3 // 1. disable RESET_OR_OVERFLOW check in charts
13 #include <sys/types.h>
18 #include <sys/resource.h>
29 #include <arpa/inet.h>
31 #include <sys/types.h>
40 #include "../config.h"
42 #ifdef NETDATA_INTERNAL_CHECKS
43 #include <sys/prctl.h>
46 #define MAX_COMPARE_NAME 100
48 #define MAX_CMDLINE 1024
51 pid_t pid_max = 32768;
55 unsigned long long file_counter = 0;
56 int proc_pid_cmdline_is_needed = 0;
57 int include_exited_childs = 1;
58 char *host_prefix = "";
59 char *config_dir = CONFIG_DIR;
62 // ----------------------------------------------------------------------------
64 void netdata_cleanup_and_exit(int ret) {
69 // ----------------------------------------------------------------------------
71 // to retrieve settings of the system
73 long get_system_cpus(void) {
78 char filename[FILENAME_MAX + 1];
79 snprintfz(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
81 ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
84 ff = procfile_readall(ff);
91 for(i = 0; i < procfile_lines(ff); i++) {
92 if(!procfile_linewords(ff, i)) continue;
94 if(strncmp(procfile_lineword(ff, i, 0), "cpu", 3) == 0) processors++;
97 if(processors < 1) processors = 1;
103 pid_t get_system_pid_max(void) {
107 char filename[FILENAME_MAX + 1];
108 snprintfz(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
109 ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
112 ff = procfile_readall(ff);
118 mpid = (pid_t)atoi(procfile_lineword(ff, 0, 0));
119 if(!mpid) mpid = 32768;
125 // ----------------------------------------------------------------------------
127 // target is the structure that process data are aggregated
130 char compare[MAX_COMPARE_NAME + 1];
131 uint32_t comparehash;
134 char id[MAX_NAME + 1];
137 char name[MAX_NAME + 1];
142 unsigned long long minflt;
143 unsigned long long cminflt;
144 unsigned long long majflt;
145 unsigned long long cmajflt;
146 unsigned long long utime;
147 unsigned long long stime;
148 unsigned long long cutime;
149 unsigned long long cstime;
150 unsigned long long num_threads;
151 unsigned long long rss;
153 long long fix_minflt;
154 long long fix_cminflt;
155 long long fix_majflt;
156 long long fix_cmajflt;
159 long long fix_cutime;
160 long long fix_cstime;
162 unsigned long long statm_size;
163 unsigned long long statm_resident;
164 unsigned long long statm_share;
165 unsigned long long statm_text;
166 unsigned long long statm_lib;
167 unsigned long long statm_data;
168 unsigned long long statm_dirty;
170 unsigned long long io_logical_bytes_read;
171 unsigned long long io_logical_bytes_written;
172 unsigned long long io_read_calls;
173 unsigned long long io_write_calls;
174 unsigned long long io_storage_bytes_read;
175 unsigned long long io_storage_bytes_written;
176 unsigned long long io_cancelled_write_bytes;
178 unsigned long long fix_io_logical_bytes_read;
179 unsigned long long fix_io_logical_bytes_written;
180 unsigned long long fix_io_read_calls;
181 unsigned long long fix_io_write_calls;
182 unsigned long long fix_io_storage_bytes_read;
183 unsigned long long fix_io_storage_bytes_written;
184 unsigned long long fix_io_cancelled_write_bytes;
187 unsigned long long openfiles;
188 unsigned long long openpipes;
189 unsigned long long opensockets;
190 unsigned long long openinotifies;
191 unsigned long long openeventfds;
192 unsigned long long opentimerfds;
193 unsigned long long opensignalfds;
194 unsigned long long openeventpolls;
195 unsigned long long openother;
197 unsigned long processes; // how many processes have been merged to this
198 int exposed; // if set, we have sent this to netdata
199 int hidden; // if set, we set the hidden flag on the dimension
202 int starts_with; // if set, the compare string matches only the
203 // beginning of the command
205 struct target *target; // the one that will be reported to netdata
210 // ----------------------------------------------------------------------------
212 // aggregate all processes in groups, to have a limited number of dimensions
214 struct target *apps_groups_root_target = NULL;
215 struct target *apps_groups_default_target = NULL;
216 long apps_groups_targets = 0;
218 struct target *users_root_target = NULL;
219 struct target *groups_root_target = NULL;
221 struct target *get_users_target(uid_t uid)
224 for(w = users_root_target ; w ; w = w->next)
225 if(w->uid == uid) return w;
227 w = calloc(sizeof(struct target), 1);
229 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
233 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", uid);
234 w->comparehash = simple_hash(w->compare);
235 w->comparelen = strlen(w->compare);
237 snprintfz(w->id, MAX_NAME, "%u", uid);
238 w->idhash = simple_hash(w->id);
240 struct passwd *pw = getpwuid(uid);
242 snprintfz(w->name, MAX_NAME, "%u", uid);
244 snprintfz(w->name, MAX_NAME, "%s", pw->pw_name);
246 netdata_fix_chart_name(w->name);
250 w->next = users_root_target;
251 users_root_target = w;
254 fprintf(stderr, "apps.plugin: added uid %u ('%s') target\n", w->uid, w->name);
259 struct target *get_groups_target(gid_t gid)
262 for(w = groups_root_target ; w ; w = w->next)
263 if(w->gid == gid) return w;
265 w = calloc(sizeof(struct target), 1);
267 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
271 snprintfz(w->compare, MAX_COMPARE_NAME, "%u", gid);
272 w->comparehash = simple_hash(w->compare);
273 w->comparelen = strlen(w->compare);
275 snprintfz(w->id, MAX_NAME, "%u", gid);
276 w->idhash = simple_hash(w->id);
278 struct group *gr = getgrgid(gid);
280 snprintfz(w->name, MAX_NAME, "%u", gid);
282 snprintfz(w->name, MAX_NAME, "%s", gr->gr_name);
284 netdata_fix_chart_name(w->name);
288 w->next = groups_root_target;
289 groups_root_target = w;
292 fprintf(stderr, "apps.plugin: added gid %u ('%s') target\n", w->gid, w->name);
297 // find or create a new target
298 // there are targets that are just aggregated to other target (the second argument)
299 struct target *get_apps_groups_target(const char *id, struct target *target)
301 int tdebug = 0, thidden = 0, ends_with = 0;
302 const char *nid = id;
304 while(nid[0] == '-' || nid[0] == '+' || nid[0] == '*') {
305 if(nid[0] == '-') thidden = 1;
306 if(nid[0] == '+') tdebug = 1;
307 if(nid[0] == '*') ends_with = 1;
310 uint32_t hash = simple_hash(id);
313 for(w = apps_groups_root_target ; w ; w = w->next) {
314 if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
318 w = calloc(sizeof(struct target), 1);
320 error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct target));
324 strncpyz(w->id, nid, MAX_NAME);
325 w->idhash = simple_hash(w->id);
327 strncpyz(w->name, nid, MAX_NAME);
329 strncpyz(w->compare, nid, MAX_COMPARE_NAME);
330 int len = strlen(w->compare);
331 if(w->compare[len - 1] == '*') {
332 w->compare[len - 1] = '\0';
335 w->ends_with = ends_with;
337 if(w->starts_with && w->ends_with)
338 proc_pid_cmdline_is_needed = 1;
340 w->comparehash = simple_hash(w->compare);
341 w->comparelen = strlen(w->compare);
347 w->next = apps_groups_root_target;
348 apps_groups_root_target = w;
351 fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
353 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
354 , w->target?w->target->id:w->id
355 , (w->hidden)?"hidden":"-"
356 , (w->debug)?"debug":"-"
362 // read the apps_groups.conf file
363 int read_apps_groups_conf(const char *name)
365 char filename[FILENAME_MAX + 1];
367 snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
370 fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
372 // ----------------------------------------
374 procfile *ff = procfile_open(filename, " :\t", PROCFILE_FLAG_DEFAULT);
377 procfile_set_quotes(ff, "'\"");
379 ff = procfile_readall(ff);
385 unsigned long line, lines = procfile_lines(ff);
387 for(line = 0; line < lines ;line++) {
388 unsigned long word, words = procfile_linewords(ff, line);
389 struct target *w = NULL;
391 char *t = procfile_lineword(ff, line, 0);
392 if(!t || !*t) continue;
394 for(word = 0; word < words ;word++) {
395 char *s = procfile_lineword(ff, line, word);
396 if(!s || !*s) continue;
401 struct target *n = get_apps_groups_target(s, w);
403 error("Cannot create target '%s' (line %lu, word %lu)", s, line, word);
411 int tdebug = 0, thidden = 0;
413 while(t[0] == '-' || t[0] == '+') {
414 if(t[0] == '-') thidden = 1;
415 if(t[0] == '+') tdebug = 1;
419 strncpyz(w->name, t, MAX_NAME);
424 fprintf(stderr, "apps.plugin: AGGREGATION TARGET NAME '%s' on ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
427 , w->compare, (w->starts_with && w->ends_with)?"substring":((w->starts_with)?"prefix":((w->ends_with)?"suffix":"exact"))
428 , w->target?w->target->id:w->id
429 , (w->hidden)?"hidden":"-"
430 , (w->debug)?"debug":"-"
437 apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL); // match nothing
438 if(!apps_groups_default_target)
439 error("Cannot create default target");
441 strncpyz(apps_groups_default_target->name, "other", MAX_NAME);
447 // ----------------------------------------------------------------------------
448 // data to store for each pid
453 char comm[MAX_COMPARE_NAME + 1];
454 char cmdline[MAX_CMDLINE + 1];
463 unsigned long long minflt;
464 unsigned long long cminflt;
465 unsigned long long majflt;
466 unsigned long long cmajflt;
467 unsigned long long utime;
468 unsigned long long stime;
469 unsigned long long cutime;
470 unsigned long long cstime;
474 // int64_t itrealvalue;
475 // unsigned long long starttime;
476 // unsigned long long vsize;
477 unsigned long long rss;
478 // unsigned long long rsslim;
479 // unsigned long long starcode;
480 // unsigned long long endcode;
481 // unsigned long long startstack;
482 // unsigned long long kstkesp;
483 // unsigned long long kstkeip;
486 // uint64_t sigignore;
487 // uint64_t sigcatch;
491 // int32_t exit_signal;
492 // int32_t processor;
493 // uint32_t rt_priority;
495 // unsigned long long delayacct_blkio_ticks;
496 // uint64_t guest_time;
497 // int64_t cguest_time;
502 unsigned long long statm_size;
503 unsigned long long statm_resident;
504 unsigned long long statm_share;
505 unsigned long long statm_text;
506 unsigned long long statm_lib;
507 unsigned long long statm_data;
508 unsigned long long statm_dirty;
510 unsigned long long io_logical_bytes_read;
511 unsigned long long io_logical_bytes_written;
512 unsigned long long io_read_calls;
513 unsigned long long io_write_calls;
514 unsigned long long io_storage_bytes_read;
515 unsigned long long io_storage_bytes_written;
516 unsigned long long io_cancelled_write_bytes;
518 // we need the last values
519 // for all incremental counters
520 // so that when a process switches users/groups
521 // we will subtract these values from the old
523 unsigned long long last_minflt;
524 unsigned long long last_majflt;
525 unsigned long long last_utime;
526 unsigned long long last_stime;
528 unsigned long long last_cminflt;
529 unsigned long long last_cmajflt;
530 unsigned long long last_cutime;
531 unsigned long long last_cstime;
533 unsigned long long last_fix_cminflt;
534 unsigned long long last_fix_cmajflt;
535 unsigned long long last_fix_cutime;
536 unsigned long long last_fix_cstime;
538 unsigned long long last_io_logical_bytes_read;
539 unsigned long long last_io_logical_bytes_written;
540 unsigned long long last_io_read_calls;
541 unsigned long long last_io_write_calls;
542 unsigned long long last_io_storage_bytes_read;
543 unsigned long long last_io_storage_bytes_written;
544 unsigned long long last_io_cancelled_write_bytes;
546 unsigned long long fix_cminflt;
547 unsigned long long fix_cmajflt;
548 unsigned long long fix_cutime;
549 unsigned long long fix_cstime;
551 int *fds; // array of fds it uses
552 int fds_size; // the size of the fds array
554 int children_count; // number of processes directly referencing this
555 int updated; // 1 when update
556 int merged; // 1 when it has been merged to its parent
559 struct target *target; // app_groups.conf targets
560 struct target *user_target; // uid based targets
561 struct target *group_target; // gid based targets
563 struct pid_stat *parent;
564 struct pid_stat *prev;
565 struct pid_stat *next;
567 } *root_of_pids = NULL, **all_pids;
569 long all_pids_count = 0;
571 struct pid_stat *get_pid_entry(pid_t pid)
574 all_pids[pid]->new_entry = 0;
575 return all_pids[pid];
578 all_pids[pid] = calloc(sizeof(struct pid_stat), 1);
580 error("Cannot allocate %zu bytes of memory", (size_t)sizeof(struct pid_stat));
584 all_pids[pid]->fds = calloc(sizeof(int), 100);
585 if(!all_pids[pid]->fds)
586 error("Cannot allocate %zu bytes of memory", (size_t)(sizeof(int) * 100));
587 else all_pids[pid]->fds_size = 100;
589 if(root_of_pids) root_of_pids->prev = all_pids[pid];
590 all_pids[pid]->next = root_of_pids;
591 root_of_pids = all_pids[pid];
593 all_pids[pid]->pid = pid;
594 all_pids[pid]->new_entry = 1;
596 return all_pids[pid];
599 void del_pid_entry(pid_t pid)
601 if(!all_pids[pid]) return;
604 fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
606 if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
607 if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
608 if(all_pids[pid]->prev) all_pids[pid]->prev->next = all_pids[pid]->next;
610 if(all_pids[pid]->fds) free(all_pids[pid]->fds);
612 all_pids[pid] = NULL;
616 // ----------------------------------------------------------------------------
617 // update pids from proc
619 int read_proc_pid_cmdline(struct pid_stat *p) {
620 char filename[FILENAME_MAX + 1];
621 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", host_prefix, p->pid);
623 int fd = open(filename, O_RDONLY, 0666);
624 if(unlikely(fd == -1)) return 1;
626 int i, bytes = read(fd, p->cmdline, MAX_CMDLINE);
630 // copy the command to the command line
631 strncpyz(p->cmdline, p->comm, MAX_CMDLINE);
635 p->cmdline[bytes] = '\0';
636 for(i = 0; i < bytes ; i++)
637 if(!p->cmdline[i]) p->cmdline[i] = ' ';
640 fprintf(stderr, "Read file '%s' contents: %s\n", filename, p->cmdline);
645 int read_proc_pid_ownership(struct pid_stat *p) {
646 char filename[FILENAME_MAX + 1];
648 snprintfz(filename, FILENAME_MAX, "%s/proc/%d", host_prefix, p->pid);
650 // ----------------------------------------
654 if(stat(filename, &st) != 0)
663 int read_proc_pid_stat(struct pid_stat *p) {
664 static procfile *ff = NULL;
666 char filename[FILENAME_MAX + 1];
668 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
670 // ----------------------------------------
672 int set_quotes = (!ff)?1:0;
674 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
677 // if(set_quotes) procfile_set_quotes(ff, "()");
678 if(set_quotes) procfile_set_open_close(ff, "(", ")");
680 ff = procfile_readall(ff);
682 // procfile_close(ff);
688 // parse the process name
690 strncpyz(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
692 // p->pid = atol(procfile_lineword(ff, 0, 0+i));
694 // p->state = *(procfile_lineword(ff, 0, 2+i));
695 p->ppid = (int32_t) atol(procfile_lineword(ff, 0, 3 + i));
696 // p->pgrp = atol(procfile_lineword(ff, 0, 4+i));
697 // p->session = atol(procfile_lineword(ff, 0, 5+i));
698 // p->tty_nr = atol(procfile_lineword(ff, 0, 6+i));
699 // p->tpgid = atol(procfile_lineword(ff, 0, 7+i));
700 // p->flags = strtoull(procfile_lineword(ff, 0, 8+i), NULL, 10);
701 p->minflt = strtoull(procfile_lineword(ff, 0, 9+i), NULL, 10);
702 p->cminflt = strtoull(procfile_lineword(ff, 0, 10+i), NULL, 10);
703 p->majflt = strtoull(procfile_lineword(ff, 0, 11+i), NULL, 10);
704 p->cmajflt = strtoull(procfile_lineword(ff, 0, 12+i), NULL, 10);
705 p->utime = strtoull(procfile_lineword(ff, 0, 13+i), NULL, 10);
706 p->stime = strtoull(procfile_lineword(ff, 0, 14+i), NULL, 10);
707 p->cutime = strtoull(procfile_lineword(ff, 0, 15+i), NULL, 10);
708 p->cstime = strtoull(procfile_lineword(ff, 0, 16+i), NULL, 10);
709 // p->priority = strtoull(procfile_lineword(ff, 0, 17+i), NULL, 10);
710 // p->nice = strtoull(procfile_lineword(ff, 0, 18+i), NULL, 10);
711 p->num_threads = (int32_t) atol(procfile_lineword(ff, 0, 19 + i));
712 // p->itrealvalue = strtoull(procfile_lineword(ff, 0, 20+i), NULL, 10);
713 // p->starttime = strtoull(procfile_lineword(ff, 0, 21+i), NULL, 10);
714 // p->vsize = strtoull(procfile_lineword(ff, 0, 22+i), NULL, 10);
715 p->rss = strtoull(procfile_lineword(ff, 0, 23+i), NULL, 10);
716 // p->rsslim = strtoull(procfile_lineword(ff, 0, 24+i), NULL, 10);
717 // p->starcode = strtoull(procfile_lineword(ff, 0, 25+i), NULL, 10);
718 // p->endcode = strtoull(procfile_lineword(ff, 0, 26+i), NULL, 10);
719 // p->startstack = strtoull(procfile_lineword(ff, 0, 27+i), NULL, 10);
720 // p->kstkesp = strtoull(procfile_lineword(ff, 0, 28+i), NULL, 10);
721 // p->kstkeip = strtoull(procfile_lineword(ff, 0, 29+i), NULL, 10);
722 // p->signal = strtoull(procfile_lineword(ff, 0, 30+i), NULL, 10);
723 // p->blocked = strtoull(procfile_lineword(ff, 0, 31+i), NULL, 10);
724 // p->sigignore = strtoull(procfile_lineword(ff, 0, 32+i), NULL, 10);
725 // p->sigcatch = strtoull(procfile_lineword(ff, 0, 33+i), NULL, 10);
726 // p->wchan = strtoull(procfile_lineword(ff, 0, 34+i), NULL, 10);
727 // p->nswap = strtoull(procfile_lineword(ff, 0, 35+i), NULL, 10);
728 // p->cnswap = strtoull(procfile_lineword(ff, 0, 36+i), NULL, 10);
729 // p->exit_signal = atol(procfile_lineword(ff, 0, 37+i));
730 // p->processor = atol(procfile_lineword(ff, 0, 38+i));
731 // p->rt_priority = strtoul(procfile_lineword(ff, 0, 39+i), NULL, 10);
732 // p->policy = strtoul(procfile_lineword(ff, 0, 40+i), NULL, 10);
733 // p->delayacct_blkio_ticks = strtoull(procfile_lineword(ff, 0, 41+i), NULL, 10);
734 // p->guest_time = strtoull(procfile_lineword(ff, 0, 42+i), NULL, 10);
735 // p->cguest_time = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
737 if(unlikely(debug || (p->target && p->target->debug)))
738 fprintf(stderr, "apps.plugin: READ PROC/PID/STAT: %s/proc/%d/stat, process: '%s' 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->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt, p->num_threads);
740 // procfile_close(ff);
744 int read_proc_pid_statm(struct pid_stat *p) {
745 static procfile *ff = NULL;
747 char filename[FILENAME_MAX + 1];
749 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
751 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
754 ff = procfile_readall(ff);
756 // procfile_close(ff);
762 p->statm_size = strtoull(procfile_lineword(ff, 0, 0), NULL, 10);
763 p->statm_resident = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
764 p->statm_share = strtoull(procfile_lineword(ff, 0, 2), NULL, 10);
765 p->statm_text = strtoull(procfile_lineword(ff, 0, 3), NULL, 10);
766 p->statm_lib = strtoull(procfile_lineword(ff, 0, 4), NULL, 10);
767 p->statm_data = strtoull(procfile_lineword(ff, 0, 5), NULL, 10);
768 p->statm_dirty = strtoull(procfile_lineword(ff, 0, 6), NULL, 10);
770 // procfile_close(ff);
774 int read_proc_pid_io(struct pid_stat *p) {
775 static procfile *ff = NULL;
777 char filename[FILENAME_MAX + 1];
779 snprintfz(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
781 ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
784 ff = procfile_readall(ff);
786 // procfile_close(ff);
792 p->io_logical_bytes_read = strtoull(procfile_lineword(ff, 0, 1), NULL, 10);
793 p->io_logical_bytes_written = strtoull(procfile_lineword(ff, 1, 1), NULL, 10);
794 p->io_read_calls = strtoull(procfile_lineword(ff, 2, 1), NULL, 10);
795 p->io_write_calls = strtoull(procfile_lineword(ff, 3, 1), NULL, 10);
796 p->io_storage_bytes_read = strtoull(procfile_lineword(ff, 4, 1), NULL, 10);
797 p->io_storage_bytes_written = strtoull(procfile_lineword(ff, 5, 1), NULL, 10);
798 p->io_cancelled_write_bytes = strtoull(procfile_lineword(ff, 6, 1), NULL, 10);
800 // procfile_close(ff);
805 // ----------------------------------------------------------------------------
807 // this is used to keep a global list of all open files of the system
808 // it is needed in order to calculate the unique files processes have open
810 #define FILE_DESCRIPTORS_INCREASE_STEP 100
812 struct file_descriptor {
814 #ifdef NETDATA_INTERNAL_CHECKS
816 #endif /* NETDATA_INTERNAL_CHECKS */
824 int all_files_len = 0;
825 int all_files_size = 0;
827 int file_descriptor_compare(void* a, void* b) {
828 #ifdef NETDATA_INTERNAL_CHECKS
829 if(((struct file_descriptor *)a)->magic != 0x0BADCAFE || ((struct file_descriptor *)b)->magic != 0x0BADCAFE)
830 error("Corrupted index data detected. Please report this.");
831 #endif /* NETDATA_INTERNAL_CHECKS */
833 if(((struct file_descriptor *)a)->hash < ((struct file_descriptor *)b)->hash)
836 else if(((struct file_descriptor *)a)->hash > ((struct file_descriptor *)b)->hash)
840 return strcmp(((struct file_descriptor *)a)->name, ((struct file_descriptor *)b)->name);
843 int file_descriptor_iterator(avl *a) { if(a) {}; return 0; }
845 avl_tree all_files_index = {
847 file_descriptor_compare
850 static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
851 struct file_descriptor tmp;
852 tmp.hash = (hash)?hash:simple_hash(name);
856 #ifdef NETDATA_INTERNAL_CHECKS
857 tmp.magic = 0x0BADCAFE;
858 #endif /* NETDATA_INTERNAL_CHECKS */
860 return (struct file_descriptor *)avl_search(&all_files_index, (avl *) &tmp);
863 #define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
864 #define file_descriptor_remove(fd) avl_remove(&all_files_index, (avl *)(fd))
866 #define FILETYPE_OTHER 0
867 #define FILETYPE_FILE 1
868 #define FILETYPE_PIPE 2
869 #define FILETYPE_SOCKET 3
870 #define FILETYPE_INOTIFY 4
871 #define FILETYPE_EVENTFD 5
872 #define FILETYPE_EVENTPOLL 6
873 #define FILETYPE_TIMERFD 7
874 #define FILETYPE_SIGNALFD 8
876 void file_descriptor_not_used(int id)
878 if(id > 0 && id < all_files_size) {
880 #ifdef NETDATA_INTERNAL_CHECKS
881 if(all_files[id].magic != 0x0BADCAFE) {
882 error("Ignoring request to remove empty file id %d.", id);
885 #endif /* NETDATA_INTERNAL_CHECKS */
888 fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
890 if(all_files[id].count > 0) {
891 all_files[id].count--;
893 if(!all_files[id].count) {
895 fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
897 file_descriptor_remove(&all_files[id]);
898 #ifdef NETDATA_INTERNAL_CHECKS
899 all_files[id].magic = 0x00000000;
900 #endif /* NETDATA_INTERNAL_CHECKS */
905 error("Request to decrease counter of fd %d (%s), while the use counter is 0", id, all_files[id].name);
907 else error("Request to decrease counter of fd %d, which is outside the array size (1 to %d)", id, all_files_size);
910 int file_descriptor_find_or_add(const char *name)
912 static int last_pos = 0;
913 uint32_t hash = simple_hash(name);
916 fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
918 struct file_descriptor *fd = file_descriptor_find(name, hash);
922 fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
929 // check we have enough memory to add it
930 if(!all_files || all_files_len == all_files_size) {
931 void *old = all_files;
934 // there is no empty slot
936 fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
938 all_files = realloc(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
940 // if the address changed, we have to rebuild the index
941 // since all pointers are now invalid
942 if(old && old != (void *)all_files) {
944 fprintf(stderr, "apps.plugin: >> re-indexing.\n");
946 all_files_index.root = NULL;
947 for(i = 0; i < all_files_size; i++) {
948 if(!all_files[i].count) continue;
949 file_descriptor_add(&all_files[i]);
953 fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
956 for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
957 all_files[i].count = 0;
958 all_files[i].name = NULL;
959 #ifdef NETDATA_INTERNAL_CHECKS
960 all_files[i].magic = 0x00000000;
961 #endif /* NETDATA_INTERNAL_CHECKS */
962 all_files[i].pos = i;
965 if(!all_files_size) all_files_len = 1;
966 all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
970 fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
972 // search for an empty slot
974 for(i = 0, c = last_pos ; i < all_files_size ; i++, c++) {
975 if(c >= all_files_size) c = 0;
978 if(!all_files[c].count) {
980 fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
982 #ifdef NETDATA_INTERNAL_CHECKS
983 if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
984 error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
985 #endif /* NETDATA_INTERNAL_CHECKS */
988 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);
990 if(all_files[c].name) free((void *)all_files[c].name);
991 all_files[c].name = NULL;
996 if(i == all_files_size) {
997 fatal("We should find an empty slot, but there isn't any");
1002 fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
1006 // else we have an empty slot in 'c'
1009 if(name[0] == '/') type = FILETYPE_FILE;
1010 else if(strncmp(name, "pipe:", 5) == 0) type = FILETYPE_PIPE;
1011 else if(strncmp(name, "socket:", 7) == 0) type = FILETYPE_SOCKET;
1012 else if(strcmp(name, "anon_inode:inotify") == 0 || strcmp(name, "inotify") == 0) type = FILETYPE_INOTIFY;
1013 else if(strcmp(name, "anon_inode:[eventfd]") == 0) type = FILETYPE_EVENTFD;
1014 else if(strcmp(name, "anon_inode:[eventpoll]") == 0) type = FILETYPE_EVENTPOLL;
1015 else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
1016 else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
1017 else if(strncmp(name, "anon_inode:", 11) == 0) {
1019 fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
1021 type = FILETYPE_OTHER;
1025 fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
1027 type = FILETYPE_OTHER;
1030 all_files[c].name = strdup(name);
1031 all_files[c].hash = hash;
1032 all_files[c].type = type;
1033 all_files[c].pos = c;
1034 all_files[c].count = 1;
1035 #ifdef NETDATA_INTERNAL_CHECKS
1036 all_files[c].magic = 0x0BADCAFE;
1037 #endif /* NETDATA_INTERNAL_CHECKS */
1038 file_descriptor_add(&all_files[c]);
1041 fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
1046 int read_pid_file_descriptors(struct pid_stat *p) {
1047 char dirname[FILENAME_MAX+1];
1049 snprintfz(dirname, FILENAME_MAX, "%s/proc/%d/fd", host_prefix, p->pid);
1050 DIR *fds = opendir(dirname);
1054 char fdname[FILENAME_MAX + 1];
1055 char linkname[FILENAME_MAX + 1];
1057 // make the array negative
1058 for(c = 0 ; c < p->fds_size ; c++)
1059 p->fds[c] = -p->fds[c];
1061 while((de = readdir(fds))) {
1062 if(strcmp(de->d_name, ".") == 0 || strcmp(de->d_name, "..") == 0)
1065 // check if the fds array is small
1066 int fdid = atoi(de->d_name);
1067 if(fdid < 0) continue;
1068 if(fdid >= p->fds_size) {
1069 // it is small, extend it
1071 fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
1073 p->fds = realloc(p->fds, (fdid + 100) * sizeof(int));
1075 fatal("Cannot re-allocate fds for %s", p->comm);
1079 // and initialize it
1080 for(c = p->fds_size ; c < (fdid + 100) ; c++) p->fds[c] = 0;
1081 p->fds_size = fdid + 100;
1084 if(p->fds[fdid] == 0) {
1085 // we don't know this fd, get it
1087 sprintf(fdname, "%s/proc/%d/fd/%s", host_prefix, p->pid, de->d_name);
1088 ssize_t l = readlink(fdname, linkname, FILENAME_MAX);
1090 if(debug || (p->target && p->target->debug)) {
1091 if(debug || (p->target && p->target->debug))
1092 error("Cannot read link %s", fdname);
1099 // if another process already has this, we will get
1101 p->fds[fdid] = file_descriptor_find_or_add(linkname);
1104 // else make it positive again, we need it
1105 // of course, the actual file may have changed, but we don't care so much
1106 // FIXME: we could compare the inode as returned by readdir direct structure
1107 else p->fds[fdid] = -p->fds[fdid];
1111 // remove all the negative file descriptors
1112 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] < 0) {
1113 file_descriptor_not_used(-p->fds[c]);
1122 // ----------------------------------------------------------------------------
1124 // 1. read all files in /proc
1125 // 2. for each numeric directory:
1126 // i. read /proc/pid/stat
1127 // ii. read /proc/pid/statm
1128 // iii. read /proc/pid/io (requires root access)
1129 // iii. read the entries in directory /proc/pid/fd (requires root access)
1131 // a. find or create a struct file_descriptor
1132 // b. cleanup any old/unused file_descriptors
1134 // after all these, some pids may be linked to targets, while others may not
1136 // in case of errors, only 1 every 1000 errors is printed
1137 // to avoid filling up all disk space
1138 // if debug is enabled, all errors are printed
1140 int collect_data_for_all_processes_from_proc(void)
1142 char dirname[FILENAME_MAX + 1];
1144 snprintfz(dirname, FILENAME_MAX, "%s/proc", host_prefix);
1145 DIR *dir = opendir(dirname);
1148 struct dirent *file = NULL;
1149 struct pid_stat *p = NULL;
1151 // mark them all as un-updated
1153 for(p = root_of_pids; p ; p = p->next) {
1157 p->children_count = 0;
1161 p->last_minflt = p->minflt;
1162 p->last_majflt = p->majflt;
1163 p->last_utime = p->utime;
1164 p->last_stime = p->stime;
1166 p->last_cminflt = p->cminflt;
1167 p->last_cmajflt = p->cmajflt;
1168 p->last_cutime = p->cutime;
1169 p->last_cstime = p->cstime;
1171 p->last_fix_cminflt = p->fix_cminflt;
1172 p->last_fix_cmajflt = p->fix_cmajflt;
1173 p->last_fix_cutime = p->fix_cutime;
1174 p->last_fix_cstime = p->fix_cstime;
1176 p->last_io_logical_bytes_read = p->io_logical_bytes_read;
1177 p->last_io_logical_bytes_written = p->io_logical_bytes_written;
1178 p->last_io_read_calls = p->io_read_calls;
1179 p->last_io_write_calls = p->io_write_calls;
1180 p->last_io_storage_bytes_read = p->io_storage_bytes_read;
1181 p->last_io_storage_bytes_written = p->io_storage_bytes_written;
1182 p->last_io_cancelled_write_bytes = p->io_cancelled_write_bytes;
1185 while((file = readdir(dir))) {
1186 char *endptr = file->d_name;
1187 pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
1189 // make sure we read a valid number
1190 if(unlikely(endptr == file->d_name || *endptr != '\0'))
1193 if(unlikely(pid <= 0 || pid > pid_max)) {
1194 error("Invalid pid %d read (expected 1 to %d). Ignoring process.", pid, pid_max);
1198 p = get_pid_entry(pid);
1199 if(unlikely(!p)) continue;
1202 // --------------------------------------------------------------------
1205 if(unlikely(read_proc_pid_stat(p))) {
1206 error("Cannot process %s/proc/%d/stat", host_prefix, pid);
1207 // there is no reason to proceed if we cannot get its status
1211 // check its parent pid
1212 if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
1213 error("Pid %d states invalid parent pid %d. Using 0.", pid, p->ppid);
1217 // --------------------------------------------------------------------
1218 // /proc/<pid>/cmdline
1220 if(proc_pid_cmdline_is_needed) {
1221 if(unlikely(read_proc_pid_cmdline(p))) {
1222 error("Cannot process %s/proc/%d/cmdline", host_prefix, pid);
1226 // --------------------------------------------------------------------
1227 // /proc/<pid>/statm
1229 if(unlikely(read_proc_pid_statm(p))) {
1230 error("Cannot process %s/proc/%d/statm", host_prefix, pid);
1232 // there is no reason to proceed if we cannot get its memory status
1237 // --------------------------------------------------------------------
1240 if(unlikely(read_proc_pid_io(p))) {
1241 error("Cannot process %s/proc/%d/io", host_prefix, pid);
1243 // on systems without /proc/X/io
1244 // allow proceeding without I/O information
1248 // --------------------------------------------------------------------
1251 if(unlikely(read_proc_pid_ownership(p))) {
1252 error("Cannot stat %s/proc/%d", host_prefix, pid);
1255 // --------------------------------------------------------------------
1258 // check if it is target
1259 // we do this only once, the first time this pid is loaded
1260 if(unlikely(p->new_entry)) {
1262 fprintf(stderr, "apps.plugin: \tJust added %s\n", p->comm);
1264 uint32_t hash = simple_hash(p->comm);
1265 size_t pclen = strlen(p->comm);
1268 for(w = apps_groups_root_target; w ; w = w->next) {
1269 // if(debug || (p->target && p->target->debug)) fprintf(stderr, "apps.plugin: \t\tcomparing '%s' with '%s'\n", w->compare, p->comm);
1271 // find it - 4 cases:
1272 // 1. the target is not a pattern
1273 // 2. the target has the prefix
1274 // 3. the target has the suffix
1275 // 4. the target is something inside cmdline
1276 if( (!w->starts_with && !w->ends_with && w->comparehash == hash && !strcmp(w->compare, p->comm))
1277 || (w->starts_with && !w->ends_with && !strncmp(w->compare, p->comm, w->comparelen))
1278 || (!w->starts_with && w->ends_with && pclen >= w->comparelen && !strcmp(w->compare, &p->comm[pclen - w->comparelen]))
1279 || (proc_pid_cmdline_is_needed && w->starts_with && w->ends_with && strstr(p->cmdline, w->compare))
1281 if(w->target) p->target = w->target;
1284 if(debug || (p->target && p->target->debug))
1285 fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
1290 // --------------------------------------------------------------------
1293 if(unlikely(read_pid_file_descriptors(p))) {
1294 error("Cannot process entries in %s/proc/%d/fd", host_prefix, pid);
1297 // --------------------------------------------------------------------
1300 // mark it as updated
1309 // ----------------------------------------------------------------------------
1310 // update statistics on the targets
1312 // 1. link all childs to their parents
1313 // 2. go from bottom to top, marking as merged all childs to their parents
1314 // this step links all parents without a target to the child target, if any
1315 // 3. link all top level processes (the ones not merged) to the default target
1316 // 4. go from top to bottom, linking all childs without a target, to their parent target
1317 // after this step, all processes have a target
1318 // [5. for each killed pid (updated = 0), remove its usage from its target]
1319 // 6. zero all apps_groups_targets
1320 // 7. concentrate all values on the apps_groups_targets
1321 // 8. remove all killed processes
1322 // 9. find the unique file count for each target
1323 // check: update_apps_groups_statistics()
1325 void link_all_processes_to_their_parents(void) {
1326 struct pid_stat *p = NULL;
1328 // link all children to their parents
1329 // and update children count on parents
1330 for(p = root_of_pids; p ; p = p->next) {
1331 // for each process found running
1333 if(likely(p->new_entry && p->updated)) {
1334 // the first time we see an entry
1335 // we remove the exited children figures
1337 p->fix_cminflt = p->cminflt;
1338 p->fix_cmajflt = p->cmajflt;
1339 p->fix_cutime = p->cutime;
1340 p->fix_cstime = p->cstime;
1343 if(likely(p->ppid > 0 && all_pids[p->ppid])) {
1344 // valid parent processes
1346 struct pid_stat *pp;
1348 p->parent = pp = all_pids[p->ppid];
1349 p->parent->children_count++;
1351 if(unlikely(debug || (p->target && p->target->debug)))
1352 fprintf(stderr, "apps.plugin: \tchild %d (%s, %s) has parent %d (%s, %s). Parent: utime=%llu, stime=%llu, minflt=%llu, majflt=%llu, cutime=%llu, cstime=%llu, cminflt=%llu, cmajflt=%llu, fix_cutime=%llu, fix_cstime=%llu, fix_cminflt=%llu, fix_cmajflt=%llu\n", p->pid, p->comm, p->updated?"running":"exited", pp->pid, pp->comm, pp->updated?"running":"exited", pp->utime, pp->stime, pp->minflt, pp->majflt, pp->cutime, pp->cstime, pp->cminflt, pp->cmajflt, pp->fix_cutime, pp->fix_cstime, pp->fix_cminflt, pp->fix_cmajflt);
1354 if(unlikely(!p->updated)) {
1355 // this process has exit
1357 // find the first parent that has been updated
1358 while(pp && !pp->updated) {
1359 // we may have to forward link it to its parent
1360 if(unlikely(!pp->parent && pp->ppid > 0 && all_pids[pp->ppid]))
1361 pp->parent = all_pids[pp->ppid];
1363 // check again for parent
1368 // this is an exited child with a parent
1369 // remove the known time from the parent's data
1370 pp->fix_cminflt += p->minflt + p->cminflt + p->fix_cminflt;
1371 pp->fix_cmajflt += p->majflt + p->cmajflt + p->fix_cmajflt;
1372 pp->fix_cutime += p->utime + p->cutime + p->fix_cutime;
1373 pp->fix_cstime += p->stime + p->cstime + p->fix_cstime;
1375 if(unlikely(pp->cminflt < pp->fix_cminflt)) pp->fix_cminflt = pp->cminflt;
1376 if(unlikely(pp->cmajflt < pp->fix_cmajflt)) pp->fix_cmajflt = pp->cmajflt;
1377 if(unlikely(pp->cutime < pp->fix_cutime)) pp->fix_cutime = pp->cutime;
1378 if(unlikely(pp->cstime < pp->fix_cstime)) pp->fix_cstime = pp->cstime;
1381 fprintf(stderr, "apps.plugin: \tupdating child metrics of %d (%s, %s) to its parent %d (%s, %s). Parent has now: utime=%llu, stime=%llu, minflt=%llu, majflt=%llu, cutime=%llu, cstime=%llu, cminflt=%llu, cmajflt=%llu, fix_cutime=%llu, fix_cstime=%llu, fix_cminflt=%llu, fix_cmajflt=%llu\n", p->pid, p->comm, p->updated?"running":"exited", pp->pid, pp->comm, pp->updated?"running":"exited", pp->utime, pp->stime, pp->minflt, pp->majflt, pp->cutime, pp->cstime, pp->cminflt, pp->cmajflt, pp->fix_cutime, pp->fix_cstime, pp->fix_cminflt, pp->fix_cmajflt);
1385 else if(unlikely(p->ppid != 0))
1386 error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
1391 void cleanup_non_existing_pids(void) {
1393 struct pid_stat *p = NULL;
1395 for(p = root_of_pids; p ;) {
1397 // fprintf(stderr, "\tEXITED %d %s [parent %d %s, target %s] utime=%llu, stime=%llu, cutime=%llu, cstime=%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->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt);
1399 for(c = 0 ; c < p->fds_size ; c++) if(p->fds[c] > 0) {
1400 file_descriptor_not_used(p->fds[c]);
1412 void apply_apps_groups_targets_inheritance(void) {
1413 struct pid_stat *p = NULL;
1415 // children that do not have a target
1416 // inherit their target from their parent
1417 int found = 1, loops = 0;
1419 if(unlikely(debug)) loops++;
1421 for(p = root_of_pids; p ; p = p->next) {
1422 // if this process does not have a target
1423 // and it has a parent
1424 // and its parent has a target
1425 // then, set the parent's target to this process
1426 if(unlikely(!p->target && p->parent && p->parent->target)) {
1427 p->target = p->parent->target;
1430 if(debug || (p->target && p->target->debug))
1431 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);
1437 // find all the procs with 0 childs and merge them to their parents
1438 // repeat, until nothing more can be done.
1441 if(unlikely(debug)) loops++;
1443 for(p = root_of_pids; p ; p = p->next) {
1444 // if this process does not have any children
1445 // and is not already merged
1447 // and its parent has children
1448 // and the target of this process and its parent is the same, or the parent does not have a target
1449 // and its parent is not init
1450 // then, mark them as merged.
1455 && p->parent->children_count
1456 && (p->target == p->parent->target || !p->parent->target)
1459 p->parent->children_count--;
1462 // the parent inherits the child's target, if it does not have a target itself
1463 if(unlikely(p->target && !p->parent->target)) {
1464 p->parent->target = p->target;
1466 if(debug || (p->target && p->target->debug))
1467 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);
1475 fprintf(stderr, "apps.plugin: merged %d processes\n", found);
1478 // init goes always to default target
1480 all_pids[1]->target = apps_groups_default_target;
1482 // give a default target on all top level processes
1483 if(unlikely(debug)) loops++;
1484 for(p = root_of_pids; p ; p = p->next) {
1485 // if the process is not merged itself
1486 // then is is a top level process
1487 if(!p->merged && !p->target)
1488 p->target = apps_groups_default_target;
1491 // give a target to all merged child processes
1494 if(unlikely(debug)) loops++;
1496 for(p = root_of_pids; p ; p = p->next) {
1497 if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
1498 p->target = p->parent->target;
1501 if(debug || (p->target && p->target->debug))
1502 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);
1508 fprintf(stderr, "apps.plugin: apply_apps_groups_targets_inheritance() made %d loops on the process tree\n", loops);
1511 long zero_all_targets(struct target *root) {
1515 for (w = root; w ; w = w->next) {
1518 if(w->fds) free(w->fds);
1534 w->statm_resident = 0;
1541 w->io_logical_bytes_read = 0;
1542 w->io_logical_bytes_written = 0;
1543 w->io_read_calls = 0;
1544 w->io_write_calls = 0;
1545 w->io_storage_bytes_read = 0;
1546 w->io_storage_bytes_written = 0;
1547 w->io_cancelled_write_bytes = 0;
1553 void aggregate_pid_on_target(struct target *w, struct pid_stat *p, struct target *o) {
1554 if(unlikely(!w->fds)) {
1555 w->fds = calloc(sizeof(int), (size_t) all_files_size);
1556 if(unlikely(!w->fds))
1557 error("Cannot allocate memory for fds in %s", w->name);
1560 if(likely(p->updated)) {
1561 if(unlikely(debug && (p->fix_cutime || p->fix_cstime || p->fix_cminflt || p->fix_cmajflt)))
1562 fprintf(stderr, "apps.plugin: \tadding child counters of %d (%s) to target %s. Currents: cutime=%llu, cstime=%llu, cminflt=%llu, cmajflt=%llu, Fixes: cutime=%llu, cstime=%llu, cminflt=%llu, cmajflt=%llu\n", p->pid, p->comm, w->name, p->cutime, p->cstime, p->cminflt, p->cmajflt, p->fix_cutime, p->fix_cstime, p->fix_cminflt, p->fix_cmajflt);
1564 w->cutime += p->cutime - p->fix_cutime;
1565 w->cstime += p->cstime - p->fix_cstime;
1566 w->cminflt += p->cminflt - p->fix_cminflt;
1567 w->cmajflt += p->cmajflt - p->fix_cmajflt;
1569 w->utime += p->utime; //+ (p->pid != 1)?(p->cutime - p->fix_cutime):0;
1570 w->stime += p->stime; //+ (p->pid != 1)?(p->cstime - p->fix_cstime):0;
1571 w->minflt += p->minflt; //+ (p->pid != 1)?(p->cminflt - p->fix_cminflt):0;
1572 w->majflt += p->majflt; //+ (p->pid != 1)?(p->cmajflt - p->fix_cmajflt):0;
1574 //if(p->num_threads < 0)
1575 // error("Negative threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1577 //if(p->num_threads > 10000)
1578 // error("Excessive threads number for pid '%s' (%d): %d", p->comm, p->pid, p->num_threads);
1580 w->num_threads += p->num_threads;
1583 w->statm_size += p->statm_size;
1584 w->statm_resident += p->statm_resident;
1585 w->statm_share += p->statm_share;
1586 w->statm_text += p->statm_text;
1587 w->statm_lib += p->statm_lib;
1588 w->statm_data += p->statm_data;
1589 w->statm_dirty += p->statm_dirty;
1591 w->io_logical_bytes_read += p->io_logical_bytes_read;
1592 w->io_logical_bytes_written += p->io_logical_bytes_written;
1593 w->io_read_calls += p->io_read_calls;
1594 w->io_write_calls += p->io_write_calls;
1595 w->io_storage_bytes_read += p->io_storage_bytes_read;
1596 w->io_storage_bytes_written += p->io_storage_bytes_written;
1597 w->io_cancelled_write_bytes += p->io_cancelled_write_bytes;
1601 if(likely(w->fds)) {
1603 for(c = 0; c < p->fds_size ;c++) {
1604 if(p->fds[c] == 0) continue;
1606 if(likely(p->fds[c] < all_files_size)) {
1607 if(w->fds) w->fds[p->fds[c]]++;
1610 error("Invalid fd number %d", p->fds[c]);
1614 if(unlikely(debug || w->debug))
1615 fprintf(stderr, "apps.plugin: \tAggregating %s pid %d on %s utime=%llu, stime=%llu, cutime=%llu, cstime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, cmajflt=%llu, fix_cutime=%llu, fix_cstime=%llu, fix_cminflt=%llu, fix_cmajflt=%llu\n", p->comm, p->pid, w->name, p->utime, p->stime, p->cutime, p->cstime, p->minflt, p->majflt, p->cminflt, p->cmajflt, p->fix_cutime, p->fix_cstime, p->fix_cminflt, p->fix_cmajflt);
1617 /* if(p->utime - p->old_utime > 100) fprintf(stderr, "BIG CHANGE: %d %s utime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->utime - p->old_utime, p->old_utime, p->utime);
1618 if(p->cutime - p->old_cutime > 100) fprintf(stderr, "BIG CHANGE: %d %s cutime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cutime - p->old_cutime, p->old_cutime, p->cutime);
1619 if(p->stime - p->old_stime > 100) fprintf(stderr, "BIG CHANGE: %d %s stime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->stime - p->old_stime, p->old_stime, p->stime);
1620 if(p->cstime - p->old_cstime > 100) fprintf(stderr, "BIG CHANGE: %d %s cstime increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cstime - p->old_cstime, p->old_cstime, p->cstime);
1621 if(p->minflt - p->old_minflt > 5000) fprintf(stderr, "BIG CHANGE: %d %s minflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->minflt - p->old_minflt, p->old_minflt, p->minflt);
1622 if(p->majflt - p->old_majflt > 5000) fprintf(stderr, "BIG CHANGE: %d %s majflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->majflt - p->old_majflt, p->old_majflt, p->majflt);
1623 if(p->cminflt - p->old_cminflt > 15000) fprintf(stderr, "BIG CHANGE: %d %s cminflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cminflt - p->old_cminflt, p->old_cminflt, p->cminflt);
1624 if(p->cmajflt - p->old_cmajflt > 15000) fprintf(stderr, "BIG CHANGE: %d %s cmajflt increased by %llu from %llu to %llu\n", p->pid, p->comm, p->cmajflt - p->old_cmajflt, p->old_cmajflt, p->cmajflt);
1628 // since the process switched target
1629 // for all incremental values
1630 // we have to subtract its OLD values from the new target
1631 // and add its OLD values to the old target
1634 // We add/subtract the last/OLD values we added to the target
1636 unsigned long long cutime = p->last_cutime - p->last_fix_cutime;
1637 unsigned long long cstime = p->last_cstime - p->last_fix_cstime;
1638 unsigned long long cminflt = p->last_cminflt - p->last_fix_cminflt;
1639 unsigned long long cmajflt = p->last_cmajflt - p->last_fix_cmajflt;
1641 w->fix_cutime -= cutime;
1642 w->fix_cstime -= cstime;
1643 w->fix_cminflt -= cminflt;
1644 w->fix_cmajflt -= cmajflt;
1646 w->fix_utime -= p->last_utime;
1647 w->fix_stime -= p->last_stime;
1648 w->fix_minflt -= p->last_minflt;
1649 w->fix_majflt -= p->last_majflt;
1651 w->fix_io_logical_bytes_read -= p->last_io_logical_bytes_read;
1652 w->fix_io_logical_bytes_written -= p->last_io_logical_bytes_written;
1653 w->fix_io_read_calls -= p->last_io_read_calls;
1654 w->fix_io_write_calls -= p->last_io_write_calls;
1655 w->fix_io_storage_bytes_read -= p->last_io_storage_bytes_read;
1656 w->fix_io_storage_bytes_written -= p->last_io_storage_bytes_written;
1657 w->fix_io_cancelled_write_bytes -= p->last_io_cancelled_write_bytes;
1661 o->fix_cutime += cutime;
1662 o->fix_cstime += cstime;
1663 o->fix_cminflt += cminflt;
1664 o->fix_cmajflt += cmajflt;
1666 o->fix_utime += p->last_utime;
1667 o->fix_stime += p->last_stime;
1668 o->fix_minflt += p->last_minflt;
1669 o->fix_majflt += p->last_majflt;
1671 o->fix_io_logical_bytes_read += p->last_io_logical_bytes_read;
1672 o->fix_io_logical_bytes_written += p->last_io_logical_bytes_written;
1673 o->fix_io_read_calls += p->last_io_read_calls;
1674 o->fix_io_write_calls += p->last_io_write_calls;
1675 o->fix_io_storage_bytes_read += p->last_io_storage_bytes_read;
1676 o->fix_io_storage_bytes_written += p->last_io_storage_bytes_written;
1677 o->fix_io_cancelled_write_bytes += p->last_io_cancelled_write_bytes;
1681 // if(o) fprintf(stderr, "apps.plugin: \t\tpid %d (%s) is not updated by OLD target %s (%s) is present.\n", p->pid, p->comm, o->id, o->name);
1683 // since the process has exited, the user
1684 // will see a drop in our charts, because the incremental
1685 // values of this process will not be there from now on
1687 // add them to the fix_* values and they will be added to
1688 // the reported values, so that the report goes steady
1690 w->fix_minflt += p->last_minflt;
1691 w->fix_majflt += p->last_majflt;
1692 w->fix_utime += p->last_utime;
1693 w->fix_stime += p->last_stime;
1695 w->fix_cminflt += (p->last_cminflt - p->last_fix_cminflt);
1696 w->fix_cmajflt += (p->last_cmajflt - p->last_fix_cmajflt);
1697 w->fix_cutime += (p->last_cutime - p->last_fix_cutime);
1698 w->fix_cstime += (p->last_cstime - p->last_fix_cstime);
1700 w->fix_io_logical_bytes_read += p->last_io_logical_bytes_read;
1701 w->fix_io_logical_bytes_written += p->last_io_logical_bytes_written;
1702 w->fix_io_read_calls += p->last_io_read_calls;
1703 w->fix_io_write_calls += p->last_io_write_calls;
1704 w->fix_io_storage_bytes_read += p->last_io_storage_bytes_read;
1705 w->fix_io_storage_bytes_written += p->last_io_storage_bytes_written;
1706 w->fix_io_cancelled_write_bytes += p->last_io_cancelled_write_bytes;
1709 //if((long long)w->cutime + w->fix_cutime < 0)
1710 // error("Negative total cutime (%llu - %lld) on target %s after adding process %d (%s, %s) with utime=%llu, stime=%llu, minflt=%llu, majflt=%llu, cutime=%llu, cstime=%llu, cminflt=%llu, cmajflt=%llu, fix_cutime=%llu, fix_cstime=%llu, fix_cminflt=%llu, fix_cmajflt=%llu\n",
1711 // w->cutime, w->fix_cutime, w->name, p->pid, p->comm, p->updated?"running":"exited", p->utime, p->stime, p->minflt, p->majflt, p->cutime, p->cstime, p->cminflt, p->cmajflt, p->fix_cutime, p->fix_cstime, p->fix_cminflt, p->fix_cmajflt);
1714 void count_targets_fds(struct target *root) {
1718 for (w = root; w ; w = w->next) {
1719 if(!w->fds) continue;
1724 w->openinotifies = 0;
1725 w->openeventfds = 0;
1726 w->opentimerfds = 0;
1727 w->opensignalfds = 0;
1728 w->openeventpolls = 0;
1731 for(c = 1; c < all_files_size ;c++) {
1733 switch(all_files[c].type) {
1742 case FILETYPE_SOCKET:
1746 case FILETYPE_INOTIFY:
1750 case FILETYPE_EVENTFD:
1754 case FILETYPE_TIMERFD:
1758 case FILETYPE_SIGNALFD:
1762 case FILETYPE_EVENTPOLL:
1763 w->openeventpolls++;
1776 void calculate_netdata_statistics(void)
1778 link_all_processes_to_their_parents();
1779 apply_apps_groups_targets_inheritance();
1781 zero_all_targets(users_root_target);
1782 zero_all_targets(groups_root_target);
1783 apps_groups_targets = zero_all_targets(apps_groups_root_target);
1785 // this has to be done, before the cleanup
1786 struct pid_stat *p = NULL;
1787 struct target *w = NULL, *o = NULL;
1789 // concentrate everything on the apps_groups_targets
1790 for(p = root_of_pids; p ; p = p->next) {
1792 // --------------------------------------------------------------------
1793 // apps_groups targets
1794 if(likely(p->target))
1795 aggregate_pid_on_target(p->target, p, NULL);
1797 error("pid %d %s was left without a target!", p->pid, p->comm);
1800 // --------------------------------------------------------------------
1803 if(likely(p->user_target && p->user_target->uid == p->uid))
1806 if(unlikely(debug && p->user_target))
1807 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);
1809 w = p->user_target = get_users_target(p->uid);
1813 aggregate_pid_on_target(w, p, o);
1815 error("pid %d %s was left without a user target!", p->pid, p->comm);
1818 // --------------------------------------------------------------------
1820 o = p->group_target;
1821 if(likely(p->group_target && p->group_target->gid == p->gid))
1822 w = p->group_target;
1824 if(unlikely(debug && p->group_target))
1825 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);
1827 w = p->group_target = get_groups_target(p->gid);
1831 aggregate_pid_on_target(w, p, o);
1833 error("pid %d %s was left without a group target!", p->pid, p->comm);
1837 count_targets_fds(apps_groups_root_target);
1838 count_targets_fds(users_root_target);
1839 count_targets_fds(groups_root_target);
1841 cleanup_non_existing_pids();
1844 // ----------------------------------------------------------------------------
1845 // update chart dimensions
1847 unsigned long long send_resource_usage_to_netdata() {
1848 static struct timeval last = { 0, 0 };
1849 static struct rusage me_last;
1854 unsigned long long usec;
1855 unsigned long long cpuuser;
1856 unsigned long long cpusyst;
1859 gettimeofday(&last, NULL);
1860 getrusage(RUSAGE_SELF, &me_last);
1862 // the first time, give a zero to allow
1863 // netdata calibrate to the current time
1864 // usec = update_every * 1000000ULL;
1870 gettimeofday(&now, NULL);
1871 getrusage(RUSAGE_SELF, &me);
1873 usec = usecdiff(&now, &last);
1874 cpuuser = me.ru_utime.tv_sec * 1000000ULL + me.ru_utime.tv_usec;
1875 cpusyst = me.ru_stime.tv_sec * 1000000ULL + me.ru_stime.tv_usec;
1877 bcopy(&now, &last, sizeof(struct timeval));
1878 bcopy(&me, &me_last, sizeof(struct rusage));
1881 fprintf(stdout, "BEGIN netdata.apps_cpu %llu\n", usec);
1882 fprintf(stdout, "SET user = %llu\n", cpuuser);
1883 fprintf(stdout, "SET system = %llu\n", cpusyst);
1884 fprintf(stdout, "END\n");
1886 fprintf(stdout, "BEGIN netdata.apps_files %llu\n", usec);
1887 fprintf(stdout, "SET files = %llu\n", file_counter);
1888 fprintf(stdout, "SET pids = %ld\n", all_pids_count);
1889 fprintf(stdout, "SET fds = %d\n", all_files_len);
1890 fprintf(stdout, "SET targets = %ld\n", apps_groups_targets);
1891 fprintf(stdout, "END\n");
1896 void send_collected_data_to_netdata(struct target *root, const char *type, unsigned long long usec)
1900 fprintf(stdout, "BEGIN %s.cpu %llu\n", type, usec);
1901 for (w = root; w ; w = w->next) {
1902 if(w->target || (!w->processes && !w->exposed)) continue;
1904 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->stime + w->fix_utime + w->fix_stime + (include_exited_childs?(w->cutime + w->cstime + w->fix_cutime + w->fix_cstime):0));
1906 fprintf(stdout, "END\n");
1908 fprintf(stdout, "BEGIN %s.cpu_user %llu\n", type, usec);
1909 for (w = root; w ; w = w->next) {
1910 if(w->target || (!w->processes && !w->exposed)) continue;
1912 fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->fix_utime + (include_exited_childs?(w->cutime + w->fix_cutime):0));
1914 fprintf(stdout, "END\n");
1916 fprintf(stdout, "BEGIN %s.cpu_system %llu\n", type, usec);
1917 for (w = root; w ; w = w->next) {
1918 if(w->target || (!w->processes && !w->exposed)) continue;
1920 fprintf(stdout, "SET %s = %llu\n", w->name, w->stime + w->fix_stime + (include_exited_childs?(w->cstime + w->fix_cstime):0));
1922 fprintf(stdout, "END\n");
1924 fprintf(stdout, "BEGIN %s.threads %llu\n", type, usec);
1925 for (w = root; w ; w = w->next) {
1926 if(w->target || (!w->processes && !w->exposed)) continue;
1928 fprintf(stdout, "SET %s = %llu\n", w->name, w->num_threads);
1930 fprintf(stdout, "END\n");
1932 fprintf(stdout, "BEGIN %s.processes %llu\n", type, usec);
1933 for (w = root; w ; w = w->next) {
1934 if(w->target || (!w->processes && !w->exposed)) continue;
1936 fprintf(stdout, "SET %s = %lu\n", w->name, w->processes);
1938 fprintf(stdout, "END\n");
1940 fprintf(stdout, "BEGIN %s.mem %llu\n", type, usec);
1941 for (w = root; w ; w = w->next) {
1942 if(w->target || (!w->processes && !w->exposed)) continue;
1944 fprintf(stdout, "SET %s = %lld\n", w->name, (long long)w->statm_resident - (long long)w->statm_share);
1946 fprintf(stdout, "END\n");
1948 fprintf(stdout, "BEGIN %s.minor_faults %llu\n", type, usec);
1949 for (w = root; w ; w = w->next) {
1950 if(w->target || (!w->processes && !w->exposed)) continue;
1952 fprintf(stdout, "SET %s = %llu\n", w->name, w->minflt + w->fix_minflt + (include_exited_childs?(w->cminflt + w->fix_cminflt):0));
1954 fprintf(stdout, "END\n");
1956 fprintf(stdout, "BEGIN %s.major_faults %llu\n", type, usec);
1957 for (w = root; w ; w = w->next) {
1958 if(w->target || (!w->processes && !w->exposed)) continue;
1960 fprintf(stdout, "SET %s = %llu\n", w->name, w->majflt + w->fix_majflt + (include_exited_childs?(w->cmajflt + w->fix_cmajflt):0));
1962 fprintf(stdout, "END\n");
1964 fprintf(stdout, "BEGIN %s.lreads %llu\n", type, usec);
1965 for (w = root; w ; w = w->next) {
1966 if(w->target || (!w->processes && !w->exposed)) continue;
1968 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_read + w->fix_io_logical_bytes_read);
1970 fprintf(stdout, "END\n");
1972 fprintf(stdout, "BEGIN %s.lwrites %llu\n", type, usec);
1973 for (w = root; w ; w = w->next) {
1974 if(w->target || (!w->processes && !w->exposed)) continue;
1976 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_logical_bytes_written + w->fix_io_logical_bytes_written);
1978 fprintf(stdout, "END\n");
1980 fprintf(stdout, "BEGIN %s.preads %llu\n", type, usec);
1981 for (w = root; w ; w = w->next) {
1982 if(w->target || (!w->processes && !w->exposed)) continue;
1984 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_read + w->fix_io_storage_bytes_read);
1986 fprintf(stdout, "END\n");
1988 fprintf(stdout, "BEGIN %s.pwrites %llu\n", type, usec);
1989 for (w = root; w ; w = w->next) {
1990 if(w->target || (!w->processes && !w->exposed)) continue;
1992 fprintf(stdout, "SET %s = %llu\n", w->name, w->io_storage_bytes_written + w->fix_io_storage_bytes_written);
1994 fprintf(stdout, "END\n");
1996 fprintf(stdout, "BEGIN %s.files %llu\n", type, usec);
1997 for (w = root; w ; w = w->next) {
1998 if(w->target || (!w->processes && !w->exposed)) continue;
2000 fprintf(stdout, "SET %s = %llu\n", w->name, w->openfiles);
2002 fprintf(stdout, "END\n");
2004 fprintf(stdout, "BEGIN %s.sockets %llu\n", type, usec);
2005 for (w = root; w ; w = w->next) {
2006 if(w->target || (!w->processes && !w->exposed)) continue;
2008 fprintf(stdout, "SET %s = %llu\n", w->name, w->opensockets);
2010 fprintf(stdout, "END\n");
2012 fprintf(stdout, "BEGIN %s.pipes %llu\n", type, usec);
2013 for (w = root; w ; w = w->next) {
2014 if(w->target || (!w->processes && !w->exposed)) continue;
2016 fprintf(stdout, "SET %s = %llu\n", w->name, w->openpipes);
2018 fprintf(stdout, "END\n");
2024 // ----------------------------------------------------------------------------
2025 // generate the charts
2027 void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title)
2030 int newly_added = 0;
2032 for(w = root ; w ; w = w->next)
2033 if(!w->exposed && w->processes) {
2036 if(debug || w->debug) fprintf(stderr, "apps.plugin: %s just added - regenerating charts.\n", w->name);
2039 // nothing more to show
2040 if(!newly_added) return;
2042 // we have something new to show
2043 // update the charts
2044 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);
2045 for (w = root; w ; w = w->next) {
2046 if(w->target || (!w->processes && !w->exposed)) continue;
2048 fprintf(stdout, "DIMENSION %s '' incremental 100 %u %s\n", w->name, hz, w->hidden ? "hidden,noreset" : "noreset");
2051 fprintf(stdout, "CHART %s.mem '' '%s Dedicated Memory (w/o shared)' 'MB' mem %s.mem stacked 20003 %d\n", type, title, type, update_every);
2052 for (w = root; w ; w = w->next) {
2053 if(w->target || (!w->processes && !w->exposed)) continue;
2055 fprintf(stdout, "DIMENSION %s '' absolute %ld %ld noreset\n", w->name, sysconf(_SC_PAGESIZE), 1024L*1024L);
2058 fprintf(stdout, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20005 %d\n", type, title, type, update_every);
2059 for (w = root; w ; w = w->next) {
2060 if(w->target || (!w->processes && !w->exposed)) continue;
2062 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
2065 fprintf(stdout, "CHART %s.processes '' '%s Processes' 'processes' processes %s.processes stacked 20004 %d\n", type, title, type, update_every);
2066 for (w = root; w ; w = w->next) {
2067 if(w->target || (!w->processes && !w->exposed)) continue;
2069 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
2072 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);
2073 for (w = root; w ; w = w->next) {
2074 if(w->target || (!w->processes && !w->exposed)) continue;
2076 fprintf(stdout, "DIMENSION %s '' incremental 100 %d noreset\n", w->name, hz);
2079 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);
2080 for (w = root; w ; w = w->next) {
2081 if(w->target || (!w->processes && !w->exposed)) continue;
2083 fprintf(stdout, "DIMENSION %s '' incremental 100 %d noreset\n", w->name, hz);
2086 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);
2087 for (w = root; w ; w = w->next) {
2088 if(w->target || (!w->processes && !w->exposed)) continue;
2090 fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
2093 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);
2094 for (w = root; w ; w = w->next) {
2095 if(w->target || (!w->processes && !w->exposed)) continue;
2097 fprintf(stdout, "DIMENSION %s '' incremental 1 1 noreset\n", w->name);
2100 fprintf(stdout, "CHART %s.lreads '' '%s Disk Logical Reads' 'kilobytes/s' disk %s.lreads stacked 20042 %d\n", type, title, type, update_every);
2101 for (w = root; w ; w = w->next) {
2102 if(w->target || (!w->processes && !w->exposed)) continue;
2104 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
2107 fprintf(stdout, "CHART %s.lwrites '' '%s I/O Logical Writes' 'kilobytes/s' disk %s.lwrites stacked 20042 %d\n", type, title, type, update_every);
2108 for (w = root; w ; w = w->next) {
2109 if(w->target || (!w->processes && !w->exposed)) continue;
2111 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
2114 fprintf(stdout, "CHART %s.preads '' '%s Disk Reads' 'kilobytes/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every);
2115 for (w = root; w ; w = w->next) {
2116 if(w->target || (!w->processes && !w->exposed)) continue;
2118 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
2121 fprintf(stdout, "CHART %s.pwrites '' '%s Disk Writes' 'kilobytes/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every);
2122 for (w = root; w ; w = w->next) {
2123 if(w->target || (!w->processes && !w->exposed)) continue;
2125 fprintf(stdout, "DIMENSION %s '' incremental 1 %d noreset\n", w->name, 1024);
2128 fprintf(stdout, "CHART %s.files '' '%s Open Files' 'open files' disk %s.files stacked 20050 %d\n", type, title, type, update_every);
2129 for (w = root; w ; w = w->next) {
2130 if(w->target || (!w->processes && !w->exposed)) continue;
2132 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
2135 fprintf(stdout, "CHART %s.sockets '' '%s Open Sockets' 'open sockets' net %s.sockets stacked 20051 %d\n", type, title, type, update_every);
2136 for (w = root; w ; w = w->next) {
2137 if(w->target || (!w->processes && !w->exposed)) continue;
2139 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
2142 fprintf(stdout, "CHART %s.pipes '' '%s Pipes' 'open pipes' processes %s.pipes stacked 20053 %d\n", type, title, type, update_every);
2143 for (w = root; w ; w = w->next) {
2144 if(w->target || (!w->processes && !w->exposed)) continue;
2146 fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
2151 // ----------------------------------------------------------------------------
2152 // parse command line arguments
2154 void parse_args(int argc, char **argv)
2159 for(i = 1; i < argc; i++) {
2161 int n = atoi(argv[i]);
2168 if(strcmp("debug", argv[i]) == 0) {
2170 debug_flags = 0xffffffff;
2174 if(strcmp("no-childs", argv[i]) == 0) {
2175 include_exited_childs = 0;
2179 if(strcmp("with-childs", argv[i]) == 0) {
2180 include_exited_childs = 1;
2189 error("Cannot understand option %s", argv[i]);
2193 if(freq > 0) update_every = freq;
2194 if(!name) name = "groups";
2196 if(read_apps_groups_conf(name)) {
2197 error("Cannot read process groups %s", name);
2202 int main(int argc, char **argv)
2204 // debug_flags = D_PROCFILE;
2206 // set the name for logging
2207 program_name = "apps.plugin";
2209 // disable syslog for apps.plugin
2210 error_log_syslog = 0;
2212 // set errors flood protection to 100 logs per hour
2213 error_log_errors_per_period = 100;
2214 error_log_throttle_period = 3600;
2216 host_prefix = getenv("NETDATA_HOST_PREFIX");
2217 if(host_prefix == NULL) {
2218 info("NETDATA_HOST_PREFIX is not passed from netdata");
2221 else info("Found NETDATA_HOST_PREFIX='%s'", host_prefix);
2223 config_dir = getenv("NETDATA_CONFIG_DIR");
2224 if(config_dir == NULL) {
2225 info("NETDATA_CONFIG_DIR is not passed from netdata");
2226 config_dir = CONFIG_DIR;
2228 else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
2230 #ifdef NETDATA_INTERNAL_CHECKS
2231 if(debug_flags != 0) {
2232 struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
2233 if(setrlimit(RLIMIT_CORE, &rl) != 0)
2234 info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
2235 prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
2237 #endif /* NETDATA_INTERNAL_CHECKS */
2239 procfile_adaptive_initial_allocation = 1;
2241 time_t started_t = time(NULL);
2244 pid_max = get_system_pid_max();
2245 processors = get_system_cpus();
2247 parse_args(argc, argv);
2249 all_pids = calloc(sizeof(struct pid_stat *), (size_t) pid_max);
2251 error("Cannot allocate %lu bytes of memory.", sizeof(struct pid_stat *) * pid_max);
2252 printf("DISABLE\n");
2256 fprintf(stdout, "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %1$d\n"
2257 "DIMENSION user '' incremental 1 1000\n"
2258 "DIMENSION system '' incremental 1 1000\n"
2259 "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %1$d\n"
2260 "DIMENSION files '' incremental 1 1\n"
2261 "DIMENSION pids '' absolute 1 1\n"
2262 "DIMENSION fds '' absolute 1 1\n"
2263 "DIMENSION targets '' absolute 1 1\n", update_every);
2265 #ifndef PROFILING_MODE
2266 unsigned long long sunext = (time(NULL) - (time(NULL) % update_every) + update_every) * 1000000ULL;
2267 unsigned long long sunow;
2268 #endif /* PROFILING_MODE */
2270 unsigned long long counter = 1;
2271 for(;1; counter++) {
2272 #ifndef PROFILING_MODE
2273 // delay until it is our time to run
2274 while((sunow = timems()) < sunext)
2275 usleep((useconds_t)(sunext - sunow));
2277 // find the next time we need to run
2278 while(timems() > sunext)
2279 sunext += update_every * 1000000ULL;
2280 #endif /* PROFILING_MODE */
2282 if(!collect_data_for_all_processes_from_proc()) {
2283 error("Cannot collect /proc data for running processes. Disabling apps.plugin...");
2284 printf("DISABLE\n");
2288 calculate_netdata_statistics();
2290 unsigned long long dt = send_resource_usage_to_netdata();
2292 // this is smart enough to show only newly added apps, when needed
2293 send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps");
2294 send_charts_updates_to_netdata(users_root_target, "users", "Users");
2295 send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups");
2297 send_collected_data_to_netdata(apps_groups_root_target, "apps", dt);
2298 send_collected_data_to_netdata(users_root_target, "users", dt);
2299 send_collected_data_to_netdata(groups_root_target, "groups", dt);
2302 fprintf(stderr, "apps.plugin: done Loop No %llu\n", counter);
2304 current_t = time(NULL);
2306 #ifndef PROFILING_MODE
2307 // restart check (14400 seconds)
2308 if(current_t - started_t > 14400) exit(0);
2310 if(current_t - started_t > 10) exit(0);
2311 #endif /* PROFILING_MODE */