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
16 #define MAX_SPARE_FDS 1
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 *config_dir = CONFIG_DIR;
27 pid_t *all_pids_sortlist = NULL;
29 // will be automatically set to 1, if guest values are collected
30 int show_guest_time = 0;
31 int show_guest_time_old = 0;
33 int enable_guest_charts = 0;
34 int enable_file_charts = 1;
35 int enable_users_charts = 1;
36 int enable_groups_charts = 1;
38 // ----------------------------------------------------------------------------
40 void netdata_cleanup_and_exit(int ret) {
45 // ----------------------------------------------------------------------------
47 // target is the structure that process data are aggregated
50 char compare[MAX_COMPARE_NAME + 1];
54 char id[MAX_NAME + 1];
57 char name[MAX_NAME + 1];
62 unsigned long long minflt;
63 unsigned long long cminflt;
64 unsigned long long majflt;
65 unsigned long long cmajflt;
66 unsigned long long utime;
67 unsigned long long stime;
68 unsigned long long gtime;
69 unsigned long long cutime;
70 unsigned long long cstime;
71 unsigned long long cgtime;
72 unsigned long long num_threads;
73 // unsigned long long rss;
75 unsigned long long statm_size;
76 unsigned long long statm_resident;
77 unsigned long long statm_share;
78 // unsigned long long statm_text;
79 // unsigned long long statm_lib;
80 // unsigned long long statm_data;
81 // unsigned long long statm_dirty;
83 unsigned long long io_logical_bytes_read;
84 unsigned long long io_logical_bytes_written;
85 // unsigned long long io_read_calls;
86 // unsigned long long io_write_calls;
87 unsigned long long io_storage_bytes_read;
88 unsigned long long io_storage_bytes_written;
89 // unsigned long long io_cancelled_write_bytes;
94 unsigned long long openfiles;
95 unsigned long long openpipes;
96 unsigned long long opensockets;
97 unsigned long long openinotifies;
98 unsigned long long openeventfds;
99 unsigned long long opentimerfds;
100 unsigned long long opensignalfds;
101 unsigned long long openeventpolls;
102 unsigned long long openother;
104 unsigned long processes; // how many processes have been merged to this
105 int exposed; // if set, we have sent this to netdata
106 int hidden; // if set, we set the hidden flag on the dimension
109 int starts_with; // if set, the compare string matches only the
110 // beginning of the command
112 struct target *target; // the one that will be reported to netdata
117 // ----------------------------------------------------------------------------
119 // aggregate all processes in groups, to have a limited number of dimensions
121 struct target *apps_groups_root_target = NULL;
122 struct target *apps_groups_default_target = NULL;
123 long apps_groups_targets = 0;
125 struct target *users_root_target = NULL;
126 struct target *groups_root_target = NULL;
128 static struct target *get_users_target(uid_t uid)
131 for(w = users_root_target ; w ; w = w->next)
132 if(w->uid == uid) return w;
134 w = callocz(sizeof(struct target), 1);
135 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", uid);
136 w->comparehash = simple_hash(w->compare);
137 w->comparelen = strlen(w->compare);
139 snprintfz(w->id, MAX_NAME, "%u", uid);
140 w->idhash = simple_hash(w->id);
142 struct passwd *pw = getpwuid(uid);
144 snprintfz(w->name, MAX_NAME, "%u", uid);
146 snprintfz(w->name, MAX_NAME, "%s", pw->pw_name);
148 netdata_fix_chart_name(w->name);
152 w->next = users_root_target;
153 users_root_target = w;
156 fprintf(stderr, "apps.plugin: added uid %u ('%s') target\n", w->uid, w->name);
161 struct target *get_groups_target(gid_t gid)
164 for(w = groups_root_target ; w ; w = w->next)
165 if(w->gid == gid) return w;
167 w = callocz(sizeof(struct target), 1);
168 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", gid);
169 w->comparehash = simple_hash(w->compare);
170 w->comparelen = strlen(w->compare);
172 snprintfz(w->id, MAX_NAME, "%u", gid);
173 w->idhash = simple_hash(w->id);
175 struct group *gr = getgrgid(gid);
177 snprintfz(w->name, MAX_NAME, "%u", gid);
179 snprintfz(w->name, MAX_NAME, "%s", gr->gr_name);
181 netdata_fix_chart_name(w->name);
185 w->next = groups_root_target;
186 groups_root_target = w;
189 fprintf(stderr, "apps.plugin: added gid %u ('%s') target\n", w->gid, w->name);
194 // find or create a new target
195 // there are targets that are just aggregated to other target (the second argument)
196 static struct target *get_apps_groups_target(const char *id, struct target *target, const char *name) {
197 int tdebug = 0, thidden = target?target->hidden:0, ends_with = 0;
198 const char *nid = id;
200 // extract the options
201 while(nid[0] == '-' || nid[0] == '+' || nid[0] == '*') {
202 if(nid[0] == '-') thidden = 1;
203 if(nid[0] == '+') tdebug = 1;
204 if(nid[0] == '*') ends_with = 1;
207 uint32_t hash = simple_hash(id);
209 // find if it already exists
210 struct target *w, *last = apps_groups_root_target;
211 for(w = apps_groups_root_target ; w ; w = w->next) {
212 if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
218 // find an existing target
219 if(unlikely(!target)) {
220 while(*name == '-') {
221 if(*name == '-') thidden = 1;
224 for(target = apps_groups_root_target ; target ; target = target->next) {
225 if(!target->target && strcmp(name, target->name) == 0)
228 if(unlikely(debug)) {
230 fprintf(stderr, "apps.plugin: REUSING TARGET NAME '%s' on ID '%s'\n", target->name, target->id);
232 fprintf(stderr, "apps.plugin: NEW TARGET NAME '%s' on ID '%s'\n", name, id);
236 if(target && target->target)
237 fatal("Internal Error: request to link process '%s' to target '%s' which is linked to target '%s'", id, target->id, target->target->id);
239 w = callocz(sizeof(struct target), 1);
240 strncpyz(w->id, nid, MAX_NAME);
241 w->idhash = simple_hash(w->id);
243 if(unlikely(!target))
245 strncpyz(w->name, name, MAX_NAME);
248 strncpyz(w->name, nid, MAX_NAME);
250 strncpyz(w->compare, nid, MAX_COMPARE_NAME);
251 size_t len = strlen(w->compare);
252 if(w->compare[len - 1] == '*') {
253 w->compare[len - 1] = '\0';
256 w->ends_with = ends_with;
258 if(w->starts_with && w->ends_with)
259 proc_pid_cmdline_is_needed = 1;
261 w->comparehash = simple_hash(w->compare);
262 w->comparelen = strlen(w->compare);
268 // append it, to maintain the order in apps_groups.conf
269 if(last) last->next = w;
270 else apps_groups_root_target = w;
273 fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
275 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
276 , w->target?w->target->name:w->name
277 , (w->hidden)?"hidden":"-"
278 , (w->debug)?"debug":"-"
284 // read the apps_groups.conf file
285 static int read_apps_groups_conf(const char *file)
287 char filename[FILENAME_MAX + 1];
289 snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, file);
292 fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
294 // ----------------------------------------
296 procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT);
299 procfile_set_quotes(ff, "'\"");
301 ff = procfile_readall(ff);
305 unsigned long line, lines = procfile_lines(ff);
307 for(line = 0; line < lines ;line++) {
308 unsigned long word, words = procfile_linewords(ff, line);
311 char *name = procfile_lineword(ff, line, 0);
312 if(!name || !*name) continue;
314 // find a possibly existing target
315 struct target *w = NULL;
317 // loop through all words, skipping the first one (the name)
318 for(word = 0; word < words ;word++) {
319 char *s = procfile_lineword(ff, line, word);
320 if(!s || !*s) continue;
323 // is this the first word? skip it
324 if(s == name) continue;
327 struct target *n = get_apps_groups_target(s, w, name);
329 error("Cannot create target '%s' (line %lu, word %lu)", s, line, word);
333 // just some optimization
334 // to avoid searching for a target for each process
335 if(!w) w = n->target?n->target:n;
341 apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL, "other"); // match nothing
342 if(!apps_groups_default_target)
343 fatal("Cannot create default target");
345 // allow the user to override group 'other'
346 if(apps_groups_default_target->target)
347 apps_groups_default_target = apps_groups_default_target->target;
353 // ----------------------------------------------------------------------------
354 // data to store for each pid
357 #define PID_LOG_IO 0x00000001
358 #define PID_LOG_STATM 0x00000002
359 #define PID_LOG_CMDLINE 0x00000004
360 #define PID_LOG_FDS 0x00000008
361 #define PID_LOG_STAT 0x00000010
365 char comm[MAX_COMPARE_NAME + 1];
366 char cmdline[MAX_CMDLINE + 1];
378 // these are raw values collected
379 unsigned long long minflt_raw;
380 unsigned long long cminflt_raw;
381 unsigned long long majflt_raw;
382 unsigned long long cmajflt_raw;
383 unsigned long long utime_raw;
384 unsigned long long stime_raw;
385 unsigned long long gtime_raw; // guest_time
386 unsigned long long cutime_raw;
387 unsigned long long cstime_raw;
388 unsigned long long cgtime_raw; // cguest_time
391 unsigned long long minflt;
392 unsigned long long cminflt;
393 unsigned long long majflt;
394 unsigned long long cmajflt;
395 unsigned long long utime;
396 unsigned long long stime;
397 unsigned long long gtime;
398 unsigned long long cutime;
399 unsigned long long cstime;
400 unsigned long long cgtime;
405 // int64_t itrealvalue;
406 // unsigned long long starttime;
407 // unsigned long long vsize;
408 // unsigned long long rss;
409 // unsigned long long rsslim;
410 // unsigned long long starcode;
411 // unsigned long long endcode;
412 // unsigned long long startstack;
413 // unsigned long long kstkesp;
414 // unsigned long long kstkeip;
417 // uint64_t sigignore;
418 // uint64_t sigcatch;
422 // int32_t exit_signal;
423 // int32_t processor;
424 // uint32_t rt_priority;
426 // unsigned long long delayacct_blkio_ticks;
431 unsigned long long statm_size;
432 unsigned long long statm_resident;
433 unsigned long long statm_share;
434 // unsigned long long statm_text;
435 // unsigned long long statm_lib;
436 // unsigned long long statm_data;
437 // unsigned long long statm_dirty;
439 unsigned long long io_logical_bytes_read_raw;
440 unsigned long long io_logical_bytes_written_raw;
441 // unsigned long long io_read_calls_raw;
442 // unsigned long long io_write_calls_raw;
443 unsigned long long io_storage_bytes_read_raw;
444 unsigned long long io_storage_bytes_written_raw;
445 // unsigned long long io_cancelled_write_bytes_raw;
447 unsigned long long io_logical_bytes_read;
448 unsigned long long io_logical_bytes_written;
449 // unsigned long long io_read_calls;
450 // unsigned long long io_write_calls;
451 unsigned long long io_storage_bytes_read;
452 unsigned long long io_storage_bytes_written;
453 // unsigned long long io_cancelled_write_bytes;
455 int *fds; // array of fds it uses
456 int fds_size; // the size of the fds array
458 int children_count; // number of processes directly referencing this
459 int keep; // 1 when we need to keep this process in memory even after it exited
460 int keeploops; // increases by 1 every time keep is 1 and updated 0
461 int updated; // 1 when the process is currently running
462 int merged; // 1 when it has been merged to its parent
463 int new_entry; // 1 when this is a new process, just saw for the first time
464 int read; // 1 when we have already read this process for this iteration
465 int sortlist; // higher numbers = top on the process tree
466 // each process gets a unique number
468 struct target *target; // app_groups.conf targets
469 struct target *user_target; // uid based targets
470 struct target *group_target; // gid based targets
472 unsigned long long stat_collected_usec;
473 unsigned long long last_stat_collected_usec;
475 unsigned long long io_collected_usec;
476 unsigned long long last_io_collected_usec;
479 char *statm_filename;
481 char *cmdline_filename;
483 struct pid_stat *parent;
484 struct pid_stat *prev;
485 struct pid_stat *next;
486 } *root_of_pids = NULL, **all_pids;
488 long all_pids_count = 0;
490 static inline struct pid_stat *get_pid_entry(pid_t pid) {
492 all_pids[pid]->new_entry = 0;
493 return all_pids[pid];
496 all_pids[pid] = callocz(sizeof(struct pid_stat), 1);
497 all_pids[pid]->fds = callocz(sizeof(int), MAX_SPARE_FDS);
498 all_pids[pid]->fds_size = MAX_SPARE_FDS;
500 if(root_of_pids) root_of_pids->prev = all_pids[pid];
501 all_pids[pid]->next = root_of_pids;
502 root_of_pids = all_pids[pid];
504 all_pids[pid]->pid = pid;
505 all_pids[pid]->new_entry = 1;
509 return all_pids[pid];
512 static inline void del_pid_entry(pid_t pid) {
514 error("attempted to free pid %d that is not allocated.", pid);
519 fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
521 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
522 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
523 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
525 if(all_pids[pid]->fds) freez(all_pids[pid]->fds);
526 if(all_pids[pid]->stat_filename) freez(all_pids[pid]->stat_filename);
527 if(all_pids[pid]->statm_filename) freez(all_pids[pid]->statm_filename);
528 if(all_pids[pid]->io_filename) freez(all_pids[pid]->io_filename);
529 if(all_pids[pid]->cmdline_filename) freez(all_pids[pid]->cmdline_filename);
530 freez(all_pids[pid]);
532 all_pids[pid] = NULL;
537 // ----------------------------------------------------------------------------
538 // update pids from proc
540 static inline int read_proc_pid_cmdline(struct pid_stat *p) {
542 if(unlikely(!p->cmdline_filename)) {
543 char filename[FILENAME_MAX + 1];
544 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", global_host_prefix, p->pid);
545 p->cmdline_filename = strdupz(filename);
548 int fd = open(p->cmdline_filename, O_RDONLY, 0666);
549 if(unlikely(fd == -1)) goto cleanup;
551 ssize_t i, bytes = read(fd, p->cmdline, MAX_CMDLINE);
554 if(unlikely(bytes < 0)) goto cleanup;
556 p->cmdline[bytes] = '\0';
557 for(i = 0; i < bytes ; i++)
558 if(unlikely(!p->cmdline[i])) p->cmdline[i] = ' ';
561 fprintf(stderr, "Read file '%s' contents: %s\n", p->cmdline_filename, p->cmdline);
566 // copy the command to the command line
567 strncpyz(p->cmdline, p->comm, MAX_CMDLINE);
571 static inline int read_proc_pid_ownership(struct pid_stat *p) {
572 if(unlikely(!p->stat_filename)) {
573 error("pid %d does not have a stat_filename", p->pid);
577 // ----------------------------------------
581 if(stat(p->stat_filename, &st) != 0) {
582 error("Cannot stat file '%s'", p->stat_filename);
592 static inline int read_proc_pid_stat(struct pid_stat *p) {
593 static procfile *ff = NULL;
595 if(unlikely(!p->stat_filename)) {
596 char filename[FILENAME_MAX + 1];
597 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", global_host_prefix, p->pid);
598 p->stat_filename = strdupz(filename);
601 int set_quotes = (!ff)?1:0;
603 ff = procfile_reopen(ff, p->stat_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
604 if(unlikely(!ff)) goto cleanup;
606 // if(set_quotes) procfile_set_quotes(ff, "()");
607 if(set_quotes) procfile_set_open_close(ff, "(", ")");
609 ff = procfile_readall(ff);
610 if(unlikely(!ff)) goto cleanup;
612 p->last_stat_collected_usec = p->stat_collected_usec;
613 p->stat_collected_usec = now_monotonic_usec();
616 // p->pid = str2ul(procfile_lineword(ff, 0, 0+i));
618 strncpyz(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
620 // p->state = *(procfile_lineword(ff, 0, 2));
621 p->ppid = (int32_t)str2ul(procfile_lineword(ff, 0, 3));
622 // p->pgrp = str2ul(procfile_lineword(ff, 0, 4));
623 // p->session = str2ul(procfile_lineword(ff, 0, 5));
624 // p->tty_nr = str2ul(procfile_lineword(ff, 0, 6));
625 // p->tpgid = str2ul(procfile_lineword(ff, 0, 7));
626 // p->flags = str2ull(procfile_lineword(ff, 0, 8));
628 unsigned long long last;
630 last = p->minflt_raw;
631 p->minflt_raw = str2ull(procfile_lineword(ff, 0, 9));
632 p->minflt = (p->minflt_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
634 last = p->cminflt_raw;
635 p->cminflt_raw = str2ull(procfile_lineword(ff, 0, 10));
636 p->cminflt = (p->cminflt_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
638 last = p->majflt_raw;
639 p->majflt_raw = str2ull(procfile_lineword(ff, 0, 11));
640 p->majflt = (p->majflt_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
642 last = p->cmajflt_raw;
643 p->cmajflt_raw = str2ull(procfile_lineword(ff, 0, 12));
644 p->cmajflt = (p->cmajflt_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
647 p->utime_raw = str2ull(procfile_lineword(ff, 0, 13));
648 p->utime = (p->utime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
651 p->stime_raw = str2ull(procfile_lineword(ff, 0, 14));
652 p->stime = (p->stime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
654 last = p->cutime_raw;
655 p->cutime_raw = str2ull(procfile_lineword(ff, 0, 15));
656 p->cutime = (p->cutime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
658 last = p->cstime_raw;
659 p->cstime_raw = str2ull(procfile_lineword(ff, 0, 16));
660 p->cstime = (p->cstime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
662 // p->priority = str2ull(procfile_lineword(ff, 0, 17));
663 // p->nice = str2ull(procfile_lineword(ff, 0, 18));
664 p->num_threads = (int32_t)str2ul(procfile_lineword(ff, 0, 19));
665 // p->itrealvalue = str2ull(procfile_lineword(ff, 0, 20));
666 // p->starttime = str2ull(procfile_lineword(ff, 0, 21));
667 // p->vsize = str2ull(procfile_lineword(ff, 0, 22));
668 // p->rss = str2ull(procfile_lineword(ff, 0, 23));
669 // p->rsslim = str2ull(procfile_lineword(ff, 0, 24));
670 // p->starcode = str2ull(procfile_lineword(ff, 0, 25));
671 // p->endcode = str2ull(procfile_lineword(ff, 0, 26));
672 // p->startstack = str2ull(procfile_lineword(ff, 0, 27));
673 // p->kstkesp = str2ull(procfile_lineword(ff, 0, 28));
674 // p->kstkeip = str2ull(procfile_lineword(ff, 0, 29));
675 // p->signal = str2ull(procfile_lineword(ff, 0, 30));
676 // p->blocked = str2ull(procfile_lineword(ff, 0, 31));
677 // p->sigignore = str2ull(procfile_lineword(ff, 0, 32));
678 // p->sigcatch = str2ull(procfile_lineword(ff, 0, 33));
679 // p->wchan = str2ull(procfile_lineword(ff, 0, 34));
680 // p->nswap = str2ull(procfile_lineword(ff, 0, 35));
681 // p->cnswap = str2ull(procfile_lineword(ff, 0, 36));
682 // p->exit_signal = str2ul(procfile_lineword(ff, 0, 37));
683 // p->processor = str2ul(procfile_lineword(ff, 0, 38));
684 // p->rt_priority = str2ul(procfile_lineword(ff, 0, 39));
685 // p->policy = str2ul(procfile_lineword(ff, 0, 40));
686 // p->delayacct_blkio_ticks = str2ull(procfile_lineword(ff, 0, 41));
688 if(enable_guest_charts) {
690 p->gtime_raw = str2ull(procfile_lineword(ff, 0, 42));
691 p->gtime = (p->gtime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
693 last = p->cgtime_raw;
694 p->cgtime_raw = str2ull(procfile_lineword(ff, 0, 43));
695 p->cgtime = (p->cgtime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
697 if (show_guest_time || p->gtime || p->cgtime) {
698 p->utime -= (p->utime >= p->gtime) ? p->gtime : p->utime;
699 p->cutime -= (p->cutime >= p->cgtime) ? p->cgtime : p->cutime;
704 if(unlikely(debug || (p->target && p->target->debug)))
705 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", global_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);
707 if(unlikely(global_iterations_counter == 1)) {
738 static inline int read_proc_pid_statm(struct pid_stat *p) {
739 static procfile *ff = NULL;
741 if(unlikely(!p->statm_filename)) {
742 char filename[FILENAME_MAX + 1];
743 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/statm", global_host_prefix, p->pid);
744 p->statm_filename = strdupz(filename);
747 ff = procfile_reopen(ff, p->statm_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
748 if(unlikely(!ff)) goto cleanup;
750 ff = procfile_readall(ff);
751 if(unlikely(!ff)) goto cleanup;
755 p->statm_size = str2ull(procfile_lineword(ff, 0, 0));
756 p->statm_resident = str2ull(procfile_lineword(ff, 0, 1));
757 p->statm_share = str2ull(procfile_lineword(ff, 0, 2));
758 // p->statm_text = str2ull(procfile_lineword(ff, 0, 3));
759 // p->statm_lib = str2ull(procfile_lineword(ff, 0, 4));
760 // p->statm_data = str2ull(procfile_lineword(ff, 0, 5));
761 // p->statm_dirty = str2ull(procfile_lineword(ff, 0, 6));
767 p->statm_resident = 0;
769 // p->statm_text = 0;
771 // p->statm_data = 0;
772 // p->statm_dirty = 0;
776 static inline int read_proc_pid_io(struct pid_stat *p) {
777 static procfile *ff = NULL;
779 if(unlikely(!p->io_filename)) {
780 char filename[FILENAME_MAX + 1];
781 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/io", global_host_prefix, p->pid);
782 p->io_filename = strdupz(filename);
786 ff = procfile_reopen(ff, p->io_filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
787 if(unlikely(!ff)) goto cleanup;
789 ff = procfile_readall(ff);
790 if(unlikely(!ff)) goto cleanup;
794 p->last_io_collected_usec = p->io_collected_usec;
795 p->io_collected_usec = now_monotonic_usec();
797 unsigned long long last;
799 last = p->io_logical_bytes_read_raw;
800 p->io_logical_bytes_read_raw = str2ull(procfile_lineword(ff, 0, 1));
801 p->io_logical_bytes_read = (p->io_logical_bytes_read_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
803 last = p->io_logical_bytes_written_raw;
804 p->io_logical_bytes_written_raw = str2ull(procfile_lineword(ff, 1, 1));
805 p->io_logical_bytes_written = (p->io_logical_bytes_written_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
807 // last = p->io_read_calls_raw;
808 // p->io_read_calls_raw = str2ull(procfile_lineword(ff, 2, 1));
809 // p->io_read_calls = (p->io_read_calls_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
811 // last = p->io_write_calls_raw;
812 // p->io_write_calls_raw = str2ull(procfile_lineword(ff, 3, 1));
813 // p->io_write_calls = (p->io_write_calls_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
815 last = p->io_storage_bytes_read_raw;
816 p->io_storage_bytes_read_raw = str2ull(procfile_lineword(ff, 4, 1));
817 p->io_storage_bytes_read = (p->io_storage_bytes_read_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
819 last = p->io_storage_bytes_written_raw;
820 p->io_storage_bytes_written_raw = str2ull(procfile_lineword(ff, 5, 1));
821 p->io_storage_bytes_written = (p->io_storage_bytes_written_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
823 // last = p->io_cancelled_write_bytes_raw;
824 // p->io_cancelled_write_bytes_raw = str2ull(procfile_lineword(ff, 6, 1));
825 // p->io_cancelled_write_bytes = (p->io_cancelled_write_bytes_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (p->io_collected_usec - p->last_io_collected_usec);
827 if(unlikely(global_iterations_counter == 1)) {
828 p->io_logical_bytes_read = 0;
829 p->io_logical_bytes_written = 0;
830 // p->io_read_calls = 0;
831 // p->io_write_calls = 0;
832 p->io_storage_bytes_read = 0;
833 p->io_storage_bytes_written = 0;
834 // p->io_cancelled_write_bytes = 0;
840 p->io_logical_bytes_read = 0;
841 p->io_logical_bytes_written = 0;
842 // p->io_read_calls = 0;
843 // p->io_write_calls = 0;
844 p->io_storage_bytes_read = 0;
845 p->io_storage_bytes_written = 0;
846 // p->io_cancelled_write_bytes = 0;
850 unsigned long long global_utime = 0;
851 unsigned long long global_stime = 0;
852 unsigned long long global_gtime = 0;
854 static inline int read_proc_stat() {
855 static char filename[FILENAME_MAX + 1] = "";
856 static procfile *ff = NULL;
857 static unsigned long long utime_raw = 0, stime_raw = 0, gtime_raw = 0, gntime_raw = 0, ntime_raw = 0;
858 static usec_t collected_usec = 0, last_collected_usec = 0;
861 snprintfz(filename, FILENAME_MAX, "%s/proc/stat", global_host_prefix);
862 ff = procfile_open(filename, " \t:", PROCFILE_FLAG_DEFAULT);
863 if(unlikely(!ff)) goto cleanup;
866 ff = procfile_readall(ff);
867 if(unlikely(!ff)) goto cleanup;
869 last_collected_usec = collected_usec;
870 collected_usec = now_monotonic_usec();
874 unsigned long long last;
877 utime_raw = str2ull(procfile_lineword(ff, 0, 1));
878 global_utime = (utime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (collected_usec - last_collected_usec);
880 // nice time, on user time
882 ntime_raw = str2ull(procfile_lineword(ff, 0, 2));
883 global_utime += (ntime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (collected_usec - last_collected_usec);
886 stime_raw = str2ull(procfile_lineword(ff, 0, 3));
887 global_stime = (stime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (collected_usec - last_collected_usec);
890 gtime_raw = str2ull(procfile_lineword(ff, 0, 10));
891 global_gtime = (gtime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (collected_usec - last_collected_usec);
893 if(enable_guest_charts) {
894 // guest nice time, on guest time
896 gntime_raw = str2ull(procfile_lineword(ff, 0, 11));
897 global_gtime += (gntime_raw - last) * (USEC_PER_SEC * RATES_DETAIL) / (collected_usec - last_collected_usec);
899 // remove guest time from user time
900 global_utime -= (global_utime > global_gtime) ? global_gtime : global_utime;
903 if(unlikely(global_iterations_counter == 1)) {
919 // ----------------------------------------------------------------------------
921 // this is used to keep a global list of all open files of the system
922 // it is needed in order to calculate the unique files processes have open
924 #define FILE_DESCRIPTORS_INCREASE_STEP 100
926 struct file_descriptor {
928 #ifdef NETDATA_INTERNAL_CHECKS
930 #endif /* NETDATA_INTERNAL_CHECKS */
938 int all_files_len = 0;
939 int all_files_size = 0;
941 int file_descriptor_compare(void* a, void* b) {
942 #ifdef NETDATA_INTERNAL_CHECKS
943 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
944 error("Corrupted index data detected. Please report this.");
945 #endif /* NETDATA_INTERNAL_CHECKS */
947 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
950 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
954 return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
957 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
959 avl_tree all_files_index = {
961 file_descriptor_compare
964 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
965 struct file_descriptor tmp;
966 tmp.hash = (hash)?hash:simple_hash(name);
970 #ifdef NETDATA_INTERNAL_CHECKS
971 tmp.magic = 0x0BADCAFE;
972 #endif /* NETDATA_INTERNAL_CHECKS */
974 return (struct file_descriptor *)avl_search(&all_files_index, (avl *) &tmp);
977 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
978 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
980 #define FILETYPE_OTHER 0
981 #define FILETYPE_FILE 1
982 #define FILETYPE_PIPE 2
983 #define FILETYPE_SOCKET 3
984 #define FILETYPE_INOTIFY 4
985 #define FILETYPE_EVENTFD 5
986 #define FILETYPE_EVENTPOLL 6
987 #define FILETYPE_TIMERFD 7
988 #define FILETYPE_SIGNALFD 8
990 static inline void file_descriptor_not_used(int id)
992 if(id > 0 && id < all_files_size) {
994 #ifdef NETDATA_INTERNAL_CHECKS
995 if(all_files[id].magic != 0x0BADCAFE) {
996 error("Ignoring request to remove empty file id %d.", id);
999 #endif /* NETDATA_INTERNAL_CHECKS */
1002 fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
1004 if(all_files[id].count > 0) {
1005 all_files[id].count--;
1007 if(!all_files[id].count) {
1009 fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
1011 if(unlikely(file_descriptor_remove(&all_files[id]) != (void *)&all_files[id]))
1012 error("INTERNAL ERROR: removal of unused fd from index, removed a different fd");
1014 #ifdef NETDATA_INTERNAL_CHECKS
1015 all_files[id].magic = 0x00000000;
1016 #endif /* NETDATA_INTERNAL_CHECKS */
1021 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
1023 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
1026 static inline void all_files_grow() {
1027 void *old = all_files;
1030 // there is no empty slot
1032 fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
1034 all_files = reallocz(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
1036 // if the address changed, we have to rebuild the index
1037 // since all pointers are now invalid
1039 if(unlikely(old && old != (void *)all_files)) {
1041 fprintf(stderr, "apps.plugin: >> re-indexing.\n");
1043 all_files_index.root = NULL;
1044 for(i = 0; i < all_files_size; i++) {
1045 if(!all_files[i].count) continue;
1046 if(unlikely(file_descriptor_add(&all_files[i]) != (void *)&all_files[i]))
1047 error("INTERNAL ERROR: duplicate indexing of fd during realloc.");
1051 fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
1054 // initialize the newly added entries
1056 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
1057 all_files[i].count = 0;
1058 all_files[i].name = NULL;
1059 #ifdef NETDATA_INTERNAL_CHECKS
1060 all_files[i].magic = 0x00000000;
1061 #endif /* NETDATA_INTERNAL_CHECKS */
1062 all_files[i].pos = i;
1065 if(unlikely(!all_files_size)) all_files_len = 1;
1066 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
1069 static inline int file_descriptor_set_on_empty_slot(const char *name, uint32_t hash, int type) {
1070 // check we have enough memory to add it
1071 if(!all_files || all_files_len == all_files_size)
1075 fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
1077 // search for an empty slot
1079 static int last_pos = 0;
1081 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
1082 if(c >= all_files_size) c = 0;
1083 if(c == 0) continue;
1085 if(!all_files[c].count) {
1087 fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
1089 #ifdef NETDATA_INTERNAL_CHECKS
1090 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
1091 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
1092 #endif /* NETDATA_INTERNAL_CHECKS */
1095 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);
1097 freez((void *)all_files[c].name);
1098 all_files[c].name = NULL;
1106 if(i == all_files_size) {
1107 fatal("We should find an empty slot, but there isn't any");
1110 // else we have an empty slot in 'c'
1113 fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
1115 all_files[c].name = strdupz(name);
1116 all_files[c].hash = hash;
1117 all_files[c].type = type;
1118 all_files[c].pos = c;
1119 all_files[c].count = 1;
1120 #ifdef NETDATA_INTERNAL_CHECKS
1121 all_files[c].magic = 0x0BADCAFE;
1122 #endif /* NETDATA_INTERNAL_CHECKS */
1123 if(unlikely(file_descriptor_add(&all_files[c]) != (void *)&all_files[c]))
1124 error("INTERNAL ERROR: duplicate indexing of fd.");
1127 fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1132 static inline int file_descriptor_find_or_add(const char *name)
1134 uint32_t hash = simple_hash(name);
1137 fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
1139 struct file_descriptor *fd = file_descriptor_find(name, hash);
1143 fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
1151 if(name[0] == '/') type = FILETYPE_FILE;
1152 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
1153 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
1154 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
1155 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
1156 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
1157 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
1158 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
1159 else if(strncmp(name, "anon_inode:", 11) == 0) {
1161 fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
1163 type = FILETYPE_OTHER;
1167 fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
1169 type = FILETYPE_OTHER;
1172 return file_descriptor_set_on_empty_slot(name, hash, type);
1175 static inline int read_pid_file_descriptors(struct pid_stat *p) {
1176 char dirname[FILENAME_MAX+1];
1178 snprintfz(dirname, FILENAME_MAX, "%s/proc/%d/fd", global_host_prefix, p->pid);
1179 DIR *fds = opendir(dirname);
1183 char fdname[FILENAME_MAX + 1];
1184 char linkname[FILENAME_MAX + 1];
1186 // make the array negative
1187 for(c = 0 ; c < p->fds_size ; c++)
1188 p->fds[c] = -p->fds[c];
1190 while((de = readdir(fds))) {
1191 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
1194 // check if the fds array is small
1195 int fdid = (int)str2l(de->d_name);
1196 if(fdid < 0) continue;
1197 if(fdid >= p->fds_size) {
1198 // it is small, extend it
1200 fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + MAX_SPARE_FDS);
1202 p->fds = reallocz(p->fds, (fdid + MAX_SPARE_FDS) * sizeof(int));
1204 // and initialize it
1205 for(c = p->fds_size ; c < (fdid + MAX_SPARE_FDS) ; c++) p->fds[c] = 0;
1206 p->fds_size = fdid + MAX_SPARE_FDS;
1209 if(p->fds[fdid] == 0) {
1210 // we don't know this fd, get it
1212 sprintf(fdname, "%s/proc/%d/fd/%s", global_host_prefix, p->pid, de->d_name);
1213 ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
1215 if(debug || (p->target && p->target->debug)) {
1216 if(debug || (p->target && p->target->debug))
1217 error("Cannot read link %s", fdname);
1224 // if another process already has this, we will get
1226 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1229 // else make it positive again, we need it
1230 // of course, the actual file may have changed, but we don't care so much
1231 // FIXME: we could compare the inode as returned by readdir dirent structure
1232 else p->fds[fdid] = -p->fds[fdid];
1236 // remove all the negative file descriptors
1237 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1238 file_descriptor_not_used(-p->fds[c]);
1247 // ----------------------------------------------------------------------------
1249 static inline int print_process_and_parents(struct pid_stat *p, unsigned long long time) {
1250 char *prefix = "\\_ ";
1254 indent = print_process_and_parents(p->parent, p->stat_collected_usec);
1258 char buffer[indent + 1];
1261 for(i = 0; i < indent ;i++) buffer[i] = ' ';
1264 fprintf(stderr, " %s %s%s (%d %s %lld"
1269 , p->updated?"running":"exited"
1270 , (long long)p->stat_collected_usec - (long long)time
1273 if(p->utime) fprintf(stderr, " utime=%llu", p->utime);
1274 if(p->stime) fprintf(stderr, " stime=%llu", p->stime);
1275 if(p->gtime) fprintf(stderr, " gtime=%llu", p->gtime);
1276 if(p->cutime) fprintf(stderr, " cutime=%llu", p->cutime);
1277 if(p->cstime) fprintf(stderr, " cstime=%llu", p->cstime);
1278 if(p->cgtime) fprintf(stderr, " cgtime=%llu", p->cgtime);
1279 if(p->minflt) fprintf(stderr, " minflt=%llu", p->minflt);
1280 if(p->cminflt) fprintf(stderr, " cminflt=%llu", p->cminflt);
1281 if(p->majflt) fprintf(stderr, " majflt=%llu", p->majflt);
1282 if(p->cmajflt) fprintf(stderr, " cmajflt=%llu", p->cmajflt);
1283 fprintf(stderr, ")\n");
1288 static inline void print_process_tree(struct pid_stat *p, char *msg) {
1290 fprintf(stderr, "%s: process %s (%d, %s) with parents:\n", msg, p->comm, p->pid, p->updated?"running":"exited");
1291 print_process_and_parents(p, p->stat_collected_usec);
1294 static inline void find_lost_child_debug(struct pid_stat *pe, unsigned long long lost, int type) {
1296 struct pid_stat *p = NULL;
1298 for(p = root_of_pids; p ; p = p->next) {
1299 if(p == pe) continue;
1303 if(p->cminflt > lost) {
1304 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);
1310 if(p->cmajflt > lost) {
1311 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);
1317 if(p->cutime > lost) {
1318 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);
1324 if(p->cstime > lost) {
1325 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);
1331 if(p->cgtime > lost) {
1332 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);
1342 fprintf(stderr, " > cannot find any process to use the lost exited child minflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1346 fprintf(stderr, " > cannot find any process to use the lost exited child majflt %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1350 fprintf(stderr, " > cannot find any process to use the lost exited child utime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1354 fprintf(stderr, " > cannot find any process to use the lost exited child stime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1358 fprintf(stderr, " > cannot find any process to use the lost exited child gtime %llu of process %d (%s)\n", lost, pe->pid, pe->comm);
1364 static inline unsigned long long remove_exited_child_from_parent(unsigned long long *field, unsigned long long *pfield) {
1365 unsigned long long absorbed = 0;
1367 if(*field > *pfield) {
1368 absorbed += *pfield;
1381 static inline void process_exited_processes() {
1384 for(p = root_of_pids; p ; p = p->next) {
1385 if(p->updated || !p->stat_collected_usec)
1388 struct pid_stat *pp = p->parent;
1390 unsigned long long utime = (p->utime_raw + p->cutime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1391 unsigned long long stime = (p->stime_raw + p->cstime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1392 unsigned long long gtime = (p->gtime_raw + p->cgtime_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1393 unsigned long long minflt = (p->minflt_raw + p->cminflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1394 unsigned long long majflt = (p->majflt_raw + p->cmajflt_raw) * (1000000ULL * RATES_DETAIL) / (p->stat_collected_usec - p->last_stat_collected_usec);
1396 if(utime + stime + gtime + minflt + majflt == 0)
1399 if(unlikely(debug)) {
1401 fprintf(stderr, "Absorb %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1404 , p->updated?"running":"exited"
1411 print_process_tree(p, "Searching parents");
1414 for(pp = p->parent; pp ; pp = pp->parent) {
1415 if(!pp->updated) continue;
1417 unsigned long long absorbed;
1418 absorbed = remove_exited_child_from_parent(&utime, &pp->cutime);
1419 if(unlikely(debug && absorbed))
1420 fprintf(stderr, " > process %s (%d %s) absorbed %llu utime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, utime);
1422 absorbed = remove_exited_child_from_parent(&stime, &pp->cstime);
1423 if(unlikely(debug && absorbed))
1424 fprintf(stderr, " > process %s (%d %s) absorbed %llu stime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, stime);
1426 absorbed = remove_exited_child_from_parent(>ime, &pp->cgtime);
1427 if(unlikely(debug && absorbed))
1428 fprintf(stderr, " > process %s (%d %s) absorbed %llu gtime (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, gtime);
1430 absorbed = remove_exited_child_from_parent(&minflt, &pp->cminflt);
1431 if(unlikely(debug && absorbed))
1432 fprintf(stderr, " > process %s (%d %s) absorbed %llu minflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, minflt);
1434 absorbed = remove_exited_child_from_parent(&majflt, &pp->cmajflt);
1435 if(unlikely(debug && absorbed))
1436 fprintf(stderr, " > process %s (%d %s) absorbed %llu majflt (remaining: %llu)\n", pp->comm, pp->pid, pp->updated?"running":"exited", absorbed, majflt);
1439 if(unlikely(utime + stime + gtime + minflt + majflt > 0)) {
1440 if(unlikely(debug)) {
1441 if(utime) find_lost_child_debug(p, utime, 3);
1442 if(stime) find_lost_child_debug(p, stime, 4);
1443 if(gtime) find_lost_child_debug(p, gtime, 5);
1444 if(minflt) find_lost_child_debug(p, minflt, 1);
1445 if(majflt) find_lost_child_debug(p, majflt, 2);
1451 fprintf(stderr, " > remaining resources - KEEP - for another loop: %s (%d %s total resources: utime=%llu stime=%llu gtime=%llu minflt=%llu majflt=%llu)\n"
1454 , p->updated?"running":"exited"
1462 for(pp = p->parent; pp ; pp = pp->parent) {
1463 if(pp->updated) break;
1467 fprintf(stderr, " > - KEEP - parent for another loop: %s (%d %s)\n"
1470 , pp->updated?"running":"exited"
1474 p->utime_raw = utime * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
1475 p->stime_raw = stime * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
1476 p->gtime_raw = gtime * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
1477 p->minflt_raw = minflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
1478 p->majflt_raw = majflt * (p->stat_collected_usec - p->last_stat_collected_usec) / (USEC_PER_SEC * RATES_DETAIL);
1479 p->cutime_raw = p->cstime_raw = p->cgtime_raw = p->cminflt_raw = p->cmajflt_raw = 0;
1482 fprintf(stderr, "\n");
1484 else if(unlikely(debug)) {
1485 fprintf(stderr, " > totally absorbed - DONE - %s (%d %s)\n"
1488 , p->updated?"running":"exited"
1494 static inline void link_all_processes_to_their_parents(void) {
1495 struct pid_stat *p, *pp;
1497 // link all children to their parents
1498 // and update children count on parents
1499 for(p = root_of_pids; p ; p = p->next) {
1500 // for each process found
1505 if(unlikely(!p->ppid)) {
1510 pp = all_pids[p->ppid];
1513 pp->children_count++;
1515 if(unlikely(debug || (p->target && p->target->debug)))
1516 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);
1520 error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1525 // ----------------------------------------------------------------------------
1527 // 1. read all files in /proc
1528 // 2. for each numeric directory:
1529 // i. read /proc/pid/stat
1530 // ii. read /proc/pid/statm
1531 // iii. read /proc/pid/io (requires root access)
1532 // iii. read the entries in directory /proc/pid/fd (requires root access)
1534 // a. find or create a struct file_descriptor
1535 // b. cleanup any old/unused file_descriptors
1537 // after all these, some pids may be linked to targets, while others may not
1539 // in case of errors, only 1 every 1000 errors is printed
1540 // to avoid filling up all disk space
1541 // if debug is enabled, all errors are printed
1543 static int compar_pid(const void *pid1, const void *pid2) {
1545 struct pid_stat *p1 = all_pids[*((pid_t *)pid1)];
1546 struct pid_stat *p2 = all_pids[*((pid_t *)pid2)];
1548 if(p1->sortlist > p2->sortlist)
1554 static inline int managed_log(struct pid_stat *p, uint32_t log, int status) {
1555 if(unlikely(!status)) {
1556 // error("command failed log %u, errno %d", log, errno);
1558 if(unlikely(debug || errno != ENOENT)) {
1559 if(unlikely(debug || !(p->log_thrown & log))) {
1560 p->log_thrown |= log;
1563 error("Cannot process %s/proc/%d/io (command '%s')", global_host_prefix, p->pid, p->comm);
1567 error("Cannot process %s/proc/%d/statm (command '%s')", global_host_prefix, p->pid, p->comm);
1570 case PID_LOG_CMDLINE:
1571 error("Cannot process %s/proc/%d/cmdline (command '%s')", global_host_prefix, p->pid, p->comm);
1575 error("Cannot process entries in %s/proc/%d/fd (command '%s')", global_host_prefix, p->pid, p->comm);
1582 error("unhandled error for pid %d, command '%s'", p->pid, p->comm);
1589 else if(unlikely(p->log_thrown & log)) {
1590 // error("unsetting log %u on pid %d", log, p->pid);
1591 p->log_thrown &= ~log;
1597 static inline int collect_data_for_pid(pid_t pid) {
1598 if(unlikely(pid <= 0 || pid > pid_max)) {
1599 error("Invalid pid %d read (expected 1 to %d). Ignoring process.", pid, pid_max);
1603 struct pid_stat *p = get_pid_entry(pid);
1604 if(unlikely(!p || p->read)) return 0;
1607 // fprintf(stderr, "Reading process %d (%s), sortlist %d\n", p->pid, p->comm, p->sortlist);
1609 // --------------------------------------------------------------------
1612 if(unlikely(!managed_log(p, PID_LOG_STAT, read_proc_pid_stat(p))))
1613 // there is no reason to proceed if we cannot get its status
1616 read_proc_pid_ownership(p);
1618 // check its parent pid
1619 if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
1620 error("Pid %d (command '%s') states invalid parent pid %d. Using 0.", pid, p->comm, p->ppid);
1624 // --------------------------------------------------------------------
1627 managed_log(p, PID_LOG_IO, read_proc_pid_io(p));
1629 // --------------------------------------------------------------------
1630 // /proc/<pid>/statm
1632 if(unlikely(!managed_log(p, PID_LOG_STATM, read_proc_pid_statm(p))))
1633 // there is no reason to proceed if we cannot get its memory status
1636 // --------------------------------------------------------------------
1639 // check if it is target
1640 // we do this only once, the first time this pid is loaded
1641 if(unlikely(p->new_entry)) {
1642 // /proc/<pid>/cmdline
1643 if(likely(proc_pid_cmdline_is_needed))
1644 managed_log(p, PID_LOG_CMDLINE, read_proc_pid_cmdline(p));
1647 fprintf(stderr, "apps.plugin: \tJust added %d (%s)\n", pid, p->comm);
1649 uint32_t hash = simple_hash(p->comm);
1650 size_t pclen = strlen(p->comm);
1653 for(w = apps_groups_root_target; w ; w = w->next) {
1654 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1656 // find it - 4 cases:
1657 // 1. the target is not a pattern
1658 // 2. the target has the prefix
1659 // 3. the target has the suffix
1660 // 4. the target is something inside cmdline
1661 if( (!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm))
1662 || (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen))
1663 || (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen]))
1664 || (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && strstr(p->cmdline, w->compare))
1666 if(w->target) p->target = w->target;
1669 if(debug || (p->target && p->target->debug))
1670 fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1677 // --------------------------------------------------------------------
1680 if(enable_file_charts)
1681 managed_log(p, PID_LOG_FDS, read_pid_file_descriptors(p));
1683 // --------------------------------------------------------------------
1686 if(unlikely(debug && include_exited_childs && all_pids_count && p->ppid && all_pids[p->ppid] && !all_pids[p->ppid]->read))
1687 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);
1689 // mark it as updated
1697 static int collect_data_for_all_processes_from_proc(void) {
1698 struct pid_stat *p = NULL;
1700 if(all_pids_count) {
1701 // read parents before childs
1702 // this is needed to prevent a situation where
1703 // a child is found running, but until we read
1704 // its parent, it has exited and its parent
1705 // has accumulated its resources
1708 for(p = root_of_pids; p ; p = p->next) {
1713 p->children_count = 0;
1716 all_pids_sortlist[slc++] = p->pid;
1719 if(unlikely(slc != all_pids_count)) {
1720 error("Internal error: I was thinking I had %ld processes in my arrays, but it seems there are more.", all_pids_count);
1721 all_pids_count = slc;
1724 if(include_exited_childs) {
1725 qsort((void *)all_pids_sortlist, (size_t)all_pids_count, sizeof(pid_t), compar_pid);
1726 for(slc = 0; slc < all_pids_count; slc++)
1727 collect_data_for_pid(all_pids_sortlist[slc]);
1731 char dirname[FILENAME_MAX + 1];
1733 snprintfz(dirname, FILENAME_MAX, "%s/proc", global_host_prefix);
1734 DIR *dir = opendir(dirname);
1737 struct dirent *file = NULL;
1739 while((file = readdir(dir))) {
1740 char *endptr = file->d_name;
1741 pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
1743 // make sure we read a valid number
1744 if(unlikely(endptr == file->d_name || *endptr != '\0'))
1747 collect_data_for_pid(pid);
1754 // normally this is done
1755 // however we may have processes exited while we collected values
1756 // so let's find the exited ones
1757 // we do this by collecting the ownership of process
1758 // if we manage to get the ownership, the process still runs
1761 link_all_processes_to_their_parents();
1762 process_exited_processes();
1767 // ----------------------------------------------------------------------------
1768 // update statistics on the targets
1770 // 1. link all childs to their parents
1771 // 2. go from bottom to top, marking as merged all childs to their parents
1772 // this step links all parents without a target to the child target, if any
1773 // 3. link all top level processes (the ones not merged) to the default target
1774 // 4. go from top to bottom, linking all childs without a target, to their parent target
1775 // after this step, all processes have a target
1776 // [5. for each killed pid (updated = 0), remove its usage from its target]
1777 // 6. zero all apps_groups_targets
1778 // 7. concentrate all values on the apps_groups_targets
1779 // 8. remove all killed processes
1780 // 9. find the unique file count for each target
1781 // check: update_apps_groups_statistics()
1783 static void cleanup_exited_pids(void) {
1785 struct pid_stat *p = NULL;
1787 for(p = root_of_pids; p ;) {
1788 if(!p->updated && (!p->keep || p->keeploops > 0)) {
1789 // 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);
1791 if(unlikely(debug && (p->keep || p->keeploops)))
1792 fprintf(stderr, " > CLEANUP cannot keep exited process %d (%s) anymore - removing it.\n", p->pid, p->comm);
1794 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1795 file_descriptor_not_used(p->fds[c]);
1804 if(unlikely(p->keep)) p->keeploops++;
1811 static void apply_apps_groups_targets_inheritance(void) {
1812 struct pid_stat *p = NULL;
1814 // children that do not have a target
1815 // inherit their target from their parent
1816 int found = 1, loops = 0;
1818 if(unlikely(debug)) loops++;
1820 for(p = root_of_pids; p ; p = p->next) {
1821 // if this process does not have a target
1822 // and it has a parent
1823 // and its parent has a target
1824 // then, set the parent's target to this process
1825 if(unlikely(!p->target && p->parent && p->parent->target)) {
1826 p->target = p->parent->target;
1829 if(debug || (p->target && p->target->debug))
1830 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);
1835 // find all the procs with 0 childs and merge them to their parents
1836 // repeat, until nothing more can be done.
1840 if(unlikely(debug)) loops++;
1843 for(p = root_of_pids; p ; p = p->next) {
1844 if(unlikely(!p->sortlist && !p->children_count))
1845 p->sortlist = sortlist++;
1847 // if this process does not have any children
1848 // and is not already merged
1850 // and its parent has children
1851 // and the target of this process and its parent is the same, or the parent does not have a target
1852 // and its parent is not init
1853 // then, mark them as merged.
1858 && p->parent->children_count
1859 && (p->target == p->parent->target || !p->parent->target)
1862 p->parent->children_count--;
1865 // the parent inherits the child's target, if it does not have a target itself
1866 if(unlikely(p->target && !p->parent->target)) {
1867 p->parent->target = p->target;
1869 if(debug || (p->target && p->target->debug))
1870 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);
1878 fprintf(stderr, "apps.plugin: TARGET INHERITANCE: merged %d processes\n", found);
1881 // init goes always to default target
1883 all_pids[1]->target = apps_groups_default_target;
1885 // give a default target on all top level processes
1886 if(unlikely(debug)) loops++;
1887 for(p = root_of_pids; p ; p = p->next) {
1888 // if the process is not merged itself
1889 // then is is a top level process
1890 if(unlikely(!p->merged && !p->target))
1891 p->target = apps_groups_default_target;
1893 // make sure all processes have a sortlist
1894 if(unlikely(!p->sortlist))
1895 p->sortlist = sortlist++;
1899 all_pids[1]->sortlist = sortlist++;
1901 // give a target to all merged child processes
1904 if(unlikely(debug)) loops++;
1906 for(p = root_of_pids; p ; p = p->next) {
1907 if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
1908 p->target = p->parent->target;
1911 if(debug || (p->target && p->target->debug))
1912 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);
1918 fprintf(stderr, "apps.plugin: apply_apps_groups_targets_inheritance() made %d loops on the process tree\n", loops);
1921 static long zero_all_targets(struct target *root) {
1925 for (w = root; w ; w = w->next) {
1943 w->statm_resident = 0;
1945 // w->statm_text = 0;
1946 // w->statm_lib = 0;
1947 // w->statm_data = 0;
1948 // w->statm_dirty = 0;
1950 w->io_logical_bytes_read = 0;
1951 w->io_logical_bytes_written = 0;
1952 // w->io_read_calls = 0;
1953 // w->io_write_calls = 0;
1954 w->io_storage_bytes_read = 0;
1955 w->io_storage_bytes_written = 0;
1956 // w->io_cancelled_write_bytes = 0;
1958 // zero file counters
1960 memset(w->target_fds, 0, sizeof(int) * w->target_fds_size);
1964 w->openinotifies = 0;
1965 w->openeventfds = 0;
1966 w->opentimerfds = 0;
1967 w->opensignalfds = 0;
1968 w->openeventpolls = 0;
1976 static inline void reallocate_target_fds(struct target *w) {
1980 if(unlikely(!w->target_fds || w->target_fds_size < all_files_size)) {
1981 w->target_fds = reallocz(w->target_fds, sizeof(int) * all_files_size);
1982 memset(&w->target_fds[w->target_fds_size], 0, sizeof(int) * (all_files_size - w->target_fds_size));
1983 w->target_fds_size = all_files_size;
1987 static inline void aggregate_fd_on_target(int fd, struct target *w) {
1991 if(unlikely(w->target_fds[fd])) {
1992 // it is already aggregated
1993 // just increase its usage counter
1994 w->target_fds[fd]++;
1998 // increase its usage counter
1999 // so that we will not add it again
2000 w->target_fds[fd]++;
2002 switch(all_files[fd].type) {
2011 case FILETYPE_SOCKET:
2015 case FILETYPE_INOTIFY:
2019 case FILETYPE_EVENTFD:
2023 case FILETYPE_TIMERFD:
2027 case FILETYPE_SIGNALFD:
2031 case FILETYPE_EVENTPOLL:
2032 w->openeventpolls++;
2041 static inline void aggregate_pid_fds_on_targets(struct pid_stat *p) {
2043 if(unlikely(!p->updated)) {
2044 // the process is not running
2048 struct target *w = p->target, *u = p->user_target, *g = p->group_target;
2050 reallocate_target_fds(w);
2051 reallocate_target_fds(u);
2052 reallocate_target_fds(g);
2054 int c, size = p->fds_size, *fds = p->fds;
2055 for(c = 0; c < size ;c++) {
2058 if(likely(fd <= 0 || fd >= all_files_size))
2061 aggregate_fd_on_target(fd, w);
2062 aggregate_fd_on_target(fd, u);
2063 aggregate_fd_on_target(fd, g);
2067 static inline void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) {
2070 if(unlikely(!p->updated)) {
2071 // the process is not running
2076 error("pid %d %s was left without a target!", p->pid, p->comm);
2080 w->cutime += p->cutime;
2081 w->cstime += p->cstime;
2082 w->cgtime += p->cgtime;
2083 w->cminflt += p->cminflt;
2084 w->cmajflt += p->cmajflt;
2086 w->utime += p->utime;
2087 w->stime += p->stime;
2088 w->gtime += p->gtime;
2089 w->minflt += p->minflt;
2090 w->majflt += p->majflt;
2092 // w->rss += p->rss;
2094 w->statm_size += p->statm_size;
2095 w->statm_resident += p->statm_resident;
2096 w->statm_share += p->statm_share;
2097 // w->statm_text += p->statm_text;
2098 // w->statm_lib += p->statm_lib;
2099 // w->statm_data += p->statm_data;
2100 // w->statm_dirty += p->statm_dirty;
2102 w->io_logical_bytes_read += p->io_logical_bytes_read;
2103 w->io_logical_bytes_written += p->io_logical_bytes_written;
2104 // w->io_read_calls += p->io_read_calls;
2105 // w->io_write_calls += p->io_write_calls;
2106 w->io_storage_bytes_read += p->io_storage_bytes_read;
2107 w->io_storage_bytes_written += p->io_storage_bytes_written;
2108 // w->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
2111 w->num_threads += p->num_threads;
2113 if(unlikely(debug || w->debug))
2114 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);
2117 static void calculate_netdata_statistics(void) {
2119 apply_apps_groups_targets_inheritance();
2121 zero_all_targets(users_root_target);
2122 zero_all_targets(groups_root_target);
2123 apps_groups_targets = zero_all_targets(apps_groups_root_target);
2125 // this has to be done, before the cleanup
2126 struct pid_stat *p = NULL;
2127 struct target *w = NULL, *o = NULL;
2129 // concentrate everything on the targets
2130 for(p = root_of_pids; p ; p = p->next) {
2132 // --------------------------------------------------------------------
2133 // apps_groups target
2135 aggregate_pid_on_target(p->target, p, NULL);
2138 // --------------------------------------------------------------------
2142 if(likely(p->user_target && p->user_target->uid == p->uid))
2145 if(unlikely(debug && p->user_target))
2146 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);
2148 w = p->user_target = get_users_target(p->uid);
2151 aggregate_pid_on_target(w, p, o);
2154 // --------------------------------------------------------------------
2155 // user group target
2157 o = p->group_target;
2158 if(likely(p->group_target && p->group_target->gid == p->gid))
2159 w = p->group_target;
2161 if(unlikely(debug && p->group_target))
2162 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);
2164 w = p->group_target = get_groups_target(p->gid);
2167 aggregate_pid_on_target(w, p, o);
2170 // --------------------------------------------------------------------
2171 // aggregate all file descriptors
2173 if(enable_file_charts)
2174 aggregate_pid_fds_on_targets(p);
2177 cleanup_exited_pids();
2180 // ----------------------------------------------------------------------------
2181 // update chart dimensions
2183 int print_calculated_number(char *str, calculated_number value) { (void)str; (void)value; return 0; }
2185 static inline void send_BEGIN(const char *type, const char *id, unsigned long long usec) {
2186 fprintf(stdout, "BEGIN %s.%s %llu\n", type, id, usec);
2189 static inline void send_SET(const char *name, unsigned long long value) {
2190 fprintf(stdout, "SET %s = %llu\n", name, value);
2193 static inline void send_END(void) {
2194 fprintf(stdout, "END\n");
2197 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;
2198 double minflt_fix_ratio = 1.0, majflt_fix_ratio = 1.0, cminflt_fix_ratio = 1.0, cmajflt_fix_ratio = 1.0;
2200 static usec_t send_resource_usage_to_netdata() {
2201 static struct timeval last = { 0, 0 };
2202 static struct rusage me_last;
2212 now_monotonic_timeval(&last);
2213 getrusage(RUSAGE_SELF, &me_last);
2215 // the first time, give a zero to allow
2216 // netdata calibrate to the current time
2217 // usec = update_every * USEC_PER_SEC;
2223 now_monotonic_timeval(&now);
2224 getrusage(RUSAGE_SELF, &me);
2226 usec = dt_usec(&now, &last);
2227 cpuuser = me.ru_utime.tv_sec * USEC_PER_SEC + me.ru_utime.tv_usec;
2228 cpusyst = me.ru_stime.tv_sec * USEC_PER_SEC + me.ru_stime.tv_usec;
2230 memmove(&last, &now, sizeof(struct timeval));
2231 memmove(&me_last, &me, sizeof(struct rusage));
2235 "BEGIN netdata.apps_cpu %llu\n"
2237 "SET system = %llu\n"
2239 "BEGIN netdata.apps_files %llu\n"
2240 "SET files = %llu\n"
2243 "SET targets = %ld\n"
2245 "BEGIN netdata.apps_fix %llu\n"
2246 "SET utime = %llu\n"
2247 "SET stime = %llu\n"
2248 "SET gtime = %llu\n"
2249 "SET minflt = %llu\n"
2250 "SET majflt = %llu\n"
2259 , apps_groups_targets
2261 , (unsigned long long)(utime_fix_ratio * 100 * RATES_DETAIL)
2262 , (unsigned long long)(stime_fix_ratio * 100 * RATES_DETAIL)
2263 , (unsigned long long)(gtime_fix_ratio * 100 * RATES_DETAIL)
2264 , (unsigned long long)(minflt_fix_ratio * 100 * RATES_DETAIL)
2265 , (unsigned long long)(majflt_fix_ratio * 100 * RATES_DETAIL)
2268 if(include_exited_childs)
2270 "BEGIN netdata.apps_children_fix %llu\n"
2271 "SET cutime = %llu\n"
2272 "SET cstime = %llu\n"
2273 "SET cgtime = %llu\n"
2274 "SET cminflt = %llu\n"
2275 "SET cmajflt = %llu\n"
2278 , (unsigned long long)(cutime_fix_ratio * 100 * RATES_DETAIL)
2279 , (unsigned long long)(cstime_fix_ratio * 100 * RATES_DETAIL)
2280 , (unsigned long long)(cgtime_fix_ratio * 100 * RATES_DETAIL)
2281 , (unsigned long long)(cminflt_fix_ratio * 100 * RATES_DETAIL)
2282 , (unsigned long long)(cmajflt_fix_ratio * 100 * RATES_DETAIL)
2288 static void normalize_data(struct target *root) {
2291 // childs processing introduces spikes
2292 // here we try to eliminate them by disabling childs processing either for specific dimensions
2293 // or entirely. Of course, either way, we disable it just a single iteration.
2295 unsigned long long max = processors * hz * RATES_DETAIL;
2296 unsigned long long utime = 0, cutime = 0, stime = 0, cstime = 0, gtime = 0, cgtime = 0, minflt = 0, cminflt = 0, majflt = 0, cmajflt = 0;
2298 if(global_utime > max) global_utime = max;
2299 if(global_stime > max) global_stime = max;
2300 if(global_gtime > max) global_gtime = max;
2302 for(w = root; w ; w = w->next) {
2303 if(w->target || (!w->processes && !w->exposed)) continue;
2308 cutime += w->cutime;
2309 cstime += w->cstime;
2310 cgtime += w->cgtime;
2312 minflt += w->minflt;
2313 majflt += w->majflt;
2314 cminflt += w->cminflt;
2315 cmajflt += w->cmajflt;
2318 if((global_utime || global_stime || global_gtime) && (utime || stime || gtime)) {
2319 if(global_utime + global_stime + global_gtime > utime + cutime + stime + cstime + gtime + cgtime) {
2320 // everything we collected fits
2326 cgtime_fix_ratio = 1.0; //(double)(global_utime + global_stime) / (double)(utime + cutime + stime + cstime);
2328 else if(global_utime + global_stime > utime + stime) {
2329 // childrens resources are too high
2330 // lower only the children resources
2333 gtime_fix_ratio = 1.0;
2336 cgtime_fix_ratio = (double)((global_utime + global_stime) - (utime + stime)) / (double)(cutime + cstime);
2339 // even running processes are unrealistic
2340 // zero the children resources
2341 // lower the running processes resources
2344 gtime_fix_ratio = (double)(global_utime + global_stime) / (double)(utime + stime);
2347 cgtime_fix_ratio = 0.0;
2356 cgtime_fix_ratio = 0.0;
2359 if(utime_fix_ratio > 1.0) utime_fix_ratio = 1.0;
2360 if(cutime_fix_ratio > 1.0) cutime_fix_ratio = 1.0;
2361 if(stime_fix_ratio > 1.0) stime_fix_ratio = 1.0;
2362 if(cstime_fix_ratio > 1.0) cstime_fix_ratio = 1.0;
2363 if(gtime_fix_ratio > 1.0) gtime_fix_ratio = 1.0;
2364 if(cgtime_fix_ratio > 1.0) cgtime_fix_ratio = 1.0;
2366 // if(utime_fix_ratio < 0.0) utime_fix_ratio = 0.0;
2367 // if(cutime_fix_ratio < 0.0) cutime_fix_ratio = 0.0;
2368 // if(stime_fix_ratio < 0.0) stime_fix_ratio = 0.0;
2369 // if(cstime_fix_ratio < 0.0) cstime_fix_ratio = 0.0;
2370 // if(gtime_fix_ratio < 0.0) gtime_fix_ratio = 0.0;
2371 // if(cgtime_fix_ratio < 0.0) cgtime_fix_ratio = 0.0;
2374 // we use cpu time to normalize page faults
2375 // the problem is that to find the proper max values
2376 // for page faults we have to parse /proc/vmstat
2377 // which is quite big to do it again (netdata does it already)
2379 // a better solution could be to somehow have netdata
2380 // do this normalization for us
2382 if(utime || stime || gtime)
2384 minflt_fix_ratio = (double)(utime * utime_fix_ratio + stime * stime_fix_ratio + gtime * gtime_fix_ratio) / (double)(utime + stime + gtime);
2387 majflt_fix_ratio = 1.0;
2389 if(cutime || cstime || cgtime)
2391 cminflt_fix_ratio = (double)(cutime * cutime_fix_ratio + cstime * cstime_fix_ratio + cgtime * cgtime_fix_ratio) / (double)(cutime + cstime + cgtime);
2394 cmajflt_fix_ratio = 1.0;
2398 if(unlikely(debug)) {
2400 "SYSTEM: u=%llu s=%llu g=%llu "
2401 "COLLECTED: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2402 "DELTA: u=%lld s=%lld g=%lld "
2403 "FIX: u=%0.2f s=%0.2f g=%0.2f cu=%0.2f cs=%0.2f cg=%0.2f "
2404 "FINALLY: u=%llu s=%llu g=%llu cu=%llu cs=%llu cg=%llu "
2415 , (long long)utime + (long long)cutime - (long long)global_utime
2416 , (long long)stime + (long long)cstime - (long long)global_stime
2417 , (long long)gtime + (long long)cgtime - (long long)global_gtime
2424 , (unsigned long long)(utime * utime_fix_ratio)
2425 , (unsigned long long)(stime * stime_fix_ratio)
2426 , (unsigned long long)(gtime * gtime_fix_ratio)
2427 , (unsigned long long)(cutime * cutime_fix_ratio)
2428 , (unsigned long long)(cstime * cstime_fix_ratio)
2429 , (unsigned long long)(cgtime * cgtime_fix_ratio)
2434 static void send_collected_data_to_netdata(struct target *root, const char *type, usec_t usec) {
2437 send_BEGIN(type, "cpu", usec);
2438 for (w = root; w ; w = w->next) {
2439 if(unlikely(w->exposed))
2440 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));
2444 send_BEGIN(type, "cpu_user", 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) + (include_exited_childs?((unsigned long long)(w->cutime * cutime_fix_ratio)):0ULL));
2451 send_BEGIN(type, "cpu_system", usec);
2452 for (w = root; w ; w = w->next) {
2453 if(unlikely(w->exposed))
2454 send_SET(w->name, (unsigned long long)(w->stime * stime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cstime * cstime_fix_ratio)):0ULL));
2458 if(show_guest_time) {
2459 send_BEGIN(type, "cpu_guest", usec);
2460 for (w = root; w ; w = w->next) {
2461 if(unlikely(w->exposed))
2462 send_SET(w->name, (unsigned long long)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((unsigned long long)(w->cgtime * cgtime_fix_ratio)):0ULL));
2467 send_BEGIN(type, "threads", usec);
2468 for (w = root; w ; w = w->next) {
2469 if(unlikely(w->exposed))
2470 send_SET(w->name, w->num_threads);
2474 send_BEGIN(type, "processes", usec);
2475 for (w = root; w ; w = w->next) {
2476 if(unlikely(w->exposed))
2477 send_SET(w->name, w->processes);
2481 send_BEGIN(type, "mem", usec);
2482 for (w = root; w ; w = w->next) {
2483 if(unlikely(w->exposed))
2484 send_SET(w->name, (w->statm_resident > w->statm_share)?(w->statm_resident - w->statm_share):0ULL);
2488 send_BEGIN(type, "vmem", usec);
2489 for (w = root; w ; w = w->next) {
2490 if(unlikely(w->exposed))
2491 send_SET(w->name, w->statm_size);
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 if(enable_file_charts) {
2538 send_BEGIN(type, "files", usec);
2539 for (w = root; w; w = w->next) {
2540 if (unlikely(w->exposed))
2541 send_SET(w->name, w->openfiles);
2545 send_BEGIN(type, "sockets", usec);
2546 for (w = root; w; w = w->next) {
2547 if (unlikely(w->exposed))
2548 send_SET(w->name, w->opensockets);
2552 send_BEGIN(type, "pipes", usec);
2553 for (w = root; w; w = w->next) {
2554 if (unlikely(w->exposed))
2555 send_SET(w->name, w->openpipes);
2562 // ----------------------------------------------------------------------------
2563 // generate the charts
2565 static void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title)
2568 int newly_added = 0;
2570 for(w = root ; w ; w = w->next) {
2571 if (w->target) continue;
2573 if (!w->exposed && w->processes) {
2576 if (debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
2580 // nothing more to show
2581 if(!newly_added && show_guest_time == show_guest_time_old) return;
2583 // we have something new to show
2584 // update the charts
2585 fprintf(stdout, "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);
2586 for (w = root; w ; w = w->next) {
2587 if(unlikely(w->exposed))
2588 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu %s\n", w->name, hz * RATES_DETAIL / 100, w->hidden ? "hidden" : "");
2591 fprintf(stdout, "CHART %s.mem '' '%s Real Memory (w/o shared)' 'MB' mem %s.mem stacked 20003 %d\n", type, title, type, update_every);
2592 for (w = root; w ; w = w->next) {
2593 if(unlikely(w->exposed))
2594 fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
2597 fprintf(stdout, "CHART %s.vmem '' '%s Virtual Memory Size' 'MB' mem %s.vmem stacked 20004 %d\n", type, title, type, update_every);
2598 for (w = root; w ; w = w->next) {
2599 if(unlikely(w->exposed))
2600 fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
2603 fprintf(stdout, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20005 %d\n", type, title, type, update_every);
2604 for (w = root; w ; w = w->next) {
2605 if(unlikely(w->exposed))
2606 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
2609 fprintf(stdout, "CHART %s.processes '' '%s Processes' 'processes' processes %s.processes stacked 20004 %d\n", type, title, type, update_every);
2610 for (w = root; w ; w = w->next) {
2611 if(unlikely(w->exposed))
2612 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
2615 fprintf(stdout, "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);
2616 for (w = root; w ; w = w->next) {
2617 if(unlikely(w->exposed))
2618 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2621 fprintf(stdout, "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);
2622 for (w = root; w ; w = w->next) {
2623 if(unlikely(w->exposed))
2624 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2627 if(show_guest_time) {
2628 fprintf(stdout, "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);
2629 for (w = root; w; w = w->next) {
2630 if(unlikely(w->exposed))
2631 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, hz * RATES_DETAIL / 100LLU);
2635 fprintf(stdout, "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);
2636 for (w = root; w ; w = w->next) {
2637 if(unlikely(w->exposed))
2638 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
2641 fprintf(stdout, "CHART %s.minor_faults '' '%s Minor Page Faults' 'page faults/s' mem %s.minor_faults stacked 20011 %d\n", type, title, type, update_every);
2642 for (w = root; w ; w = w->next) {
2643 if(unlikely(w->exposed))
2644 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL);
2647 fprintf(stdout, "CHART %s.lreads '' '%s Disk Logical Reads' 'kilobytes/s' disk %s.lreads stacked 20042 %d\n", type, title, type, update_every);
2648 for (w = root; w ; w = w->next) {
2649 if(unlikely(w->exposed))
2650 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2653 fprintf(stdout, "CHART %s.lwrites '' '%s I/O Logical Writes' 'kilobytes/s' disk %s.lwrites stacked 20042 %d\n", type, title, type, update_every);
2654 for (w = root; w ; w = w->next) {
2655 if(unlikely(w->exposed))
2656 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2659 fprintf(stdout, "CHART %s.preads '' '%s Disk Reads' 'kilobytes/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every);
2660 for (w = root; w ; w = w->next) {
2661 if(unlikely(w->exposed))
2662 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2665 fprintf(stdout, "CHART %s.pwrites '' '%s Disk Writes' 'kilobytes/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
2666 for (w = root; w ; w = w->next) {
2667 if(unlikely(w->exposed))
2668 fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL);
2671 if(enable_file_charts) {
2672 fprintf(stdout, "CHART %s.files '' '%s Open Files' 'open files' disk %s.files stacked 20050 %d\n", type,
2673 title, type, update_every);
2674 for (w = root; w; w = w->next) {
2675 if (unlikely(w->exposed))
2676 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
2679 fprintf(stdout, "CHART %s.sockets '' '%s Open Sockets' 'open sockets' net %s.sockets stacked 20051 %d\n",
2680 type, title, type, update_every);
2681 for (w = root; w; w = w->next) {
2682 if (unlikely(w->exposed))
2683 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
2686 fprintf(stdout, "CHART %s.pipes '' '%s Pipes' 'open pipes' processes %s.pipes stacked 20053 %d\n", type,
2687 title, type, update_every);
2688 for (w = root; w; w = w->next) {
2689 if (unlikely(w->exposed))
2690 fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name);
2696 // ----------------------------------------------------------------------------
2697 // parse command line arguments
2699 static void parse_args(int argc, char **argv)
2704 for(i = 1; i < argc; i++) {
2706 int n = (int)str2l(argv[i]);
2713 if(strcmp("version", argv[i]) == 0 || strcmp("-v", argv[i]) == 0) {
2714 printf("apps.plugin %s\n", VERSION);
2718 if(strcmp("debug", argv[i]) == 0) {
2720 // debug_flags = 0xffffffff;
2724 if(strcmp("no-childs", argv[i]) == 0 || strcmp("without-childs", argv[i]) == 0) {
2725 include_exited_childs = 0;
2729 if(strcmp("with-childs", argv[i]) == 0) {
2730 include_exited_childs = 1;
2734 if(strcmp("with-guest", argv[i]) == 0) {
2735 enable_guest_charts = 1;
2739 if(strcmp("no-guest", argv[i]) == 0 || strcmp("without-guest", argv[i]) == 0) {
2740 enable_guest_charts = 0;
2744 if(strcmp("with-files", argv[i]) == 0) {
2745 enable_file_charts = 1;
2749 if(strcmp("no-files", argv[i]) == 0 || strcmp("without-files", argv[i]) == 0) {
2750 enable_file_charts = 0;
2754 if(strcmp("no-users", argv[i]) == 0 || strcmp("without-users", argv[i]) == 0) {
2755 enable_users_charts = 0;
2759 if(strcmp("no-groups", argv[i]) == 0 || strcmp("without-groups", argv[i]) == 0) {
2760 enable_groups_charts = 0;
2764 if(strcmp("-h", argv[i]) == 0 || strcmp("--help", argv[i]) == 0) {
2767 " netdata apps.plugin %s\n"
2768 " Copyright (C) 2016-2017 Costa Tsaousis <costa@tsaousis.gr>\n"
2769 " Released under GNU Public License v3 or later.\n"
2770 " All rights reserved.\n"
2772 " This program is a data collector plugin for netdata.\n"
2774 " Valid command line options:\n"
2776 " SECONDS set the data collection frequency\n"
2778 " debug enable debugging (lot of output)\n"
2781 " without-childs enable / disable aggregating exited\n"
2782 " children resources into parents\n"
2783 " (default is enabled)\n"
2786 " without-guest enable / disable reporting guest charts\n"
2787 " (default is disabled)\n"
2790 " without-files enable / disable reporting files, sockets, pipes\n"
2791 " (default is enabled)\n"
2793 " NAME read apps_NAME.conf instead of\n"
2794 " apps_groups.conf\n"
2795 " (default NAME=groups)\n"
2797 " version print program version and exit\n"
2809 error("Cannot understand option %s", argv[i]);
2813 if(freq > 0) update_every = freq;
2814 if(!name) name = "groups";
2816 if(read_apps_groups_conf(name)) {
2817 error("Cannot read process groups '%s/apps_%s.conf'. There are no internal defaults. Failing.", config_dir, name);
2822 int main(int argc, char **argv)
2824 // debug_flags = D_PROCFILE;
2826 // set the name for logging
2827 program_name = "apps.plugin";
2829 info("started on pid %d", getpid());
2831 // disable syslog for apps.plugin
2832 error_log_syslog = 0;
2834 // set errors flood protection to 100 logs per hour
2835 error_log_errors_per_period = 100;
2836 error_log_throttle_period = 3600;
2838 global_host_prefix = getenv("NETDATA_HOST_PREFIX");
2839 if(global_host_prefix == NULL) {
2840 // info("NETDATA_HOST_PREFIX is not passed from netdata");
2841 global_host_prefix = "";
2843 // else info("Found NETDATA_HOST_PREFIX='%s'", global_host_prefix);
2845 config_dir = getenv("NETDATA_CONFIG_DIR");
2846 if(config_dir == NULL) {
2847 // info("NETDATA_CONFIG_DIR is not passed from netdata");
2848 config_dir = CONFIG_DIR;
2850 // else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
2852 #ifdef NETDATA_INTERNAL_CHECKS
2853 if(debug_flags != 0) {
2854 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
2855 if(setrlimit(RLIMIT_CORE, &rl) != 0)
2856 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
2857 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
2859 #endif /* NETDATA_INTERNAL_CHECKS */
2861 procfile_adaptive_initial_allocation = 1;
2863 time_t started_t = now_monotonic_sec();
2865 get_system_pid_max();
2868 parse_args(argc, argv);
2870 all_pids_sortlist = callocz(sizeof(pid_t), (size_t)pid_max);
2871 all_pids = callocz(sizeof(struct pid_stat *), (size_t) pid_max);
2874 "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %1$d\n"
2875 "DIMENSION user '' incremental 1 1000\n"
2876 "DIMENSION system '' incremental 1 1000\n"
2877 "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %1$d\n"
2878 "DIMENSION files '' incremental 1 1\n"
2879 "DIMENSION pids '' absolute 1 1\n"
2880 "DIMENSION fds '' absolute 1 1\n"
2881 "DIMENSION targets '' absolute 1 1\n"
2882 "CHART netdata.apps_fix '' 'Apps Plugin Normalization Ratios' 'percentage' apps.plugin netdata.apps_fix line 140002 %1$d\n"
2883 "DIMENSION utime '' absolute 1 %2$llu\n"
2884 "DIMENSION stime '' absolute 1 %2$llu\n"
2885 "DIMENSION gtime '' absolute 1 %2$llu\n"
2886 "DIMENSION minflt '' absolute 1 %2$llu\n"
2887 "DIMENSION majflt '' absolute 1 %2$llu\n"
2892 if(include_exited_childs)
2894 "CHART netdata.apps_children_fix '' 'Apps Plugin Exited Children Normalization Ratios' 'percentage' apps.plugin netdata.apps_children_fix line 140003 %1$d\n"
2895 "DIMENSION cutime '' absolute 1 %2$llu\n"
2896 "DIMENSION cstime '' absolute 1 %2$llu\n"
2897 "DIMENSION cgtime '' absolute 1 %2$llu\n"
2898 "DIMENSION cminflt '' absolute 1 %2$llu\n"
2899 "DIMENSION cmajflt '' absolute 1 %2$llu\n"
2904 usec_t step = update_every * USEC_PER_SEC;
2905 global_iterations_counter = 1;
2907 heartbeat_init(&hb);
2908 for(;1; global_iterations_counter++) {
2910 #ifdef NETDATA_PROFILING
2911 #warning "compiling for profiling"
2912 static int profiling_count=0;
2914 if(unlikely(profiling_count > 1000)) exit(0);
2916 heartbeat_next(&hb, step);
2919 if(!collect_data_for_all_processes_from_proc()) {
2920 error("Cannot collect /proc data for running processes. Disabling apps.plugin...");
2921 printf("DISABLE\n");
2925 calculate_netdata_statistics();
2926 normalize_data(apps_groups_root_target);
2928 usec_t dt = send_resource_usage_to_netdata();
2930 // this is smart enough to show only newly added apps, when needed
2931 send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps");
2933 if(likely(enable_users_charts))
2934 send_charts_updates_to_netdata(users_root_target, "users", "Users");
2936 if(likely(enable_groups_charts))
2937 send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups");
2939 send_collected_data_to_netdata(apps_groups_root_target, "apps", dt);
2941 if(likely(enable_users_charts))
2942 send_collected_data_to_netdata(users_root_target, "users", dt);
2944 if(likely(enable_groups_charts))
2945 send_collected_data_to_netdata(groups_root_target, "groups", dt);
2949 show_guest_time_old = show_guest_time;
2952 fprintf(stderr, "apps.plugin: done Loop No %llu\n", global_iterations_counter);
2954 // restart check (14400 seconds)
2955 if(now_monotonic_sec() - started_t > 14400) exit(0);