w->name);
213 struct target *get_groups_target(gid_t gid)
216 for(w = groups_root_target ; w ; w = w->next)
217 if(w->gid == gid) return w;
219 w = callocz(sizeof(struct target), 1);
220 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", gid);
221 w->comparehash = simple_hash(w->compare);
222 w->comparelen = strlen(w->compare);
224 snprintfz(w->id, MAX_NAME, "%u", gid);
225 w->idhash = simple_hash(w->id);
227 struct group *gr = getgrgid(gid);
229 snprintfz(w->name, MAX_NAME, "%u", gid);
231 snprintfz(w->name, MAX_NAME, "%s", gr->gr_name);
233 netdata_fix_chart_name(w->name);
237 w->next = groups_root_target;
238 groups_root_target = w;
241 fprintf(stderr, "apps.plugin: added gid %u ('%s') target\n", w->gid, w->name);
246 // find or create a new target
247 // there are targets that are just aggregated to other target (the second argument)
248 struct target *get_apps_groups_target(const char *id, struct target *target) {
249 int tdebug = 0, thidden = 0, ends_with = 0;
250 const char *nid = id;
252 while(nid[0] == '-' || nid[0] == '+' || nid[0] == '*') {
253 if(nid[0] == '-') thidden = 1;
254 if(nid[0] == '+') tdebug = 1;
255 if(nid[0] == '*') ends_with = 1;
258 uint32_t hash = simple_hash(id);
260 struct target *w, *last = apps_groups_root_target;
261 for(w = apps_groups_root_target ; w ; w = w->next) {
262 if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
268 w = callocz(sizeof(struct target), 1);
269 strncpyz(w->id, nid, MAX_NAME);
270 w->idhash = simple_hash(w->id);
272 strncpyz(w->name, nid, MAX_NAME);
274 strncpyz(w->compare, nid, MAX_COMPARE_NAME);
275 size_t len = strlen(w->compare);
276 if(w->compare[len - 1] == '*') {
277 w->compare[len - 1] = '\0';
280 w->ends_with = ends_with;
282 if(w->starts_with && w->ends_with)
283 proc_pid_cmdline_is_needed = 1;
285 w->comparehash = simple_hash(w->compare);
286 w->comparelen = strlen(w->compare);
292 // append it, to maintain the order in apps_groups.conf
293 if(last) last->next = w;
294 else apps_groups_root_target = w;
297 fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
299 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
300 , w->target?w->target->id:w->id
301 , (w->hidden)?"hidden":"-"
302 , (w->debug)?"debug":"-"
308 // read the apps_groups.conf file
309 int read_apps_groups_conf(const char *name)
311 char filename[FILENAME_MAX + 1];
313 snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
316 fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
318 // ----------------------------------------
320 procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT);
323 procfile_set_quotes(ff, "'\"");
325 ff = procfile_readall(ff);
331 unsigned long line, lines = procfile_lines(ff);
333 for(line = 0; line < lines ;line++) {
334 unsigned long word, words = procfile_linewords(ff, line);
335 struct target *w = NULL;
337 char *t = procfile_lineword(ff, line, 0);
338 if(!t || !*t) continue;
340 for(word = 0; word < words ;word++) {
341 char *s = procfile_lineword(ff, line, word);
342 if(!s || !*s) continue;
347 struct target *n = get_apps_groups_target(s, w);
349 error("Cannot create target '%s' (line %lu, word %lu)", s, line, word);
357 int tdebug = 0, thidden = 0;
359 while(t[0] == '-' || t[0] == '+') {
360 if(t[0] == '-') thidden = 1;
361 if(t[0] == '+') tdebug = 1;
365 strncpyz(w->name, t, MAX_NAME);
370 fprintf(stderr, "apps.plugin: AGGREGATION TARGET NAME '%s' on ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
373 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
374 , w->target?w->target->id:w->id
375 , (w->hidden)?"hidden":"-"
376 , (w->debug)?"debug":"-"
383 apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
384 if(!apps_groups_default_target)
385 error("Cannot create default target");
387 strncpyz(apps_groups_default_target->name, "other", MAX_NAME);
393 // ----------------------------------------------------------------------------
394 // data to store for each pid
397 #define PID_LOG_IO 0x00000001
398 #define PID_LOG_STATM 0x00000002
399 #define PID_LOG_CMDLINE 0x00000004
400 #define PID_LOG_FDS 0x00000008
401 #define PID_LOG_STAT 0x00000010
405 char comm[MAX_COMPARE_NAME + 1];
406 char cmdline[MAX_CMDLINE + 1];
418 // these are raw values collected
419 unsigned long long minflt_raw;
420 unsigned long long cminflt_raw;
421 unsigned long long majflt_raw;
422 unsigned long long cmajflt_raw;
423 unsigned long long utime_raw;
424 unsigned long long stime_raw;
425 unsigned long long gtime_raw; // guest_time
426 unsigned long long cutime_raw;
427 unsigned long long cstime_raw;
428 unsigned long long cgtime_raw; // cguest_time
431 unsigned long long minflt;
432 unsigned long long cminflt;
433 unsigned long long majflt;
434 unsigned long long cmajflt;
435 unsigned long long utime;
436 unsigned long long stime;
437 unsigned long long gtime;
438 unsigned long long cutime;
439 unsigned long long cstime;
440 unsigned long long cgtime;
445 // int64_t itrealvalue;
446 // unsigned long long starttime;
447 // unsigned long long vsize;
448 unsigned long long rss;
449 // unsigned long long rsslim;
450 // unsigned long long starcode;
451 // unsigned long long endcode;
452 // unsigned long long startstack;
453 // unsigned long long kstkesp;
454 // unsigned long long kstkeip;
457 // uint64_t sigignore;
458 // uint64_t sigcatch;
462 // int32_t exit_signal;
463 // int32_t processor;
464 // uint32_t rt_priority;
466 // unsigned long long delayacct_blkio_ticks;
471 unsigned long long statm_size;
472 unsigned long long statm_resident;
473 unsigned long long statm_share;
474 unsigned long long statm_text;
475 unsigned long long statm_lib;
476 unsigned long long statm_data;
477 unsigned long long statm_dirty;
479 unsigned long long io_logical_bytes_read_raw;
480 unsigned long long io_logical_bytes_written_raw;
481 unsigned long long io_read_calls_raw;
482 unsigned long long io_write_calls_raw;
483 unsigned long long io_storage_bytes_read_raw;
484 unsigned long long io_storage_bytes_written_raw;
485 unsigned long long io_cancelled_write_bytes_raw;
487 unsigned long long io_logical_bytes_read;
488 unsigned long long io_logical_bytes_written;
489 unsigned long long io_read_calls;
490 unsigned long long io_write_calls;
491 unsigned long long io_storage_bytes_read;
492 unsigned long long io_storage_bytes_written;
493 unsigned long long io_cancelled_write_bytes;
495 int *fds; // array of fds it uses
496 int fds_size; // the size of the fds array
498 int children_count; // number of processes directly referencing this
499 int keep; // 1 when we need to keep this process in memory even after it exited
500 int keeploops; // increases by 1 every time keep is 1 and updated 0
501 int updated; // 1 when the process is currently running
502 int merged; // 1 when it has been merged to its parent
503 int new_entry; // 1 when this is a new process, just saw for the first time
504 int read; // 1 when we have already read this process for this iteration
505 int sortlist; // higher numbers = top on the process tree
506 // each process gets a unique number
508 struct target *target; // app_groups.conf targets
509 struct target *user_target; // uid based targets
510 struct target *group_target; // gid based targets
512 unsigned long long stat_collected_usec;
513 unsigned long long last_stat_collected_usec;
515 unsigned long long io_collected_usec;
516 unsigned long long last_io_collected_usec;
519 char *statm_filename;
521 char *cmdline_filename;
523 struct pid_stat *parent;
524 struct pid_stat *prev;
525 struct pid_stat *next;
526 } *root_of_pids = NULL, **all_pids;
528 long all_pids_count = 0;
530 struct pid_stat *get_pid_entry(pid_t pid) {
532 all_pids[pid]->new_entry = 0;
533 return all_pids[pid];
536 all_pids[pid] = callocz(sizeof(struct pid_stat), 1);
537 all_pids[pid]->fds = callocz(sizeof(int), 100);
538 all_pids[pid]->fds_size = 100;
540 if(root_of_pids) root_of_pids->prev = all_pids[pid];
541 all_pids[pid]->next = root_of_pids;
542 root_of_pids = all_pids[pid];
544 all_pids[pid]->pid = pid;
545 all_pids[pid]->new_entry = 1;
549 return all_pids[pid];
552 void del_pid_entry(pid_t pid) {
554 error("attempted to free pid %d that is not allocated.", pid);
559 fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
561 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
562 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
563 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
565 if(all_pids[pid]->fds) freez(all_pids[pid]->fds);
566 if(all_pids[pid]->stat_filename) freez(all_pids[pid]->stat_filename);
567 if(all_pids[pid]->statm_filename) freez(all_pids[pid]->statm_filename);
568 if(all_pids[pid]->io_filename) freez(all_pids[pid]->io_filename);
569 if(all_pids[pid]->cmdline_filename) freez(all_pids[pid]->cmdline_filename);
570 freez(all_pids[pid]);
572 all_pids[pid] = NULL;
577 // ----------------------------------------------------------------------------
578 // update pids from proc
580 int read_proc_pid_cmdline(struct pid_stat *p) {
582 if(unlikely(!p->cmdline_filename)) {
583 char filename[FILENAME_MAX + 1];
584 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", host_prefix, p->pid);
585 p->cmdline_filename = strdupz(filename);
588 int fd = open(p->cmdline_filename, O_RDONLY, 0666);
589 if(unlikely(fd == -1)) goto cleanup;
591 ssize_t i, bytes = read(fd, p->cmdline, MAX_CMDLINE);
594 if(unlikely(bytes < 0)) goto cleanup;
596 p->cmdline[bytes] = '\0';
597 for(i = 0; i < bytes ; i++)
598 if(unlikely(!p->cmdline[i])) p->cmdline[i] = ' ';
601 fprintf(stderr, "Read file '%s' contents: %s\n", p->cmdline_filename, p->cmdline);
606 // copy the command to the command line
607 strncpyz(p->cmdline, p->comm, MAX_CMDLINE);
611 int read_proc_pid_ownership(struct pid_stat *p) {
612 if(unlikely(!p->stat_filename)) {
613 error("pid %d does not have a stat_filename", p->pid);
617 // ----------------------------------------
621 if(stat(p->stat_filename, &st) != 0) {
622 error("Cannot stat file '%s'", p->stat_filename);
632 int read_proc_pid_stat(struct pid_stat *p) {
633 static procfile *ff = NULL;
635 if(unlikely(!p->stat_filename)) {
636 char filename[FILENAME_MAX + 1];
637 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
638 p->stat_filename = strdupz(filename);
641 int set_quotes = (!ff)?1:0;
643 ff = procfile_reopen(ff, p->stat_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
644 if(unlikely(!ff)) goto cleanup;
646 // if(set_quotes) procfile_set_quotes(ff, "()");
647 if(set_quotes) procfile_set_open_close(ff, "(", ")");
649 ff = procfile_readall(ff);
650 if(unlikely(!ff)) goto cleanup;
652 p->last_stat_collected_usec = p->stat_collected_usec;
653 p->stat_collected_usec = time_usec();
656 // p->pid = atol(procfile_lineword(ff, 0, 0+i));
658 strncpyz(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
660 // p->state = *(procfile_lineword(ff, 0, 2));
661 p->ppid = (int32_t) atol(procfile_lineword(ff, 0, 3));
662 // p->pgrp = atol(procfile_lineword(ff, 0, 4));
663 // p->session = atol(procfile_lineword(ff, 0, 5));
664 // p->tty_nr = atol(procfile_lineword(ff, 0, 6));
665 // p->tpgid = atol(procfile_lineword(ff, 0, 7));
666 // p->flags = strtoull(procfile_lineword(ff, 0, 8), NULL, 10);
668 unsigned long long last;
670 last = p->minflt_raw;
671 p->minflt_raw = strtoull(procfile_lineword(ff, 0, 9), NULL, 10);
672 p->minflt = (p->minflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
674 last = p->cminflt_raw;
675 p->cminflt_raw = strtoull(procfile_lineword(ff, 0, 10), NULL, 10);
676 p->cminflt = (p->cminflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
678 last = p->majflt_raw;
679 p->majflt_raw = strtoull(procfile_lineword(ff, 0, 11), NULL, 10);
680 p->majflt = (p->majflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
682 last = p->cmajflt_raw;
683 p->cmajflt_raw = strtoull(procfile_lineword(ff, 0, 12), NULL, 10);
684 p->cmajflt = (p->cmajflt_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
687 p->utime_raw = strtoull(procfile_lineword(ff, 0, 13), NULL, 10);
688 p->utime = (p->utime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
691 p->stime_raw = strtoull(procfile_lineword(ff, 0, 14), NULL, 10);
692 p->stime = (p->stime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
694 last = p->cutime_raw;
695 p->cutime_raw = strtoull(procfile_lineword(ff, 0, 15), NULL, 10);
696 p->cutime = (p->cutime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
698 last = p->cstime_raw;
699 p->cstime_raw = strtoull(procfile_lineword(ff, 0, 16), NULL, 10);
700 p->cstime = (p->cstime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
702 // p->priority = strtoull(procfile_lineword(ff, 0, 17), NULL, 10);
703 // p->nice = strtoull(procfile_lineword(ff, 0, 18), NULL, 10);
704 p->num_threads = (int32_t) atol(procfile_lineword(ff, 0, 19));
705 // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20), NULL, 10);
706 // p->starttime = strtoull(procfile_lineword(ff, 0, 21), NULL, 10);
707 // p->vsize = strtoull(procfile_lineword(ff, 0, 22), NULL, 10);
708 p->rss = strtoull(procfile_lineword(ff, 0, 23), NULL, 10);
709 // p->rsslim = strtoull(procfile_lineword(ff, 0, 24), NULL, 10);
710 // p->starcode = strtoull(procfile_lineword(ff, 0, 25), NULL, 10);
711 // p->endcode = strtoull(procfile_lineword(ff, 0, 26), NULL, 10);
712 // p->startstack = strtoull(procfile_lineword(ff, 0, 27), NULL, 10);
713 // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28), NULL, 10);
714 // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29), NULL, 10);
715 // p->signal = strtoull(procfile_lineword(ff, 0, 30), NULL, 10);
716 // p->blocked = strtoull(procfile_lineword(ff, 0, 31), NULL, 10);
717 // p->sigignore = strtoull(procfile_lineword(ff, 0, 32), NULL, 10);
718 // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33), NULL, 10);
719 // p->wchan = strtoull(procfile_lineword(ff, 0, 34), NULL, 10);
720 // p->nswap = strtoull(procfile_lineword(ff, 0, 35), NULL, 10);
721 // p->cnswap = strtoull(procfile_lineword(ff, 0, 36), NULL, 10);
722 // p->exit_signal = atol(procfile_lineword(ff, 0, 37));
723 // p->processor = atol(procfile_lineword(ff, 0, 38));
724 // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39), NULL, 10);
725 // p->policy = strtoul(procfile_lineword(ff, 0, 40), NULL, 10);
726 // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41), NULL, 10);
728 if(enable_guest_charts) {
730 p->gtime_raw = strtoull(procfile_lineword(ff, 0, 42), NULL, 10);
731 p->gtime = (p->gtime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
733 last = p->cgtime_raw;
734 p->cgtime_raw = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
735 p->cgtime = (p->cgtime_raw - last) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
737 if (show_guest_time || p->gtime || p->cgtime) {
738 p->utime -= (p->utime >= p->gtime) ? p->gtime : p->utime;
739 p->cutime -= (p->cutime >= p->cgtime) ? p->cgtime : p->cutime;
744 if(unlikely(debug || (p->target && p->target->debug)))
745 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);
747 if(unlikely(global_iterations_counter == 1)) {
778 int read_proc_pid_statm(struct pid_stat *p) {
779 static procfile *ff = NULL;
781 if(unlikely(!p->statm_filename)) {
782 char filename[FILENAME_MAX + 1];
783 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
784 p->statm_filename = strdupz(filename);
787 ff = procfile_reopen(ff, p->statm_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
788 if(unlikely(!ff)) goto cleanup;
790 ff = procfile_readall(ff);
791 if(unlikely(!ff)) goto cleanup;
795 p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
796 p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
797 p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
798 p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
799 p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
800 p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
801 p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
807 p->statm_resident = 0;
816 int read_proc_pid_io(struct pid_stat *p) {
817 static procfile *ff = NULL;
819 if(unlikely(!p->io_filename)) {
820 char filename[FILENAME_MAX + 1];
821 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
822 p->io_filename = strdupz(filename);
826 ff = procfile_reopen(ff, p->io_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
827 if(unlikely(!ff)) goto cleanup;
829 ff = procfile_readall(ff);
830 if(unlikely(!ff)) goto cleanup;
834 p->last_io_collected_usec = p->io_collected_usec;
835 p->io_collected_usec = time_usec();
837 unsigned long long last;
839 last = p->io_logical_bytes_read_raw;
840 p->io_logical_bytes_read_raw = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
841 p->io_logical_bytes_read = (p->io_logical_bytes_read_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
843 last = p->io_logical_bytes_written_raw;
844 p->io_logical_bytes_written_raw = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
845 p->io_logical_bytes_written = (p->io_logical_bytes_written_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
847 last = p->io_read_calls_raw;
848 p->io_read_calls_raw = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
849 p->io_read_calls = (p->io_read_calls_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
851 last = p->io_write_calls_raw;
852 p->io_write_calls_raw = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
853 p->io_write_calls = (p->io_write_calls_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
855 last = p->io_storage_bytes_read_raw;
856 p->io_storage_bytes_read_raw = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
857 p->io_storage_bytes_read = (p->io_storage_bytes_read_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
859 last = p->io_storage_bytes_written_raw;
860 p->io_storage_bytes_written_raw = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
861 p->io_storage_bytes_written = (p->io_storage_bytes_written_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
863 last = p->io_cancelled_write_bytes_raw;
864 p->io_cancelled_write_bytes_raw = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
865 p->io_cancelled_write_bytes = (p->io_cancelled_write_bytes_raw - last) * (1000000ULL * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
867 if(unlikely(global_iterations_counter == 1)) {
868 p->io_logical_bytes_read = 0;
869 p->io_logical_bytes_written = 0;
870 p->io_read_calls = 0;
871 p->io_write_calls = 0;
872 p->io_storage_bytes_read = 0;
873 p->io_storage_bytes_written = 0;
874 p->io_cancelled_write_bytes = 0;
880 p->io_logical_bytes_read = 0;
881 p->io_logical_bytes_written = 0;
882 p->io_read_calls = 0;
883 p->io_write_calls = 0;
884 p->io_storage_bytes_read = 0;
885 p->io_storage_bytes_written = 0;
886 p->io_cancelled_write_bytes = 0;
890 unsigned long long global_utime = 0;
891 unsigned long long global_stime = 0;
892 unsigned long long global_gtime = 0;
894 int read_proc_stat() {
895 static char filename[FILENAME_MAX + 1] = "";
896 static procfile *ff = NULL;
897 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;
900 snprintfz(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
901 ff = procfile_open(filename, " \t:", PROCFILE_FLAG_DEFAULT);
902 if(unlikely(!ff)) goto cleanup;
905 ff = procfile_readall(ff);
906 if(unlikely(!ff)) goto cleanup;
908 last_collected_usec = collected_usec;
909 collected_usec = time_usec();
913 unsigned long long last;
916 utime_raw = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
917 global_utime = (utime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
919 // nice time, on user time
921 ntime_raw = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
922 global_utime += (ntime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
925 stime_raw = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
926 global_stime = (stime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
929 gtime_raw = strtoull(procfile_lineword(ff, 0, 10), NULL, 10);
930 global_gtime = (gtime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
932 if(enable_guest_charts) {
933 // guest nice time, on guest time
935 gntime_raw = strtoull(procfile_lineword(ff, 0, 11), NULL, 10);
936 global_gtime += (gntime_raw - last) * (1000000ULL * RATES_DETAIL) / (collected_usec - last_collected_usec);
938 // remove guest time from user time
939 global_utime -= (global_utime > global_gtime) ? global_gtime : global_utime;
942 if(unlikely(global_iterations_counter == 1)) {
958 // ----------------------------------------------------------------------------
960 // this is used to keep a global list of all open files of the system
961 // it is needed in order to calculate the unique files processes have open
963 #define FILE_DESCRIPTORS_INCREASE_STEP 100
965 struct file_descriptor {
967 #ifdef NETDATA_INTERNAL_CHECKS
969 #endif /* NETDATA_INTERNAL_CHECKS */
977 int all_files_len = 0;
978 int all_files_size = 0;
980 int file_descriptor_compare(void* a, void* b) {
981 #ifdef NETDATA_INTERNAL_CHECKS
982 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
983 error("Corrupted index data detected. Please report this.");
984 #endif /* NETDATA_INTERNAL_CHECKS */
986 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
989 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
993 return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
996 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
998 avl_tree all_files_index = {
1000 file_descriptor_compare
1003 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
1004 struct file_descriptor tmp;
1005 tmp.hash = (hash)?hash:simple_hash(name);
1009 #ifdef NETDATA_INTERNAL_CHECKS
1010 tmp.magic = 0x0BADCAFE;
1011 #endif /* NETDATA_INTERNAL_CHECKS */
1013 return (struct file_descriptor *)avl_search(&all_files_index, (avl *) &tmp);
1016 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
1017 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
1019 #define FILETYPE_OTHER 0
1020 #define FILETYPE_FILE 1
1021 #define FILETYPE_PIPE 2
1022 #define FILETYPE_SOCKET 3
1023 #define FILETYPE_INOTIFY 4
1024 #define FILETYPE_EVENTFD 5
1025 #define FILETYPE_EVENTPOLL 6
1026 #define FILETYPE_TIMERFD 7
1027 #define FILETYPE_SIGNALFD 8
1029 void file_descriptor_not_used(int id)
1031 if(id > 0 && id < all_files_size) {
1033 #ifdef NETDATA_INTERNAL_CHECKS
1034 if(all_files[id].magic != 0x0BADCAFE) {
1035 error("Ignoring request to remove empty file id %d.", id);
1038 #endif /* NETDATA_INTERNAL_CHECKS */
1041 fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
1043 if(all_files[id].count > 0) {
1044 all_files[id].count--;
1046 if(!all_files[id].count) {
1048 fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
1050 file_descriptor_remove(&all_files[id]);
1051 #ifdef NETDATA_INTERNAL_CHECKS
1052 all_files[id].magic = 0x00000000;
1053 #endif /* NETDATA_INTERNAL_CHECKS */
1058 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
1060 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
1063 int file_descriptor_find_or_add(const char *name)
1065 static int last_pos = 0;
1066 uint32_t hash = simple_hash(name);
1069 fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
1071 struct file_descriptor *fd = file_descriptor_find(name, hash);
1075 fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
1082 // check we have enough memory to add it
1083 if(!all_files || all_files_len == all_files_size) {
1084 void *old = all_files;
1087 // there is no empty slot
1089 fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
1091 all_files = reallocz(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
1093 // if the address changed, we have to rebuild the index
1094 // since all pointers are now invalid
1095 if(old && old != (void *)all_files) {
1097 fprintf(stderr, "apps.plugin: >> re-indexing.\n");
1099 all_files_index.root = NULL;
1100 for(i = 0; i < all_files_size; i++) {
1101 if(!all_files[i].count) continue;
1102 file_descriptor_add(&all_files[i]);
1106 fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
1109 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
1110 all_files[i].count = 0;
1111 all_files[i].name = NULL;
1112 #ifdef NETDATA_INTERNAL_CHECKS
1113 all_files[i].magic = 0x00000000;
1114 #endif /* NETDATA_INTERNAL_CHECKS */
1115 all_files[i].pos = i;
1118 if(!all_files_size) all_files_len = 1;
1119 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
1123 fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
1125 // search for an empty slot
1127 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
1128 if(c >= all_files_size) c = 0;
1129 if(c == 0) continue;
1131 if(!all_files[c].count) {
1133 fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
1135 #ifdef NETDATA_INTERNAL_CHECKS
1136 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
1137 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
1138 #endif /* NETDATA_INTERNAL_CHECKS */
1141 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);
1143 if(all_files[c].name) freez((void *)all_files[c].name);
1144 all_files[c].name = NULL;
1149 if(i == all_files_size) {
1150 fatal("We should find an empty slot, but there isn't any");
1155 fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
1159 // else we have an empty slot in 'c'
1162 if(name[0] == '/') type = FILETYPE_FILE;
1163 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
1164 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
1165 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
1166 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
1167 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
1168 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
1169 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
1170 else if(strncmp(name, "anon_inode:", 11) == 0) {
1172 fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
1174 type = FILETYPE_OTHER;
1178 fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
1180 type = FILETYPE_OTHER;
1183 all_files[c].name = strdupz(name);
1184 all_files[c].hash = hash;
1185 all_files[c].type = type;
1186 all_files[c].pos = c;
1187 all_files[c].count = 1;
1188 #ifdef NETDATA_INTERNAL_CHECKS
1189 all_files[c].magic = 0x0BADCAFE;
1190 #endif /* NETDATA_INTERNAL_CHECKS */
1191 file_descriptor_add(&all_files[c]);
1194 fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1199 int read_pid_file_descriptors(struct pid_stat *p) {
1200 char dirname[FILENAME_MAX+1];
1202 snprintfz(dirname, FILENAME_MAX, "%s/proc/%d/fd", host_prefix, p->pid);
1203 DIR *fds = opendir(dirname);
1207 char fdname[FILENAME_MAX + 1];
1208 char linkname[FILENAME_MAX + 1];
1210 // make the array negative
1211 for(c = 0 ; c < p->fds_size ; c++)
1212 p->fds[c] = -p->fds[c];
1214 while((de = readdir(fds))) {
1215 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
1218 // check if the fds array is small
1219 int fdid = atoi(de->d_name);
1220 if(fdid < 0) continue;
1221 if(fdid >= p->fds_size) {
1222 // it is small, extend it
1224 fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
1226 p->fds = reallocz(p->fds, (fdid + 100) * sizeof(int));
1228 fatal("Cannot re-allocate fds for %s", p->comm);
1232 // and initialize it
1233 for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
1234 p->fds_size = fdid + 100;
1237 if(p->fds[fdid] == 0) {
1238 // we don't know this fd, get it
1240 sprintf(fdname, "%s/proc/%d/fd/%s", host_prefix, p->pid, de->d_name);
1241 ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
1243 if(debug || (p->target && p->target->debug)) {
1244 if(debug || (p->target && p->target->debug))
1245 error("Cannot read link %s", fdname);
1252 // if another process already has this, we will get
1254 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1257 // else make it positive again, we need it
1258 // of course, the actual file may have changed, but we don't care so much
1259 // FIXME: we could compare the inode as returned by readdir dirent structure
1260 else p->fds[fdid] = -p->fds[fdid];
1264 // remove all the negative file descriptors
1265 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1266 file_descriptor_not_used(-p->fds[c]);
1275 // ----------------------------------------------------------------------------
1277 int print_process_and_parents(struct pid_stat *p, unsigned long long time) {
1278 char *prefix = "\\_ ";
1282 indent = print_process_and_parents(p->parent, p->stat_collected_usec);
1286 char buffer[indent + 1];
1289 for(i = 0; i < indent ;i++) buffer[i] = ' ';
1292 fprintf(stderr, " %s %s%s (%d %s %lld"
1297 , p->updated?"running":"exited"
1298 , (long long)p->stat_collected_usec - (long long)time
1301 if(p->utime) fprintf(stderr, " utime=%llu", p->utime);
1302 if(p->stime) fprintf(stderr, " stime=%llu", p->stime);
1303 if(p->gtime) fprintf(stderr, " gtime=%llu", p->gtime);
1304 if(p->cutime) fprintf(stderr, " cutime=%llu", p->cutime);
1305 if(p->cstime) fprintf(stderr, " cstime=%llu", p->cstime);
1306 if(p->cgtime) fprintf(stderr, " cgtime=%llu", p->cgtime);
1307 if(p->minflt) fprintf(stderr, " minflt=%llu", p->minflt);
1308 if(p->cminflt) fprintf(stderr, " cminflt=%llu", p->cminflt);
1309 if(p->majflt) fprintf(stderr, " majflt=%llu", p->majflt);
1310 if(p->cmajflt) fprintf(stderr, " cmajflt=%llu", p->cmajflt);
1311 fprintf(stderr, ")\n");
1316 void print_process_tree(struct pid_stat *p, char *msg) {
1318 fprintf(stderr, "%s: process %s (%d, %s) with parents:\n", msg, p->comm, p->pid, p->updated?"running":"exited");
1319 print_process_and_parents(p, p->stat_collected_usec);
1322 void find_lost_child_debug(struct pid_stat *pe, unsigned long long lost, int type) {
1324 struct pid_stat *p = NULL;
1326 for(p = root_of_pids; p ; p = p->next) {
1327 if(p == pe) continue;
1331 if(p->cminflt > lost) {
1332 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);
1338 if(p->cmajflt > lost) {
1339 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);
1345 if(p->cutime > lost) {
1346 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);
1352 if(p->cstime > lost) {
1353 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);
1359 if(p->cgtime > lost) {
1360 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);
1370 fprintf(stderr, " > cannot find any process to use the lost exited child minflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1374 fprintf(stderr, " > cannot find any process to use the lost exited child majflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1378 fprintf(stderr, " > cannot find any process to use the lost exited child utime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1382 fprintf(stderr, " > cannot find any process to use the lost exited child stime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1386 fprintf(stderr, " > cannot find any process to use the lost exited child gtime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1392 unsigned long long remove_exited_child_from_parent(unsigned long long *field, unsigned long long *pfield) {
1393 unsigned long long absorbed = 0;
1395 if(*field > *pfield) {
1396 absorbed += *pfield;
1409 void process_exited_processes() {
1412 for(p = root_of_pids; p ; p = p->next) {
1413 if(p->updated || !p->stat_collected_usec)
1416 struct pid_stat *pp = p->parent;
1418 unsigned long long utime = (p->utime_raw + p->cutime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1419 unsigned long long stime = (p->stime_raw + p->cstime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1420 unsigned long long gtime = (p->gtime_raw + p->cgtime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1421 unsigned long long minflt = (p->minflt_raw + p->cminflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1422 unsigned long long majflt = (p->majflt_raw + p->cmajflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1424 if(utime + stime + gtime + minflt + majflt == 0)
1427 if(unlikely(debug)) {
1429 fprintf(stderr, "Absorb %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1432 , p->updated?"running":"exited"
1439 print_process_tree(p, "Searching parents");
1442 for(pp = p->parent; pp ; pp = pp->parent) {
1443 if(!pp->updated) continue;
1445 unsigned long long absorbed;
1446 absorbed = remove_exited_child_from_parent(&utime, &pp->cutime);
1447 if(unlikely(debug && absorbed))
1448 fprintf(stderr, " > process %s (%d %s) absorbed %llu utime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, utime);
1450 absorbed = remove_exited_child_from_parent(&stime, &pp->cstime);
1451 if(unlikely(debug && absorbed))
1452 fprintf(stderr, " > process %s (%d %s) absorbed %llu stime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, stime);
1454 absorbed = remove_exited_child_from_parent(>ime, &pp->cgtime);
1455 if(unlikely(debug && absorbed))
1456 fprintf(stderr, " > process %s (%d %s) absorbed %llu gtime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, gtime);
1458 absorbed = remove_exited_child_from_parent(&minflt, &pp->cminflt);
1459 if(unlikely(debug && absorbed))
1460 fprintf(stderr, " > process %s (%d %s) absorbed %llu minflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, minflt);
1462 absorbed = remove_exited_child_from_parent(&majflt, &pp->cmajflt);
1463 if(unlikely(debug && absorbed))
1464 fprintf(stderr, " > process %s (%d %s) absorbed %llu majflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, majflt);
1467 if(unlikely(utime + stime + gtime + minflt + majflt > 0)) {
1468 if(unlikely(debug)) {
1469 if(utime) find_lost_child_debug(p, utime, 3);
1470 if(stime) find_lost_child_debug(p, stime, 4);
1471 if(gtime) find_lost_child_debug(p, gtime, 5);
1472 if(minflt) find_lost_child_debug(p, minflt, 1);
1473 if(majflt) find_lost_child_debug(p, majflt, 2);
1479 fprintf(stderr, " > remaining resources - KEEP - for another loop: %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1482 , p->updated?"running":"exited"
1490 for(pp = p->parent; pp ; pp = pp->parent) {
1491 if(pp->updated) break;
1495 fprintf(stderr, " > - KEEP - parent for another loop: %s (%d %s)\n"
1498 , pp->updated?"running":"exited"
1502 p->utime_raw = utime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1503 p->stime_raw = stime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1504 p->gtime_raw = gtime * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1505 p->minflt_raw = minflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1506 p->majflt_raw = majflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (1000000ULL * RATES_DETAIL);
1507 p->cutime_raw = p->cstime_raw = p->cgtime_raw = p->cminflt_raw = p->cmajflt_raw = 0;
1510 fprintf(stderr, "\n");
1512 else if(unlikely(debug)) {
1513 fprintf(stderr, " > totally absorbed - DONE - %s (%d %s)\n"
1516 , p->updated?"running":"exited"
1522 void link_all_processes_to_their_parents(void) {
1523 struct pid_stat *p, *pp;
1525 // link all children to their parents
1526 // and update children count on parents
1527 for(p = root_of_pids; p ; p = p->next) {
1528 // for each process found
1533 if(unlikely(!p->ppid)) {
1538 pp = all_pids[p->ppid];
1541 pp->children_count++;
1543 if(unlikely(debug || (p->target && p->target->debug)))
1544 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);
1548 error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1553 // ----------------------------------------------------------------------------
1555 // 1. read all files in /proc
1556 // 2. for each numeric directory:
1557 // i. read /proc/pid/stat
1558 // ii. read /proc/pid/statm
1559 // iii. read /proc/pid/io (requires root access)
1560 // iii. read the entries in directory /proc/pid/fd (requires root access)
1562 // a. find or create a struct file_descriptor
1563 // b. cleanup any old/unused file_descriptors
1565 // after all these, some pids may be linked to targets, while others may not
1567 // in case of errors, only 1 every 1000 errors is printed
1568 // to avoid filling up all disk space
1569 // if debug is enabled, all errors are printed
1571 static int compar_pid(const void *pid1, const void *pid2) {
1573 struct pid_stat *p1 = all_pids[*((pid_t *)pid1)];
1574 struct pid_stat *p2 = all_pids[*((pid_t *)pid2)];
1576 if(p1->sortlist > p2->sortlist)
1582 static inline int managed_log(struct pid_stat *p, uint32_t log, int status) {
1583 if(unlikely(!status)) {
1584 // error("command failed log %u, errno %d", log, errno);
1586 if(unlikely(debug || errno != ENOENT)) {
1587 if(unlikely(debug || !(p->log_thrown & log))) {
1588 p->log_thrown |= log;
1591 error("Cannot process %s/proc/%d/io (command '%s')", host_prefix, p->pid, p->comm);
1595 error("Cannot process %s/proc/%d/statm (command '%s')", host_prefix, p->pid, p->comm);
1598 case PID_LOG_CMDLINE:
1599 error("Cannot process %s/proc/%d/cmdline (command '%s')", host_prefix, p->pid, p->comm);
1603 error("Cannot process entries in %s/proc/%d/fd (command '%s')", host_prefix, p->pid, p->comm);
1610 error("unhandled error for pid %d, command '%s'", p->pid, p->comm);
1617 else if(unlikely(p->log_thrown & log)) {
1618 // error("unsetting log %u on pid %d", log, p->pid);
1619 p->log_thrown &= ~log;
1625 void collect_data_for_pid(pid_t pid) {
1626 if(unlikely(pid <= 0 || pid > pid_max)) {
1627 error("Invalid pid %d read (expected 1 to %d). Ignoring process.", pid, pid_max);
1631 struct pid_stat *p = get_pid_entry(pid);
1632 if(unlikely(!p || p->read)) return;
1635 // fprintf(stderr, "Reading process %d (%s), sortlist %d\n", p->pid, p->comm, p->sortlist);
1637 // --------------------------------------------------------------------
1640 if(unlikely(!managed_log(p, PID_LOG_STAT, read_proc_pid_stat(p))))
1641 // there is no reason to proceed if we cannot get its status
1644 read_proc_pid_ownership(p);
1646 // check its parent pid
1647 if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
1648 error("Pid %d (command '%s') states invalid parent pid %d. Using 0.", pid, p->comm, p->ppid);
1652 // --------------------------------------------------------------------
1655 managed_log(p, PID_LOG_IO, read_proc_pid_io(p));
1657 // --------------------------------------------------------------------
1658 // /proc/<pid>/statm
1660 if(unlikely(!managed_log(p, PID_LOG_STATM, read_proc_pid_statm(p))))
1661 // there is no reason to proceed if we cannot get its memory status
1664 // --------------------------------------------------------------------
1667 // check if it is target
1668 // we do this only once, the first time this pid is loaded
1669 if(unlikely(p->new_entry)) {
1670 // /proc/<pid>/cmdline
1671 if(likely(proc_pid_cmdline_is_needed))
1672 managed_log(p, PID_LOG_CMDLINE, read_proc_pid_cmdline(p));
1675 fprintf(stderr, "apps.plugin: \tJust added %d (%s)\n", pid, p->comm);
1677 uint32_t hash = simple_hash(p->comm);
1678 size_t pclen = strlen(p->comm);
1681 for(w = apps_groups_root_target; w ; w = w->next) {
1682 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1684 // find it - 4 cases:
1685 // 1. the target is not a pattern
1686 // 2. the target has the prefix
1687 // 3. the target has the suffix
1688 // 4. the target is something inside cmdline
1689 if( (!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm))
1690 || (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen))
1691 || (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen]))
1692 || (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && strstr(p->cmdline, w->compare))
1694 if(w->target) p->target = w->target;
1697 if(debug || (p->target && p->target->debug))
1698 fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1705 // --------------------------------------------------------------------
1708 managed_log(p, PID_LOG_FDS, read_pid_file_descriptors(p));
1710 // --------------------------------------------------------------------
1713 if(unlikely(debug && include_exited_childs && all_pids_count && p->ppid && all_pids[p->ppid] && !all_pids[p->ppid]->read))
1714 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);
1716 // mark it as updated
1722 int collect_data_for_all_processes_from_proc(void) {
1723 struct pid_stat *p = NULL;
1725 if(all_pids_count) {
1726 // read parents before childs
1727 // this is needed to prevent a situation where
1728 // a child is found running, but until we read
1729 // its parent, it has exited and its parent
1730 // has accumulated its resources
1733 for(p = root_of_pids; p ; p = p->next) {
1738 p->children_count = 0;
1741 all_pids_sortlist[slc++] = p->pid;
1744 if(unlikely(slc != all_pids_count)) {
1745 error("Internal error: I was thinking I had %ld processes in my arrays, but it seems there are more.", all_pids_count);
1746 all_pids_count = slc;
1749 if(include_exited_childs) {
1750 qsort((void *)all_pids_sortlist, all_pids_count, sizeof(pid_t), compar_pid);
1751 for(slc = 0; slc < all_pids_count; slc++)
1752 collect_data_for_pid(all_pids_sortlist[slc]);
1756 char dirname[FILENAME_MAX + 1];
1758 snprintfz(dirname, FILENAME_MAX, "%s/proc", host_prefix);
1759 DIR *dir = opendir(dirname);
1762 struct dirent *file = NULL;
1764 while((file = readdir(dir))) {
1765 char *endptr = file->d_name;
1766 pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
1768 // make sure we read a valid number
1769 if(unlikely(endptr == file->d_name || *endptr != '\0'))
1772 collect_data_for_pid(pid);
1776 // normally this is done
1777 // however we may have processes exited while we collected values
1778 // so let's find the exited ones
1779 // we do this by collecting the ownership of process
1780 // if we manage to get the ownership, the process still runs
1783 link_all_processes_to_their_parents();
1784 process_exited_processes();
1789 // ----------------------------------------------------------------------------
1790 // update statistics on the targets
1792 // 1. link all childs to their parents
1793 // 2. go from bottom to top, marking as merged all childs to their parents
1794 // this step links all parents without a target to the child target, if any
1795 // 3. link all top level processes (the ones not merged) to the default target
1796 // 4. go from top to bottom, linking all childs without a target, to their parent target
1797 // after this step, all processes have a target
1798 // [5. for each killed pid (updated = 0), remove its usage from its target]
1799 // 6. zero all apps_groups_targets
1800 // 7. concentrate all values on the apps_groups_targets
1801 // 8. remove all killed processes
1802 // 9. find the unique file count for each target
1803 // check: update_apps_groups_statistics()
1805 void cleanup_exited_pids(void) {
1807 struct pid_stat *p = NULL;
1809 for(p = root_of_pids; p ;) {
1810 if(!p->updated && (!p->keep || p->keeploops > 0)) {
1811 // 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);
1813 if(unlikely(debug && (p->keep || p->keeploops)))
1814 fprintf(stderr, " > CLEANUP cannot keep exited process %d (%s) anymore - removing it.\n", p->pid, p->comm);
1816 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1817 file_descriptor_not_used(p->fds[c]);
1826 if(unlikely(p->keep)) p->keeploops++;
1833 void apply_apps_groups_targets_inheritance(void) {
1834 struct pid_stat *p = NULL;
1836 // children that do not have a target
1837 // inherit their target from their parent
1838 int found = 1, loops = 0;
1840 if(unlikely(debug)) loops++;
1842 for(p = root_of_pids; p ; p = p->next) {
1843 // if this process does not have a target
1844 // and it has a parent
1845 // and its parent has a target
1846 // then, set the parent's target to this process
1847 if(unlikely(!p->target && p->parent && p->parent->target)) {
1848 p->target = p->parent->target;
1851 if(debug || (p->target && p->target->debug))
1852 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);
1857 // find all the procs with 0 childs and merge them to their parents
1858 // repeat, until nothing more can be done.
1862 if(unlikely(debug)) loops++;
1865 for(p = root_of_pids; p ; p = p->next) {
1866 if(unlikely(!p->sortlist && !p->children_count))
1867 p->sortlist = sortlist++;
1869 // if this process does not have any children
1870 // and is not already merged
1872 // and its parent has children
1873 // and the target of this process and its parent is the same, or the parent does not have a target
1874 // and its parent is not init
1875 // then, mark them as merged.
1880 && p->parent->children_count
1881 && (p->target == p->parent->target || !p->parent->target)
1884 p->parent->children_count--;
1887 // the parent inherits the child's target, if it does not have a target itself
1888 if(unlikely(p->target && !p->parent->target)) {
1889 p->parent->target = p->target;
1891 if(debug || (p->target && p->target->debug))
1892 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);
1900 fprintf(stderr, "apps.plugin: TARGET INHERITANCE: merged %d processes\n", found);
1903 // init goes always to default target
1905 all_pids[1]->target = apps_groups_default_target;
1907 // give a default target on all top level processes
1908 if(unlikely(debug)) loops++;
1909 for(p = root_of_pids; p ; p = p->next) {
1910 // if the process is not merged itself
1911 // then is is a top level process
1912 if(unlikely(!p->merged && !p->target))
1913 p->target = apps_groups_default_target;
1915 // make sure all processes have a sortlist
1916 if(unlikely(!p->sortlist))
1917 p->sortlist = sortlist++;
1921 all_pids[1]->sortlist = sortlist++;
1923 // give a target to all merged child processes
1926 if(unlikely(debug)) loops++;
1928 for(p = root_of_pids; p ; p = p->next) {
1929 if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
1930 p->target = p->parent->target;
1933 if(debug || (p->target && p->target->debug))
1934 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);
1940 fprintf(stderr, "apps.plugin: apply_apps_groups_targets_inheritance() made %d loops on the process tree\n", loops);
1943 long zero_all_targets(struct target *root) {
1947 for (w = root; w ; w = w->next) {
1950 if(w->fds) freez(w->fds);
1968 w->statm_resident = 0;
1975 w->io_logical_bytes_read = 0;
1976 w->io_logical_bytes_written = 0;
1977 w->io_read_calls = 0;
1978 w->io_write_calls = 0;
1979 w->io_storage_bytes_read = 0;
1980 w->io_storage_bytes_written = 0;
1981 w->io_cancelled_write_bytes = 0;
1987 void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) {
1990 if(unlikely(!w->fds))
1991 w->fds = callocz(sizeof(int), (size_t) all_files_size);
1993 if(likely(p->updated)) {
1994 w->cutime += p->cutime;
1995 w->cstime += p->cstime;
1996 w->cgtime += p->cgtime;
1997 w->cminflt += p->cminflt;
1998 w->cmajflt += p->cmajflt;
2000 w->utime += p->utime;
2001 w->stime += p->stime;
2002 w->gtime += p->gtime;
2003 w->minflt += p->minflt;
2004 w->majflt += p->majflt;
2008 w->statm_size += p->statm_size;
2009 w->statm_resident += p->statm_resident;
2010 w->statm_share += p->statm_share;
2011 w->statm_text += p->statm_text;
2012 w->statm_lib += p->statm_lib;
2013 w->statm_data += p->statm_data;
2014 w->statm_dirty += p->statm_dirty;
2016 w->io_logical_bytes_read += p->io_logical_bytes_read;
2017 w->io_logical_bytes_written += p->io_logical_bytes_written;
2018 w->io_read_calls += p->io_read_calls;
2019 w->io_write_calls += p->io_write_calls;
2020 w->io_storage_bytes_read += p->io_storage_bytes_read;
2021 w->io_storage_bytes_written += p->io_storage_bytes_written;
2022 w->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
2025 w->num_threads += p->num_threads;
2027 if(likely(w->fds)) {
2029 for(c = 0; c < p->fds_size ;c++) {
2030 if(p->fds[c] == 0) continue;
2032 if(likely(p->fds[c] < all_files_size)) {
2033 if(w->fds) w->fds[p->fds[c]]++;
2036 error("Invalid fd number %d", p->fds[c]);
2040 if(unlikely(debug || w->debug))
2041 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);
2045 void count_targets_fds(struct target *root) {
2049 for (w = root; w ; w = w->next) {
2050 if(!w->fds) continue;
2055 w->openinotifies = 0;
2056 w->openeventfds = 0;
2057 w->opentimerfds = 0;
2058 w->opensignalfds = 0;
2059 w->openeventpolls = 0;
2062 for(c = 1; c < all_files_size ;c++) {
2064 switch(all_files[c].type) {
2073 case FILETYPE_SOCKET:
2077 case FILETYPE_INOTIFY:
2081 case FILETYPE_EVENTFD:
2085 case FILETYPE_TIMERFD:
2089 case FILETYPE_SIGNALFD:
2093 case FILETYPE_EVENTPOLL:
2094 w->openeventpolls++;
2107 void calculate_netdata_statistics(void) {
2108 apply_apps_groups_targets_inheritance();
2110 zero_all_targets(users_root_target);
2111 zero_all_targets(groups_root_target);
2112 apps_groups_targets = zero_all_targets(apps_groups_root_target);
2114 // this has to be done, before the cleanup
2115 struct pid_stat *p = NULL;
2116 struct target *w = NULL, *o = NULL;
2118 // concentrate everything on the apps_groups_targets
2119 for(p = root_of_pids; p ; p = p->next) {
2121 // --------------------------------------------------------------------
2122 // apps_groups targets
2123 if(likely(p->target))
2124 aggregate_pid_on_target(p->target, p, NULL);
2126 error("pid %d %s was left without a target!", p->pid, p->comm);
2129 // --------------------------------------------------------------------
2132 if(likely(p->user_target && p->user_target->uid == p->uid))
2135 if(unlikely(debug && p->user_target))
2136 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);
2138 w = p->user_target = get_users_target(p->uid);
2142 aggregate_pid_on_target(w, p, o);
2144 error("pid %d %s was left without a user target!", p->pid, p->comm);
2147 // --------------------------------------------------------------------
2149 o = p->group_target;
2150 if(likely(p->group_target && p->group_target->gid == p->gid))
2151 w = p->group_target;
2153 if(unlikely(debug && p->group_target))
2154 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);
2156 w = p->group_target = get_groups_target(p->gid);
2160 aggregate_pid_on_target(w, p, o);
2162 error("pid %d %s was left without a group target!", p->pid, p->comm);
2166 count_targets_fds(apps_groups_root_target);
2167 count_targets_fds(users_root_target);
2168 count_targets_fds(groups_root_target);
2170 cleanup_exited_pids();
2173 // ----------------------------------------------------------------------------
2174 // update chart dimensions
2176 BUFFER *output = NULL;
2177 int print_calculated_number(char *str, calculated_number value) { (void)str; (void)value; return 0; }
2179 static inline void send_BEGIN(const char *type, const char *id, unsigned long long usec) {
2180 // fprintf(stdout, "BEGIN %s.%s %llu\n", type, id, usec);
2181 buffer_strcat(output, "BEGIN ");
2182 buffer_strcat(output, type);
2183 buffer_strcat(output, ".");
2184 buffer_strcat(output, id);
2185 buffer_strcat(output, " ");
2186 buffer_print_llu(output, usec);
2187 buffer_strcat(output, "\n");
2190 static inline void send_SET(const char *name, unsigned long long value) {
2191 // fprintf(stdout, "SET %s = %llu\n", name, value);
2192 buffer_strcat(output, "SET ");
2193 buffer_strcat(output, name);
2194 buffer_strcat(output, " = ");
2195 buffer_print_llu(output, value);
2196 buffer_strcat(output, "\n");
2199 static inline void send_END(void) {
2200 // fprintf(stdout, "END\n");
2201 buffer_strcat(output, "END\n");
2204 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;
2205 double minflt_fix_ratio = 1.0, majflt_fix_ratio = 1.0, cminflt_fix_ratio = 1.0, cmajflt_fix_ratio = 1.0;
2207 unsigned long long send_resource_usage_to_netdata() {
2208 static struct timeval last = { 0, 0 };
2209 static struct rusage me_last;
2214 unsigned long long usec;
2215 unsigned long long cpuuser;
2216 unsigned long long cpusyst;
2219 gettimeofday(&last, NULL);
2220 getrusage(RUSAGE_SELF, &me_last);
2222 // the first time, give a zero to allow
2223 // netdata calibrate to the current time
2224 // usec = update_every * 1000000ULL;
2230 gettimeofday(&now, NULL);
2231 getrusage(RUSAGE_SELF, &me);
2233 usec = usec_dt(&now, &last);
2234 cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
2235 cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
2237 bcopy(&now, &last, sizeof(struct timeval));
2238 bcopy(&me, &me_last, sizeof(struct rusage));
2241 buffer_sprintf(output,
2242 "BEGIN netdata.apps_cpu %llu\n"
2244 "SET system = %llu\n"
2246 "BEGIN netdata.apps_files %llu\n"
2247 "SET files = %llu\n"
2250 "SET targets = %ld\n"
2252 "BEGIN netdata.apps_fix %llu\n"
2253 "SET utime = %llu\n"
2254 "SET stime = %llu\n"
2255 "SET gtime = %llu\n"
2256 "SET minflt = %llu\n"
2257 "SET majflt = %llu\n"
2266 , apps_groups_targets
2268 , (unsigned long long)(utime_fix_ratio * 100 * RATES_DETAIL)
2269 , (unsigned long long)(stime_fix_ratio * 100 * RATES_DETAIL)
2270 , (unsigned long long)(gtime_fix_ratio * 100 * RATES_DETAIL)
2271 , (unsigned long long)(minflt_fix_ratio * 100 * RATES_DETAIL)
2272 , (unsigned long long)(majflt_fix_ratio * 100 * RATES_DETAIL)
2275 if(include_exited_childs)
2276 buffer_sprintf(output,
2277 "BEGIN netdata.apps_children_fix %llu\n"
2278 "SET cutime = %llu\n"
2279 "SET cstime = %llu\n"
2280 "SET cgtime = %llu\n"
2281 "SET cminflt = %llu\n"
2282 "SET cmajflt = %llu\n"
2285 , (unsigned long long)(cutime_fix_ratio * 100 * RATES_DETAIL)
2286 , (unsigned long long)(cstime_fix_ratio * 100 * RATES_DETAIL)
2287 , (unsigned long long)(cgtime_fix_ratio * 100 * RATES_DETAIL)
2288 , (unsigned long long)(cminflt_fix_ratio * 100 * RATES_DETAIL)
2289 , (unsigned long long)(cmajflt_fix_ratio * 100 * RATES_DETAIL)
2295 void normalize_data(struct target *root) {
2298 // childs processing introduces spikes
2299 // here we try to eliminate them by disabling childs processing either for specific dimensions
2300 // or entirely. Of course, either way, we disable it just a single iteration.
2302 unsigned long long max = processors * hz * RATES_DETAIL;
2303 unsigned long long utime = 0, cutime = 0, stime = 0, cstime = 0, gtime = 0, cgtime = 0, minflt = 0, cminflt = 0, majflt = 0, cmajflt = 0;
2305 if(global_utime > max) global_utime = max;
2306 if(global_stime > max) global_stime = max;
2307 if(global_gtime > max) global_gtime = max;
2309 for(w = root; w ; w = w->next) {
2310 if(w->target || (!w->processes && !w->exposed)) continue;
2315 cutime += w->cutime;
2316 cstime += w->cstime;
2317 cgtime += w->cgtime;
2319 minflt += w->minflt;
2320 majflt += w->majflt;
2321 cminflt += w->cminflt;
2322 cmajflt += w->cmajflt;
2325 if((global_utime || global_stime || global_gtime) && (utime || stime || gtime)) {
2326 if(global_utime + global_stime + global_gtime > utime + cutime + stime + cstime + gtime + cgtime) {
2327 // everything we collected fits
2333 cgtime_fix_ratio = 1.0; //(double)(global_utime + global_stime) / (double)(utime + cutime + stime + cstime);
2335 else if(global_utime + global_stime > utime + stime) {
2336 // childrens resources are too high
2337 // lower only the children resources
2340 gtime_fix_ratio = 1.0;
2343 cgtime_fix_ratio = (double)((global_utime + global_stime) - (utime + stime)) / (double)(cutime + cstime);
2346 // even running processes are unrealistic
2347 // zero the children resources
2348 // lower the running processes resources
2351 gtime_fix_ratio = (double)(global_utime + global_stime) / (double)(utime + stime);
2354 cgtime_fix_ratio = 0.0;
2363 cgtime_fix_ratio = 0.0;
2366 if(utime_fix_ratio > 1.0) utime_fix_ratio = 1.0;
2367 if(cutime_fix_ratio > 1.0) cutime_fix_ratio = 1.0;
2368 if(stime_fix_ratio > 1.0) stime_fix_ratio = 1.0;
2369 if(cstime_fix_ratio > 1.0) cstime_fix_ratio = 1.0;
2370 if(gtime_fix_ratio > 1.0) gtime_fix_ratio = 1.0;
2371 if(cgtime_fix_ratio > 1.0) cgtime_fix_ratio = 1.0;
2373 // if(utime_fix_ratio < 0.0) utime_fix_ratio = 0.0;
2374 // if(cutime_fix_ratio < 0.0) cutime_fix_ratio = 0.0;
2375 // if(stime_fix_ratio < 0.0) stime_fix_ratio = 0.0;
2376 // if(cstime_fix_ratio < 0.0) cstime_fix_ratio = 0.0;
2377 // if(gtime_fix_ratio < 0.0) gtime_fix_ratio = 0.0;
2378 // if(cgtime_fix_ratio < 0.0) cgtime_fix_ratio = 0.0;
2381 // we use cpu time to normalize page faults
2382 // the problem is that to find the proper max values
2383 // for page faults we have to parse /proc/vmstat
2384 // which is quite big to do it again (netdata does it already)
2386 // a better solution could be to somehow have netdata
2387 // do this normalization for us
2389 if(utime || stime || gtime)
2391 minflt_fix_ratio = (double)(utime * utime_fix_ratio + stime * stime_fix_ratio + gtime * gtime_fix_ratio) / (double)(utime + stime + gtime);
2394 majflt_fix_ratio = 1.0;
2396 if(cutime || cstime || cgtime)
2398 cminflt_fix_ratio = (double)(cutime * cutime_fix_ratio + cstime * cstime_fix_ratio + cgtime * cgtime_fix_ratio) / (double)(cutime + cstime + cgtime);
2401 cmajflt_fix_ratio = 1.0;
2405 if(unlikely(debug)) {
2407 "SYSTEM: u=%llu s=%llu g=%llu "
2408 "COLLECTED: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2409 "DELTA: u=%lld s=%lld g=%lld "
2410 "FIX: u=%0.2f s=%0.2f g=%0.2f cu=%0.2f cs=%0.2f cg=%0.2f "
2411 "FINALLY: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2422 , (long long)utime + (long long)cutime - (long long)global_utime
2423 , (long long)stime + (long long)cstime - (long long)global_stime
2424 , (long long)gtime + (long long)cgtime - (long long)global_gtime
2431 , (unsigned long long)(utime * utime_fix_ratio)
2432 , (unsigned long long)(stime * stime_fix_ratio)
2433 , (unsigned long long)(gtime * gtime_fix_ratio)
2434 , (unsigned long long)(cutime * cutime_fix_ratio)
2435 , (unsigned long long)(cstime * cstime_fix_ratio)
2436 , (unsigned long long)(cgtime * cgtime_fix_ratio)
2441 void send_collected_data_to_netdata(struct target *root, const char *type, unsigned long long usec) {
2444 send_BEGIN(type, "cpu", usec);
2445 for (w = root; w ; w = w->next) {
2446 if(unlikely(w->exposed))
2447 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));
2451 send_BEGIN(type, "cpu_user", usec);
2452 for (w = root; w ; w = w->next) {
2453 if(unlikely(w->exposed))
2454 send_SET(w->name, (unsigned long long)(w->utime * utime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cutime * cutime_fix_ratio)):0ULL));
2458 send_BEGIN(type, "cpu_system", usec);
2459 for (w = root; w ; w = w->next) {
2460 if(unlikely(w->exposed))
2461 send_SET(w->name, (unsigned long long)(w->stime * stime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cstime * cstime_fix_ratio)):0ULL));
2465 if(show_guest_time) {
2466 send_BEGIN(type, "cpu_guest", usec);
2467 for (w = root; w ; w = w->next) {
2468 if(unlikely(w->exposed))
2469 send_SET(w->name, (unsigned long long)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cgtime * cgtime_fix_ratio)):0ULL));
2474 send_BEGIN(type, "threads", usec);
2475 for (w = root; w ; w = w->next) {
2476 if(unlikely(w->exposed))
2477 send_SET(w->name, w->num_threads);
2481 send_BEGIN(type, "processes", usec);
2482 for (w = root; w ; w = w->next) {
2483 if(unlikely(w->exposed))
2484 send_SET(w->name, w->processes);
2488 send_BEGIN(type, "mem", usec);
2489 for (w = root; w ; w = w->next) {
2490 if(unlikely(w->exposed))
2491 send_SET(w->name, (w->statm_resident > w->statm_share)?(w->statm_resident - w->statm_share):0ULL);
2495 send_BEGIN(type, "minor_faults", usec);
2496 for (w = root; w ; w = w->next) {
2497 if(unlikely(w->exposed))
2498 send_SET(w->name, (unsigned long long)(w->minflt * minflt_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cminflt * cminflt_fix_ratio)):0ULL));
2502 send_BEGIN(type, "major_faults", usec);
2503 for (w = root; w ; w = w->next) {
2504 if(unlikely(w->exposed))
2505 send_SET(w->name, (unsigned long long)(w->majflt * majflt_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cmajflt * cmajflt_fix_ratio)):0ULL));
2509 send_BEGIN(type, "lreads", usec);
2510 for (w = root; w ; w = w->next) {
2511 if(unlikely(w->exposed))
2512 send_SET(w->name, w->io_logical_bytes_read);
2516 send_BEGIN(type, "lwrites", usec);
2517 for (w = root; w ; w = w->next) {
2518 if(unlikely(w->exposed))
2519 send_SET(w->name, w->io_logical_bytes_written);
2523 send_BEGIN(type, "preads", usec);
2524 for (w = root; w ; w = w->next) {
2525 if(unlikely(w->exposed))
2526 send_SET(w->name, w->io_storage_bytes_read);
2530 send_BEGIN(type, "pwrites", usec);
2531 for (w = root; w ; w = w->next) {
2532 if(unlikely(w->exposed))
2533 send_SET(w->name, w->io_storage_bytes_written);
2537 send_BEGIN(type, "files", usec);
2538 for (w = root; w ; w = w->next) {
2539 if(unlikely(w->exposed))
2540 send_SET(w->name, w->openfiles);
2544 send_BEGIN(type, "sockets", usec);
2545 for (w = root; w ; w = w->next) {
2546 if(unlikely(w->exposed))
2547 send_SET(w->name, w->opensockets);
2551 send_BEGIN(type, "pipes", usec);
2552 for (w = root; w ; w = w->next) {
2553 if(unlikely(w->exposed))
2554 send_SET(w->name, w->openpipes);
2560 // ----------------------------------------------------------------------------
2561 // generate the charts
2563 void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title)
2566 int newly_added = 0;
2568 for(w = root ; w ; w = w->next) {
2569 if (w->target) continue;
2571 if (!w->exposed && w->processes) {
2574 if (debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
2578 // nothing more to show
2579 if(!newly_added && show_guest_time == show_guest_time_old) return;
2581 // we have something new to show
2582 // update the charts
2583 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);
2584 for (w = root; w ; w = w->next) {
2585 if(unlikely(w->exposed))
2586 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu %s\n", w->name, hz * RATES_DETAIL / 100, w->hidden ? "hidden" : "");
2589 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);
2590 for (w = root; w ; w = w->next) {
2591 if(unlikely(w->exposed))
2592 buffer_sprintf(output, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
2595 buffer_sprintf(output, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20005 %d\n", type, title, type, update_every);
2596 for (w = root; w ; w = w->next) {
2597 if(unlikely(w->exposed))
2598 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2601 buffer_sprintf(output, "CHART %s.processes '' '%s Processes' 'processes' processes %s.processes stacked 20004 %d\n", type, title, type, update_every);
2602 for (w = root; w ; w = w->next) {
2603 if(unlikely(w->exposed))
2604 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2607 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);
2608 for (w = root; w ; w = w->next) {
2609 if(unlikely(w->exposed))
2610 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2613 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);
2614 for (w = root; w ; w = w->next) {
2615 if(unlikely(w->exposed))
2616 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2619 if(show_guest_time) {
2620 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);
2621 for (w = root; w; w = w->next) {
2622 if(unlikely(w->exposed))
2623 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2627 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);
2628 for (w = root; w ; w = w->next) {
2629 if(unlikely(w->exposed))
2630 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
2633 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);
2634 for (w = root; w ; w = w->next) {
2635 if(unlikely(w->exposed))
2636 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
2639 buffer_sprintf(output, "CHART %s.lreads '' '%s Disk Logical Reads' 'kilobytes/s' disk %s.lreads stacked 20042 %d\n", type, title, type, update_every);
2640 for (w = root; w ; w = w->next) {
2641 if(unlikely(w->exposed))
2642 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2645 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);
2646 for (w = root; w ; w = w->next) {
2647 if(unlikely(w->exposed))
2648 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2651 buffer_sprintf(output, "CHART %s.preads '' '%s Disk Reads' 'kilobytes/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every);
2652 for (w = root; w ; w = w->next) {
2653 if(unlikely(w->exposed))
2654 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2657 buffer_sprintf(output, "CHART %s.pwrites '' '%s Disk Writes' 'kilobytes/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
2658 for (w = root; w ; w = w->next) {
2659 if(unlikely(w->exposed))
2660 buffer_sprintf(output, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2663 buffer_sprintf(output, "CHART %s.files '' '%s Open Files' 'open files' disk %s.files stacked 20050 %d\n", type, 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", type, 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.pipes '' '%s Pipes' 'open pipes' processes %s.pipes stacked 20053 %d\n", type, title, type, update_every);
2676 for (w = root; w ; w = w->next) {
2677 if(unlikely(w->exposed))
2678 buffer_sprintf(output, "DIMENSION %s '' absolute 1 1\n", w->name);
2683 // ----------------------------------------------------------------------------
2684 // parse command line arguments
2686 void parse_args(int argc, char **argv)
2691 for(i = 1; i < argc; i++) {
2693 int n = atoi(argv[i]);
2700 if(strcmp("debug", argv[i]) == 0) {
2702 // debug_flags = 0xffffffff;
2706 if(strcmp("no-childs", argv[i]) == 0 || strcmp("without-childs", argv[i]) == 0) {
2707 include_exited_childs = 0;
2711 if(strcmp("with-childs", argv[i]) == 0) {
2712 include_exited_childs = 1;
2716 if(strcmp("with-guest", argv[i]) == 0) {
2717 enable_guest_charts = 1;
2721 if(strcmp("no-guest", argv[i]) == 0 || strcmp("without-guest", argv[i]) == 0) {
2722 enable_guest_charts = 0;
2726 if(strcmp("-h", argv[i]) == 0 || strcmp("--help", argv[i]) == 0) {
2729 "(C) 2016 Costa Tsaousis"
2731 "This program is a data collector plugin for netdata.\n"
2733 "Valid command line options:\n"
2735 "SECONDS set the data collection frequency\n"
2737 "debug enable debugging (lot of output)\n"
2740 "without-childs enable / disable aggregating exited\n"
2741 " children resources into parents\n"
2742 " (default is enabled)\n"
2745 "without-guest enable / disable reporting guest charts\n"
2746 " (default is disabled)\n"
2748 "NAME read apps_NAME.conf instead of\n"
2749 " apps_groups.conf\n"
2750 " (default NAME=groups)\n"
2760 error("Cannot understand option %s", argv[i]);
2764 if(freq > 0) update_every = freq;
2765 if(!name) name = "groups";
2767 if(read_apps_groups_conf(name)) {
2768 error("Cannot read process groups %s", name);
2773 int main(int argc, char **argv)
2775 // debug_flags = D_PROCFILE;
2777 // set the name for logging
2778 program_name = "apps.plugin";
2780 info("started on pid %d", getpid());
2782 // disable syslog for apps.plugin
2783 error_log_syslog = 0;
2785 // set errors flood protection to 100 logs per hour
2786 error_log_errors_per_period = 100;
2787 error_log_throttle_period = 3600;
2789 host_prefix = getenv("NETDATA_HOST_PREFIX");
2790 if(host_prefix == NULL) {
2791 // info("NETDATA_HOST_PREFIX is not passed from netdata");
2794 // else info("Found NETDATA_HOST_PREFIX='%s'", host_prefix);
2796 config_dir = getenv("NETDATA_CONFIG_DIR");
2797 if(config_dir == NULL) {
2798 // info("NETDATA_CONFIG_DIR is not passed from netdata");
2799 config_dir = CONFIG_DIR;
2801 // else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
2803 #ifdef NETDATA_INTERNAL_CHECKS
2804 if(debug_flags != 0) {
2805 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
2806 if(setrlimit(RLIMIT_CORE, &rl) != 0)
2807 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
2808 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
2810 #endif /* NETDATA_INTERNAL_CHECKS */
2812 procfile_adaptive_initial_allocation = 1;
2814 time_t started_t = time(NULL);
2817 pid_max = get_system_pid_max();
2818 processors = get_system_cpus();
2820 parse_args(argc, argv);
2822 all_pids_sortlist = callocz(sizeof(pid_t), (size_t)pid_max);
2823 all_pids = callocz(sizeof(struct pid_stat *), (size_t) pid_max);
2825 output = buffer_create(1024);
2826 buffer_sprintf(output,
2827 "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %1$d\n"
2828 "DIMENSION user '' incremental 1 1000\n"
2829 "DIMENSION system '' incremental 1 1000\n"
2830 "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %1$d\n"
2831 "DIMENSION files '' incremental 1 1\n"
2832 "DIMENSION pids '' absolute 1 1\n"
2833 "DIMENSION fds '' absolute 1 1\n"
2834 "DIMENSION targets '' absolute 1 1\n"
2835 "CHART netdata.apps_fix '' 'Apps Plugin Normalization Ratios' 'percentage' apps.plugin netdata.apps_fix line 140002 %1$d\n"
2836 "DIMENSION utime '' absolute 1 %2$llu\n"
2837 "DIMENSION stime '' absolute 1 %2$llu\n"
2838 "DIMENSION gtime '' absolute 1 %2$llu\n"
2839 "DIMENSION minflt '' absolute 1 %2$llu\n"
2840 "DIMENSION majflt '' absolute 1 %2$llu\n"
2845 if(include_exited_childs)
2846 buffer_sprintf(output,
2847 "CHART netdata.apps_children_fix '' 'Apps Plugin Exited Children Normalization Ratios' 'percentage' apps.plugin netdata.apps_children_fix line 140003 %1$d\n"
2848 "DIMENSION cutime '' absolute 1 %2$llu\n"
2849 "DIMENSION cstime '' absolute 1 %2$llu\n"
2850 "DIMENSION cgtime '' absolute 1 %2$llu\n"
2851 "DIMENSION cminflt '' absolute 1 %2$llu\n"
2852 "DIMENSION cmajflt '' absolute 1 %2$llu\n"
2857 #ifndef PROFILING_MODE
2858 unsigned long long sunext = (time(NULL) - (time(NULL) % update_every) + update_every) * 1000000ULL;
2859 unsigned long long sunow;
2860 #endif /* PROFILING_MODE */
2862 global_iterations_counter = 1;
2863 for(;1; global_iterations_counter++) {
2864 #ifndef PROFILING_MODE
2865 // delay until it is our time to run
2866 while((sunow = time_usec()) < sunext)
2867 sleep_usec(sunext - sunow);
2869 // find the next time we need to run
2870 while(time_usec() > sunext)
2871 sunext += update_every * 1000000ULL;
2872 #endif /* PROFILING_MODE */
2874 if(!collect_data_for_all_processes_from_proc()) {
2875 error("Cannot collect /proc data for running processes. Disabling apps.plugin...");
2876 printf("DISABLE\n");
2880 calculate_netdata_statistics();
2881 normalize_data(apps_groups_root_target);
2883 unsigned long long dt = send_resource_usage_to_netdata();
2885 // this is smart enough to show only newly added apps, when needed
2886 send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps");
2887 send_charts_updates_to_netdata(users_root_target, "users", "Users");
2888 send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups");
2890 send_collected_data_to_netdata(apps_groups_root_target, "apps", dt);
2891 send_collected_data_to_netdata(users_root_target, "users", dt);
2892 send_collected_data_to_netdata(groups_root_target, "groups", dt);
2894 show_guest_time_old = show_guest_time;
2896 //if(puts(buffer_tostring(output)) == EOF)
2897 if(write(STDOUT_FILENO, buffer_tostring(output), buffer_strlen(output)) == -1)
2898 fatal("Cannot send chart values to netdata.");
2901 buffer_flush(output);
2904 fprintf(stderr, "apps.plugin: done Loop No %llu\n", global_iterations_counter);
2906 current_t = time(NULL);
2908 #ifndef PROFILING_MODE
2909 // restart check (14400 seconds)
2910 if(current_t - started_t > 14400) exit(0);
2912 if(current_t - started_t > 10) exit(0);
2913 #endif /* PROFILING_MODE */