-// TODO
-//
-// 1. disable RESET_OR_OVERFLOW check in charts
-
#ifdef HAVE_CONFIG_H
#include <config.h>
#endif
#include "procfile.h"
#include "../config.h"
+#ifdef NETDATA_INTERNAL_CHECKS
+#include <sys/prctl.h>
+#endif
+
#define MAX_COMPARE_NAME 100
#define MAX_NAME 100
#define MAX_CMDLINE 1024
-unsigned long long Hertz = 1;
-
-long processors = 1;
-long pid_max = 32768;
+int processors = 1;
+pid_t pid_max = 32768;
int debug = 0;
int update_every = 1;
unsigned long long file_counter = 0;
int proc_pid_cmdline_is_needed = 0;
-
+int include_exited_childs = 1;
char *host_prefix = "";
char *config_dir = CONFIG_DIR;
-#ifdef NETDATA_INTERNAL_CHECKS
-// ----------------------------------------------------------------------------
-// memory debugger
-// do not use in production systems - it mis-aligns allocated memory
-
-struct allocations {
- size_t allocations;
- size_t allocated;
- size_t allocated_max;
-} allocations = { 0, 0, 0 };
-
-#define MALLOC_MARK (uint32_t)(0x0BADCAFE)
-#define MALLOC_PREFIX (sizeof(uint32_t) * 2)
-#define MALLOC_SUFFIX (sizeof(uint32_t))
-#define MALLOC_OVERHEAD (MALLOC_PREFIX + MALLOC_SUFFIX)
-
-void *mark_allocation(void *allocated_ptr, size_t size_without_overheads) {
- uint32_t *real_ptr = (uint32_t *)allocated_ptr;
- real_ptr[0] = MALLOC_MARK;
- real_ptr[1] = (uint32_t) size_without_overheads;
-
- uint32_t *end_ptr = (uint32_t *)(allocated_ptr + MALLOC_PREFIX + size_without_overheads);
- end_ptr[0] = MALLOC_MARK;
-
- // fprintf(stderr, "MEMORY_POINTER: Allocated at %p, returning %p.\n", allocated_ptr, (void *)(allocated_ptr + MALLOC_PREFIX));
-
- return allocated_ptr + MALLOC_PREFIX;
-}
-
-void *check_allocation(const char *file, int line, const char *function, void *marked_ptr, size_t *size_without_overheads_ptr) {
- uint32_t *real_ptr = (uint32_t *)(marked_ptr - MALLOC_PREFIX);
-
- // fprintf(stderr, "MEMORY_POINTER: Checking pointer at %p, real %p for %s/%u@%s.\n", marked_ptr, (void *)(marked_ptr - MALLOC_PREFIX), function, line, file);
-
- if(real_ptr[0] != MALLOC_MARK) fatal("MEMORY: prefix MARK is not valid for %s/%u@%s.", function, line, file);
-
- size_t size = real_ptr[1];
-
- uint32_t *end_ptr = (uint32_t *)(marked_ptr + size);
- if(end_ptr[0] != MALLOC_MARK) fatal("MEMORY: suffix MARK of allocation with size %zu is not valid for %s/%u@%s.", size, function, line, file);
-
- if(size_without_overheads_ptr) *size_without_overheads_ptr = size;
-
- return real_ptr;
-}
-
-void *malloc_debug(const char *file, int line, const char *function, size_t size) {
- void *ptr = malloc(size + MALLOC_OVERHEAD);
- if(!ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
-
- allocations.allocated += size;
- allocations.allocations++;
-
- debug(D_MEMORY, "MEMORY: Allocated %zu bytes for %s/%u@%s."
- " Status: allocated %zu in %zu allocs."
- , size
- , function, line, file
- , allocations.allocated
- , allocations.allocations
- );
-
- if(allocations.allocated > allocations.allocated_max) {
- debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
- allocations.allocated_max = allocations.allocated;
- }
-
- size_t csize;
- check_allocation(file, line, function, mark_allocation(ptr, size), &csize);
- if(size != csize) {
- fatal("Invalid size.");
- }
-
- return mark_allocation(ptr, size);
-}
-
-void *calloc_debug(const char *file, int line, const char *function, size_t nmemb, size_t size) {
- void *ptr = malloc_debug(file, line, function, (nmemb * size));
- bzero(ptr, nmemb * size);
- return ptr;
-}
-
-void free_debug(const char *file, int line, const char *function, void *ptr) {
- size_t size;
- void *real_ptr = check_allocation(file, line, function, ptr, &size);
-
- bzero(real_ptr, size + MALLOC_OVERHEAD);
-
- free(real_ptr);
- allocations.allocated -= size;
- allocations.allocations--;
-
- debug(D_MEMORY, "MEMORY: freed %zu bytes for %s/%u@%s."
- " Status: allocated %zu in %zu allocs."
- , size
- , function, line, file
- , allocations.allocated
- , allocations.allocations
- );
-}
-
-void *realloc_debug(const char *file, int line, const char *function, void *ptr, size_t size) {
- if(!ptr) return malloc_debug(file, line, function, size);
- if(!size) { free_debug(file, line, function, ptr); return NULL; }
-
- size_t old_size;
- void *real_ptr = check_allocation(file, line, function, ptr, &old_size);
-
- void *new_ptr = realloc(real_ptr, size + MALLOC_OVERHEAD);
- if(!new_ptr) fatal("MEMORY: Cannot allocate %zu bytes for %s/%u@%s.", size, function, line, file);
-
- allocations.allocated += size;
- allocations.allocated -= old_size;
-
- debug(D_MEMORY, "MEMORY: Re-allocated from %zu to %zu bytes for %s/%u@%s."
- " Status: allocated %z in %zu allocs."
- , old_size, size
- , function, line, file
- , allocations.allocated
- , allocations.allocations
- );
-
- if(allocations.allocated > allocations.allocated_max) {
- debug(D_MEMORY, "MEMORY: total allocation peak increased from %zu to %zu", allocations.allocated_max, allocations.allocated);
- allocations.allocated_max = allocations.allocated;
- }
-
- return mark_allocation(new_ptr, size);
-}
-
-char *strdup_debug(const char *file, int line, const char *function, const char *ptr) {
- size_t size = 0;
- const char *s = ptr;
-
- while(*s++) size++;
- size++;
- char *p = malloc_debug(file, line, function, size);
- if(!p) fatal("Cannot allocate %zu bytes.", size);
+// ----------------------------------------------------------------------------
- memcpy(p, ptr, size);
- return p;
+void netdata_cleanup_and_exit(int ret) {
+ exit(ret);
}
-#define malloc(size) malloc_debug(__FILE__, __LINE__, __FUNCTION__, (size))
-#define calloc(nmemb, size) calloc_debug(__FILE__, __LINE__, __FUNCTION__, (nmemb), (size))
-#define realloc(ptr, size) realloc_debug(__FILE__, __LINE__, __FUNCTION__, (ptr), (size))
-#define free(ptr) free_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
-
-#ifdef strdup
-#undef strdup
-#endif
-#define strdup(ptr) strdup_debug(__FILE__, __LINE__, __FUNCTION__, (ptr))
-
-#endif /* NETDATA_INTERNAL_CHECKS */
-
// ----------------------------------------------------------------------------
// system functions
int processors = 0;
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
+ snprintfz(filename, FILENAME_MAX, "%s/proc/stat", host_prefix);
ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
if(!ff) return 1;
return processors;
}
-long get_system_pid_max(void) {
+pid_t get_system_pid_max(void) {
procfile *ff = NULL;
- long mpid = 32768;
+ pid_t mpid = 32768;
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
+ snprintfz(filename, FILENAME_MAX, "%s/proc/sys/kernel/pid_max", host_prefix);
ff = procfile_open(filename, NULL, PROCFILE_FLAG_DEFAULT);
if(!ff) return mpid;
return mpid;
}
- mpid = atol(procfile_lineword(ff, 0, 0));
+ mpid = (pid_t)atoi(procfile_lineword(ff, 0, 0));
if(!mpid) mpid = 32768;
procfile_close(ff);
return mpid;
}
-unsigned long long get_system_hertz(void)
-{
- unsigned long long myhz = 1;
-
-#ifdef _SC_CLK_TCK
- if((myhz = (unsigned long long int) sysconf(_SC_CLK_TCK)) > 0) {
- return myhz;
- }
-#endif
-
-#ifdef HZ
- myhz = HZ; /* <asm/param.h> */
-#else /* HZ */
- /* If 32-bit or big-endian (not Alpha or ia64), assume HZ is 100. */
- hz = (sizeof(long)==sizeof(int) || htons(999)==999) ? 100UL : 1024UL;
-#endif /* HZ */
-
- error("Unknown HZ value. Assuming %llu.", myhz);
- return myhz;
-}
-
-
// ----------------------------------------------------------------------------
// target
// target is the structure that process data are aggregated
unsigned long long num_threads;
unsigned long long rss;
- unsigned long long fix_minflt;
- unsigned long long fix_cminflt;
- unsigned long long fix_majflt;
- unsigned long long fix_cmajflt;
- unsigned long long fix_utime;
- unsigned long long fix_stime;
- unsigned long long fix_cutime;
- unsigned long long fix_cstime;
+ long long fix_minflt;
+ long long fix_cminflt;
+ long long fix_majflt;
+ long long fix_cmajflt;
+ long long fix_utime;
+ long long fix_stime;
+ long long fix_cutime;
+ long long fix_cstime;
unsigned long long statm_size;
unsigned long long statm_resident;
return NULL;
}
- snprintf(w->compare, MAX_COMPARE_NAME, "%d", uid);
+ snprintfz(w->compare, MAX_COMPARE_NAME, "%u", uid);
w->comparehash = simple_hash(w->compare);
w->comparelen = strlen(w->compare);
- snprintf(w->id, MAX_NAME, "%d", uid);
+ snprintfz(w->id, MAX_NAME, "%u", uid);
w->idhash = simple_hash(w->id);
struct passwd *pw = getpwuid(uid);
if(!pw)
- snprintf(w->name, MAX_NAME, "%d", uid);
+ snprintfz(w->name, MAX_NAME, "%u", uid);
else
- snprintf(w->name, MAX_NAME, "%s", pw->pw_name);
+ snprintfz(w->name, MAX_NAME, "%s", pw->pw_name);
+
+ netdata_fix_chart_name(w->name);
w->uid = uid;
users_root_target = w;
if(unlikely(debug))
- fprintf(stderr, "apps.plugin: added uid %d ('%s') target\n", w->uid, w->name);
+ fprintf(stderr, "apps.plugin: added uid %u ('%s') target\n", w->uid, w->name);
return w;
}
return NULL;
}
- snprintf(w->compare, MAX_COMPARE_NAME, "%d", gid);
+ snprintfz(w->compare, MAX_COMPARE_NAME, "%u", gid);
w->comparehash = simple_hash(w->compare);
w->comparelen = strlen(w->compare);
- snprintf(w->id, MAX_NAME, "%d", gid);
+ snprintfz(w->id, MAX_NAME, "%u", gid);
w->idhash = simple_hash(w->id);
struct group *gr = getgrgid(gid);
if(!gr)
- snprintf(w->name, MAX_NAME, "%d", gid);
+ snprintfz(w->name, MAX_NAME, "%u", gid);
else
- snprintf(w->name, MAX_NAME, "%s", gr->gr_name);
+ snprintfz(w->name, MAX_NAME, "%s", gr->gr_name);
+
+ netdata_fix_chart_name(w->name);
w->gid = gid;
groups_root_target = w;
if(unlikely(debug))
- fprintf(stderr, "apps.plugin: added gid %d ('%s') target\n", w->gid, w->name);
+ fprintf(stderr, "apps.plugin: added gid %u ('%s') target\n", w->gid, w->name);
return w;
}
}
uint32_t hash = simple_hash(id);
- struct target *w;
+ struct target *w, *last = apps_groups_root_target;
for(w = apps_groups_root_target ; w ; w = w->next) {
if(w->idhash == hash && strncmp(nid, w->id, MAX_NAME) == 0)
return w;
+
+ last = w;
}
w = calloc(sizeof(struct target), 1);
return NULL;
}
- strncpy(w->id, nid, MAX_NAME);
+ strncpyz(w->id, nid, MAX_NAME);
w->idhash = simple_hash(w->id);
- strncpy(w->name, nid, MAX_NAME);
+ strncpyz(w->name, nid, MAX_NAME);
- strncpy(w->compare, nid, MAX_COMPARE_NAME);
+ strncpyz(w->compare, nid, MAX_COMPARE_NAME);
int len = strlen(w->compare);
if(w->compare[len - 1] == '*') {
w->compare[len - 1] = '\0';
w->debug = tdebug;
w->target = target;
- w->next = apps_groups_root_target;
- apps_groups_root_target = w;
+ // append it, to maintain the order in apps_groups.conf
+ if(last) last->next = w;
+ else apps_groups_root_target = w;
if(unlikely(debug))
fprintf(stderr, "apps.plugin: ADDING TARGET ID '%s', process name '%s' (%s), aggregated on target '%s', options: %s %s\n"
{
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
+ snprintfz(filename, FILENAME_MAX, "%s/apps_%s.conf", config_dir, name);
if(unlikely(debug))
fprintf(stderr, "apps.plugin: process groups file: '%s'\n", filename);
struct target *n = get_apps_groups_target(s, w);
if(!n) {
- error("Cannot create target '%s' (line %d, word %d)", s, line, word);
+ error("Cannot create target '%s' (line %lu, word %lu)", s, line, word);
continue;
}
t++;
}
- strncpy(w->name, t, MAX_NAME);
- w->name[MAX_NAME] = '\0';
+ strncpyz(w->name, t, MAX_NAME);
w->hidden = thidden;
w->debug = tdebug;
if(!apps_groups_default_target)
error("Cannot create default target");
else
- strncpy(apps_groups_default_target->name, "other", MAX_NAME);
+ strncpyz(apps_groups_default_target->name, "other", MAX_NAME);
return 0;
}
// we will subtract these values from the old
// target
unsigned long long last_minflt;
- unsigned long long last_cminflt;
unsigned long long last_majflt;
- unsigned long long last_cmajflt;
unsigned long long last_utime;
unsigned long long last_stime;
+
+ unsigned long long last_cminflt;
+ unsigned long long last_cmajflt;
unsigned long long last_cutime;
unsigned long long last_cstime;
+ unsigned long long last_fix_cminflt;
+ unsigned long long last_fix_cmajflt;
+ unsigned long long last_fix_cutime;
+ unsigned long long last_fix_cstime;
+
unsigned long long last_io_logical_bytes_read;
unsigned long long last_io_logical_bytes_written;
unsigned long long last_io_read_calls;
unsigned long long last_io_storage_bytes_written;
unsigned long long last_io_cancelled_write_bytes;
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- unsigned long long old_utime;
- unsigned long long old_stime;
- unsigned long long old_minflt;
- unsigned long long old_majflt;
-
- unsigned long long old_cutime;
- unsigned long long old_cstime;
- unsigned long long old_cminflt;
- unsigned long long old_cmajflt;
-
- unsigned long long fix_cutime;
- unsigned long long fix_cstime;
unsigned long long fix_cminflt;
unsigned long long fix_cmajflt;
-
- unsigned long long diff_cutime;
- unsigned long long diff_cstime;
- unsigned long long diff_cminflt;
- unsigned long long diff_cmajflt;
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
+ unsigned long long fix_cutime;
+ unsigned long long fix_cstime;
int *fds; // array of fds it uses
int fds_size; // the size of the fds array
all_pids[pid] = calloc(sizeof(struct pid_stat), 1);
if(!all_pids[pid]) {
- error("Cannot allocate %lu bytes of memory", (unsigned long)sizeof(struct pid_stat));
+ error("Cannot allocate %zu bytes of memory", (size_t)sizeof(struct pid_stat));
return NULL;
}
all_pids[pid]->fds = calloc(sizeof(int), 100);
if(!all_pids[pid]->fds)
- error("Cannot allocate %ld bytes of memory", (unsigned long)(sizeof(int) * 100));
+ error("Cannot allocate %zu bytes of memory", (size_t)(sizeof(int) * 100));
else all_pids[pid]->fds_size = 100;
if(root_of_pids) root_of_pids->prev = all_pids[pid];
{
if(!all_pids[pid]) return;
- if(debug) fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: process %d %s exited, deleting it.\n", pid, all_pids[pid]->comm);
if(root_of_pids == all_pids[pid]) root_of_pids = all_pids[pid]->next;
if(all_pids[pid]->next) all_pids[pid]->next->prev = all_pids[pid]->prev;
int read_proc_pid_cmdline(struct pid_stat *p) {
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/cmdline", host_prefix, p->pid);
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/cmdline", host_prefix, p->pid);
int fd = open(filename, O_RDONLY, 0666);
if(unlikely(fd == -1)) return 1;
if(bytes <= 0) {
// copy the command to the command line
- strncpy(p->cmdline, p->comm, MAX_CMDLINE);
- p->cmdline[MAX_CMDLINE] = '\0';
+ strncpyz(p->cmdline, p->comm, MAX_CMDLINE);
return 0;
}
int read_proc_pid_ownership(struct pid_stat *p) {
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/%d", host_prefix, p->pid);
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d", host_prefix, p->pid);
// ----------------------------------------
// read uid and gid
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/stat", host_prefix, p->pid);
// ----------------------------------------
// parse the process name
unsigned int i = 0;
- strncpy(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
- p->comm[MAX_COMPARE_NAME] = '\0';
+ strncpyz(p->comm, procfile_lineword(ff, 0, 1), MAX_COMPARE_NAME);
// p->pid = atol(procfile_lineword(ff, 0, 0+i));
// comm is at 1
// p->guest_time = strtoull(procfile_lineword(ff, 0, 42+i), NULL, 10);
// p->cguest_time = strtoull(procfile_lineword(ff, 0, 43), NULL, 10);
- if(debug || (p->target && p->target->debug))
+ if(unlikely(debug || (p->target && p->target->debug)))
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);
// procfile_close(ff);
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/statm", host_prefix, p->pid);
ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
if(!ff) return 1;
char filename[FILENAME_MAX + 1];
- snprintf(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
+ snprintfz(filename, FILENAME_MAX, "%s/proc/%d/io", host_prefix, p->pid);
ff = procfile_reopen(ff, filename, NULL, PROCFILE_FLAG_NO_ERROR_ON_FILE_IO);
if(!ff) return 1;
avl_tree all_files_index = {
NULL,
- file_descriptor_compare,
-#ifndef AVL_WITHOUT_PTHREADS
-#ifdef AVL_LOCK_WITH_MUTEX
- PTHREAD_MUTEX_INITIALIZER
-#else
- PTHREAD_RWLOCK_INITIALIZER
-#endif
-#endif /* AVL_WITHOUT_PTHREADS */
+ file_descriptor_compare
};
static struct file_descriptor *file_descriptor_find(const char *name, uint32_t hash) {
- struct file_descriptor *result = NULL, tmp;
+ struct file_descriptor tmp;
tmp.hash = (hash)?hash:simple_hash(name);
tmp.name = name;
tmp.count = 0;
tmp.magic = 0x0BADCAFE;
#endif /* NETDATA_INTERNAL_CHECKS */
- avl_search(&all_files_index, (avl *)&tmp, file_descriptor_iterator, (avl **)&result);
- return result;
+ return (struct file_descriptor *)avl_search(&all_files_index, (avl *) &tmp);
}
#define file_descriptor_add(fd) avl_insert(&all_files_index, (avl *)(fd))
}
#endif /* NETDATA_INTERNAL_CHECKS */
- if(debug) fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: decreasing slot %d (count = %d).\n", id, all_files[id].count);
if(all_files[id].count > 0) {
all_files[id].count--;
if(!all_files[id].count) {
- if(debug) fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> slot %d is empty.\n", id);
+
file_descriptor_remove(&all_files[id]);
#ifdef NETDATA_INTERNAL_CHECKS
all_files[id].magic = 0x00000000;
static int last_pos = 0;
uint32_t hash = simple_hash(name);
- if(debug) fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: adding or finding name '%s' with hash %u\n", name, hash);
struct file_descriptor *fd = file_descriptor_find(name, hash);
if(fd) {
// found
- if(debug) fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> found on slot %d\n", fd->pos);
+
fd->count++;
return fd->pos;
}
int i;
// there is no empty slot
- if(debug) fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: extending fd array to %d entries\n", all_files_size + FILE_DESCRIPTORS_INCREASE_STEP);
+
all_files = realloc(all_files, (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP) * sizeof(struct file_descriptor));
// if the address changed, we have to rebuild the index
// since all pointers are now invalid
if(old && old != (void *)all_files) {
- if(debug) fprintf(stderr, "apps.plugin: >> re-indexing.\n");
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> re-indexing.\n");
+
all_files_index.root = NULL;
for(i = 0; i < all_files_size; i++) {
if(!all_files[i].count) continue;
file_descriptor_add(&all_files[i]);
}
- if(debug) fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> re-indexing done.\n");
}
for(i = all_files_size; i < (all_files_size + FILE_DESCRIPTORS_INCREASE_STEP); i++) {
all_files_size += FILE_DESCRIPTORS_INCREASE_STEP;
}
- if(debug) fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> searching for empty slot.\n");
// search for an empty slot
int i, c;
if(c == 0) continue;
if(!all_files[c].count) {
- if(debug) fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> Examining slot %d.\n", c);
#ifdef NETDATA_INTERNAL_CHECKS
if(all_files[c].magic == 0x0BADCAFE && all_files[c].name && file_descriptor_find(all_files[c].name, all_files[c].hash))
error("fd on position %d is not cleared properly. It still has %s in it.\n", c, all_files[c].name);
#endif /* NETDATA_INTERNAL_CHECKS */
- if(debug) 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);
+ if(unlikely(debug))
+ 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);
+
if(all_files[c].name) free((void *)all_files[c].name);
all_files[c].name = NULL;
last_pos = c;
fatal("We should find an empty slot, but there isn't any");
exit(1);
}
- if(debug) fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: >> updating slot %d.\n", c);
all_files_len++;
else if(strcmp(name, "anon_inode:[timerfd]") == 0) type = FILETYPE_TIMERFD;
else if(strcmp(name, "anon_inode:[signalfd]") == 0) type = FILETYPE_SIGNALFD;
else if(strncmp(name, "anon_inode:", 11) == 0) {
- if(debug) fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: FIXME: unknown anonymous inode: %s\n", name);
+
type = FILETYPE_OTHER;
}
else {
- if(debug) fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: FIXME: cannot understand linkname: %s\n", name);
+
type = FILETYPE_OTHER;
}
#endif /* NETDATA_INTERNAL_CHECKS */
file_descriptor_add(&all_files[c]);
- if(debug) fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: using fd position %d (name: %s)\n", c, all_files[c].name);
return c;
}
int read_pid_file_descriptors(struct pid_stat *p) {
char dirname[FILENAME_MAX+1];
- snprintf(dirname, FILENAME_MAX, "%s/proc/%d/fd", host_prefix, p->pid);
+ snprintfz(dirname, FILENAME_MAX, "%s/proc/%d/fd", host_prefix, p->pid);
DIR *fds = opendir(dirname);
if(fds) {
int c;
if(fdid < 0) continue;
if(fdid >= p->fds_size) {
// it is small, extend it
- if(debug) fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: extending fd memory slots for %s from %d to %d\n", p->comm, p->fds_size, fdid + 100);
+
p->fds = realloc(p->fds, (fdid + 100) * sizeof(int));
if(!p->fds) {
- error("Cannot re-allocate fds for %s", p->comm);
+ fatal("Cannot re-allocate fds for %s", p->comm);
break;
}
{
char dirname[FILENAME_MAX + 1];
- snprintf(dirname, FILENAME_MAX, "%s/proc", host_prefix);
+ snprintfz(dirname, FILENAME_MAX, "%s/proc", host_prefix);
DIR *dir = opendir(dirname);
if(!dir) return 0;
all_pids_count = 0;
for(p = root_of_pids; p ; p = p->next) {
all_pids_count++;
- p->parent = NULL;
- p->updated = 0;
- p->children_count = 0;
- p->merged = 0;
- p->new_entry = 0;
-
- p->last_minflt = p->minflt;
- p->last_cminflt = p->cminflt;
- p->last_majflt = p->majflt;
- p->last_cmajflt = p->cmajflt;
- p->last_utime = p->utime;
- p->last_stime = p->stime;
- p->last_cutime = p->cutime;
- p->last_cstime = p->cstime;
-
- p->last_io_logical_bytes_read = p->io_logical_bytes_read;
+
+ p->parent = NULL;
+
+ p->updated = 0;
+ p->children_count = 0;
+ p->merged = 0;
+ p->new_entry = 0;
+
+ p->last_minflt = p->minflt;
+ p->last_majflt = p->majflt;
+ p->last_utime = p->utime;
+ p->last_stime = p->stime;
+
+ p->last_cminflt = p->cminflt;
+ p->last_cmajflt = p->cmajflt;
+ p->last_cutime = p->cutime;
+ p->last_cstime = p->cstime;
+
+ p->last_fix_cminflt = p->fix_cminflt;
+ p->last_fix_cmajflt = p->fix_cmajflt;
+ p->last_fix_cutime = p->fix_cutime;
+ p->last_fix_cstime = p->fix_cstime;
+
+ p->last_io_logical_bytes_read = p->io_logical_bytes_read;
p->last_io_logical_bytes_written = p->io_logical_bytes_written;
- p->last_io_read_calls = p->io_read_calls;
- p->last_io_write_calls = p->io_write_calls;
- p->last_io_storage_bytes_read = p->io_storage_bytes_read;
+ p->last_io_read_calls = p->io_read_calls;
+ p->last_io_write_calls = p->io_write_calls;
+ p->last_io_storage_bytes_read = p->io_storage_bytes_read;
p->last_io_storage_bytes_written = p->io_storage_bytes_written;
p->last_io_cancelled_write_bytes = p->io_cancelled_write_bytes;
}
pid_t pid = (pid_t) strtoul(file->d_name, &endptr, 10);
// make sure we read a valid number
- if(unlikely(pid <= 0 || pid > pid_max || endptr == file->d_name || *endptr != '\0'))
+ if(unlikely(endptr == file->d_name || *endptr != '\0'))
+ continue;
+
+ if(unlikely(pid <= 0 || pid > pid_max)) {
+ error("Invalid pid %d read (expected 1 to %d). Ignoring process.", pid, pid_max);
continue;
+ }
p = get_pid_entry(pid);
if(unlikely(!p)) continue;
// /proc/<pid>/stat
if(unlikely(read_proc_pid_stat(p))) {
- error("Cannot process %s/proc/%d/stat", host_prefix, pid);
-
+ error("Cannot process %s/proc/%d/stat", host_prefix, pid);
// there is no reason to proceed if we cannot get its status
continue;
}
// check its parent pid
if(unlikely(p->ppid < 0 || p->ppid > pid_max)) {
- error("Pid %d states invalid parent pid %d. Using 0.", pid, p->ppid);
-
+ error("Pid %d states invalid parent pid %d. Using 0.", pid, p->ppid);
p->ppid = 0;
}
- // --------------------------------------------------------------------
- // /proc/<pid>/cmdline
-
- if(proc_pid_cmdline_is_needed) {
- if(unlikely(read_proc_pid_cmdline(p))) {
- error("Cannot process %s/proc/%d/cmdline", host_prefix, pid);
- }
- }
-
// --------------------------------------------------------------------
// /proc/<pid>/statm
if(unlikely(read_proc_pid_statm(p))) {
- error("Cannot process %s/proc/%d/statm", host_prefix, pid);
-
+ error("Cannot process %s/proc/%d/statm", host_prefix, pid);
// there is no reason to proceed if we cannot get its memory status
continue;
}
// check if it is target
// we do this only once, the first time this pid is loaded
if(unlikely(p->new_entry)) {
- if(debug) fprintf(stderr, "apps.plugin: \tJust added %s\n", p->comm);
+ // /proc/<pid>/cmdline
+ if(proc_pid_cmdline_is_needed) {
+ if(unlikely(read_proc_pid_cmdline(p))) {
+ error("Cannot process %s/proc/%d/cmdline", host_prefix, pid);
+ }
+ }
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: \tJust added %d (%s)\n", pid, p->comm);
+
uint32_t hash = simple_hash(p->comm);
- size_t pclen = strlen(p->comm);
+ size_t pclen = strlen(p->comm);
struct target *w;
for(w = apps_groups_root_target; w ; w = w->next) {
if(debug || (p->target && p->target->debug))
fprintf(stderr, "apps.plugin: \t\t%s linked to target %s\n", p->comm, p->target->name);
+
+ break;
}
}
}
return 1;
}
-
-// ----------------------------------------------------------------------------
-
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
-// print a tree view of all processes
-int debug_childrens_aggregations(pid_t pid, int level) {
- struct pid_stat *p = NULL;
- char b[level+3];
- int i, ret = 0;
-
- for(i = 0; i < level; i++) b[i] = '\t';
- b[level] = '|';
- b[level+1] = '-';
- b[level+2] = '\0';
-
- for(p = root_of_pids; p ; p = p->next) {
- if(p->ppid == pid) {
- ret += debug_childrens_aggregations(p->pid, level+1);
- }
- }
-
- p = all_pids[pid];
- if(p) {
- if(!p->updated) ret += 1;
- if(ret) fprintf(stderr, "%s %s %d [%s, %s] c=%d u=%llu+%llu, s=%llu+%llu, cu=%llu+%llu, cs=%llu+%llu, n=%llu+%llu, j=%llu+%llu, cn=%llu+%llu, cj=%llu+%llu\n"
- , b, p->comm, p->pid, p->updated?"OK":"KILLED", p->target->name, p->children_count
- , p->utime, p->utime - p->old_utime
- , p->stime, p->stime - p->old_stime
- , p->cutime, p->cutime - p->old_cutime
- , p->cstime, p->cstime - p->old_cstime
- , p->minflt, p->minflt - p->old_minflt
- , p->majflt, p->majflt - p->old_majflt
- , p->cminflt, p->cminflt - p->old_cminflt
- , p->cmajflt, p->cmajflt - p->old_cmajflt
- );
- }
-
- return ret;
-}
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
-
-
// ----------------------------------------------------------------------------
// update statistics on the targets
// check: update_apps_groups_statistics()
void link_all_processes_to_their_parents(void) {
+ struct pid_stat *init = all_pids[1];
struct pid_stat *p = NULL;
// link all children to their parents
for(p = root_of_pids; p ; p = p->next) {
// for each process found running
- if(p->ppid > 0
- && p->ppid <= pid_max
- && all_pids[p->ppid]
- ) {
- // for valid processes
+ if(likely(p->new_entry && p->updated)) {
+ // the first time we see an entry
+ // we remove the exited children figures
+ // to avoid spikes
+ p->fix_cminflt = p->cminflt;
+ p->fix_cmajflt = p->cmajflt;
+ p->fix_cutime = p->cutime;
+ p->fix_cstime = p->cstime;
+ }
+
+ if(likely(p->ppid > 0 && all_pids[p->ppid])) {
+ // valid parent processes
- if(debug || (p->target && p->target->debug))
- fprintf(stderr, "apps.plugin: \tparent of %d (%s) is %d (%s)\n", p->pid, p->comm, p->ppid, all_pids[p->ppid]->comm);
+ struct pid_stat *pp;
- p->parent = all_pids[p->ppid];
+ p->parent = pp = all_pids[p->ppid];
p->parent->children_count++;
- }
- else if(p->ppid != 0)
- error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
- }
-}
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
-void aggregate_children_to_parents(void) {
- struct pid_stat *p = NULL;
+ if(unlikely(debug || (p->target && p->target->debug)))
+ 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);
- // for each killed process, remove its values from the parents
- // sums (we had already added them in a previous loop)
- for(p = root_of_pids; p ; p = p->next) {
- if(p->updated) continue;
-
- if(debug) fprintf(stderr, "apps.plugin: UNMERGING %d %s\n", p->pid, p->comm);
-
- unsigned long long diff_utime = p->utime + p->cutime + p->fix_cutime;
- unsigned long long diff_stime = p->stime + p->cstime + p->fix_cstime;
- unsigned long long diff_minflt = p->minflt + p->cminflt + p->fix_cminflt;
- unsigned long long diff_majflt = p->majflt + p->cmajflt + p->fix_cmajflt;
-
- struct pid_stat *t = p;
- while((t = t->parent)) {
- if(!t->updated) continue;
-
- unsigned long long x;
- if(diff_utime && t->diff_cutime) {
- x = (t->diff_cutime < diff_utime)?t->diff_cutime:diff_utime;
- diff_utime -= x;
- t->diff_cutime -= x;
- t->fix_cutime += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cutime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
- }
- if(diff_stime && t->diff_cstime) {
- x = (t->diff_cstime < diff_stime)?t->diff_cstime:diff_stime;
- diff_stime -= x;
- t->diff_cstime -= x;
- t->fix_cstime += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cstime %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
- }
- if(diff_minflt && t->diff_cminflt) {
- x = (t->diff_cminflt < diff_minflt)?t->diff_cminflt:diff_minflt;
- diff_minflt -= x;
- t->diff_cminflt -= x;
- t->fix_cminflt += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cminflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
- }
- if(diff_majflt && t->diff_cmajflt) {
- x = (t->diff_cmajflt < diff_majflt)?t->diff_cmajflt:diff_majflt;
- diff_majflt -= x;
- t->diff_cmajflt -= x;
- t->fix_cmajflt += x;
- if(debug) fprintf(stderr, "apps.plugin: \t cmajflt %llu from %d %s %s\n", x, t->pid, t->comm, t->target->name);
+ if(unlikely(!p->updated)) {
+ // this process has exit
+
+ // find the first parent that has been updated
+ while(pp && !pp->updated) {
+ // we may have to forward link it to its parent
+ if(unlikely(!pp->parent && pp->ppid > 0 && all_pids[pp->ppid]))
+ pp->parent = all_pids[pp->ppid];
+
+ // check again for parent
+ pp = pp->parent;
+ }
+
+ if(likely(pp)) {
+ // this is an exited child with a parent
+ // remove the known time from the parent's data
+ pp->fix_cminflt += p->last_minflt + p->last_cminflt + p->last_fix_cminflt;
+ pp->fix_cmajflt += p->last_majflt + p->last_cmajflt + p->last_fix_cmajflt;
+ pp->fix_cutime += p->last_utime + p->last_cutime + p->last_fix_cutime;
+ pp->fix_cstime += p->last_stime + p->last_cstime + p->last_fix_cstime;
+
+ // The known exited children (the ones we track) may have
+ // contributed more than the value accumulated into the process
+ // by the kernel.
+ // This can happen if the parent process has not waited-for
+ // its children (check: man 2 times).
+ // In this case, the kernel adds these resources to init (pid 1).
+ //
+ // The following code, attempts to fix this.
+ // Without this code, the charts will have random spikes
+ // for example, when an SSH session ends (sshd forks a child
+ // to serve the session, but when this session ends, sshd
+ // does not wait-for its child, thus all the resources of the
+ // ssh session get added to init, resulting in a huge spike on
+ // the charts).
+
+ if(unlikely(pp->cminflt < pp->fix_cminflt)) {
+ if(likely(init && pp != init)) {
+ unsigned long long have = pp->fix_cminflt - pp->cminflt;
+ unsigned long long max = init->cminflt - init->fix_cminflt;
+ if(have > max) have = max;
+ init->fix_cminflt += have;
+ }
+ pp->fix_cminflt = pp->cminflt;
+ }
+ if(unlikely(pp->cmajflt < pp->fix_cmajflt)) {
+ if(likely(init && pp != init)) {
+ unsigned long long have = pp->fix_cmajflt - pp->cmajflt;
+ unsigned long long max = init->cmajflt - init->fix_cmajflt;
+ if(have > max) have = max;
+ init->fix_cmajflt += have;
+ }
+ pp->fix_cmajflt = pp->cmajflt;
+ }
+ if(unlikely(pp->cutime < pp->fix_cutime)) {
+ if(likely(init && pp != init)) {
+ unsigned long long have = pp->fix_cutime - pp->cutime;
+ unsigned long long max = init->cutime - init->fix_cutime;
+ if(have > max) have = max;
+ init->fix_cutime += have;
+ }
+ pp->fix_cutime = pp->cutime;
+ }
+ if(unlikely(pp->cstime < pp->fix_cstime)) {
+ if(likely(init && pp != init)) {
+ unsigned long long have = pp->fix_cstime - pp->cstime;
+ unsigned long long max = init->cstime - init->fix_cstime;
+ if(have > max) have = max;
+ init->fix_cstime += have;
+ }
+ pp->fix_cstime = pp->cstime;
+ }
+
+ if(unlikely(debug))
+ 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);
+ }
}
}
-
- if(diff_utime) error("Cannot fix up utime %llu", diff_utime);
- if(diff_stime) error("Cannot fix up stime %llu", diff_stime);
- if(diff_minflt) error("Cannot fix up minflt %llu", diff_minflt);
- if(diff_majflt) error("Cannot fix up majflt %llu", diff_majflt);
+ else if(unlikely(p->ppid != 0))
+ error("pid %d %s states parent %d, but the later does not exist.", p->pid, p->comm, p->ppid);
}
}
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
+
void cleanup_non_existing_pids(void) {
int c;
// children that do not have a target
// inherit their target from their parent
- int found = 1;
+ int found = 1, loops = 0;
while(found) {
+ if(unlikely(debug)) loops++;
found = 0;
for(p = root_of_pids; p ; p = p->next) {
// if this process does not have a target
// repeat, until nothing more can be done.
found = 1;
while(found) {
+ if(unlikely(debug)) loops++;
found = 0;
for(p = root_of_pids; p ; p = p->next) {
// if this process does not have any children
}
}
- if(debug)
+ if(unlikely(debug))
fprintf(stderr, "apps.plugin: merged %d processes\n", found);
}
all_pids[1]->target = apps_groups_default_target;
// give a default target on all top level processes
+ if(unlikely(debug)) loops++;
for(p = root_of_pids; p ; p = p->next) {
// if the process is not merged itself
// then is is a top level process
if(!p->merged && !p->target)
p->target = apps_groups_default_target;
-
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- // by the way, update the diffs
- // will be used later for subtracting killed process times
- p->diff_cutime = p->utime - p->cutime;
- p->diff_cstime = p->stime - p->cstime;
- p->diff_cminflt = p->minflt - p->cminflt;
- p->diff_cmajflt = p->majflt - p->cmajflt;
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
}
// give a target to all merged child processes
found = 1;
while(found) {
+ if(unlikely(debug)) loops++;
found = 0;
for(p = root_of_pids; p ; p = p->next) {
if(unlikely(!p->target && p->merged && p->parent && p->parent->target)) {
}
}
}
+
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: apply_apps_groups_targets_inheritance() made %d loops on the process tree\n", loops);
}
long zero_all_targets(struct target *root) {
}
if(likely(p->updated)) {
- w->cutime += p->cutime; // - p->fix_cutime;
- w->cstime += p->cstime; // - p->fix_cstime;
- w->cminflt += p->cminflt; // - p->fix_cminflt;
- w->cmajflt += p->cmajflt; // - p->fix_cmajflt;
+ if(unlikely(debug && (p->fix_cutime || p->fix_cstime || p->fix_cminflt || p->fix_cmajflt)))
+ 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);
+
+ w->cutime += p->cutime - p->fix_cutime;
+ w->cstime += p->cstime - p->fix_cstime;
+ w->cminflt += p->cminflt - p->fix_cminflt;
+ w->cmajflt += p->cmajflt - p->fix_cmajflt;
w->utime += p->utime; //+ (p->pid != 1)?(p->cutime - p->fix_cutime):0;
w->stime += p->stime; //+ (p->pid != 1)?(p->cstime - p->fix_cstime):0;
}
if(unlikely(debug || w->debug))
- fprintf(stderr, "apps.plugin: \tAgregating %s pid %d on %s utime=%llu, stime=%llu, cutime=%llu, cstime=%llu, minflt=%llu, majflt=%llu, cminflt=%llu, 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);
+ 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);
/* 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);
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);
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);
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);
*/
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- p->old_utime = p->utime;
- p->old_cutime = p->cutime;
- p->old_stime = p->stime;
- p->old_cstime = p->cstime;
- p->old_minflt = p->minflt;
- p->old_majflt = p->majflt;
- p->old_cminflt = p->cminflt;
- p->old_cmajflt = p->cmajflt;
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
if(o) {
// since the process switched target
// IMPORTANT
// We add/subtract the last/OLD values we added to the target
- w->fix_cutime -= p->last_cutime;
- w->fix_cstime -= p->last_cstime;
- w->fix_cminflt -= p->last_cminflt;
- w->fix_cmajflt -= p->last_cmajflt;
+ unsigned long long cutime = p->last_cutime - p->last_fix_cutime;
+ unsigned long long cstime = p->last_cstime - p->last_fix_cstime;
+ unsigned long long cminflt = p->last_cminflt - p->last_fix_cminflt;
+ unsigned long long cmajflt = p->last_cmajflt - p->last_fix_cmajflt;
+
+ w->fix_cutime -= cutime;
+ w->fix_cstime -= cstime;
+ w->fix_cminflt -= cminflt;
+ w->fix_cmajflt -= cmajflt;
- w->fix_utime -= p->last_utime;
- w->fix_stime -= p->last_stime;
+ w->fix_utime -= p->last_utime;
+ w->fix_stime -= p->last_stime;
w->fix_minflt -= p->last_minflt;
w->fix_majflt -= p->last_majflt;
-
- w->fix_io_logical_bytes_read -= p->last_io_logical_bytes_read;
+ w->fix_io_logical_bytes_read -= p->last_io_logical_bytes_read;
w->fix_io_logical_bytes_written -= p->last_io_logical_bytes_written;
- w->fix_io_read_calls -= p->last_io_read_calls;
- w->fix_io_write_calls -= p->last_io_write_calls;
- w->fix_io_storage_bytes_read -= p->last_io_storage_bytes_read;
+ w->fix_io_read_calls -= p->last_io_read_calls;
+ w->fix_io_write_calls -= p->last_io_write_calls;
+ w->fix_io_storage_bytes_read -= p->last_io_storage_bytes_read;
w->fix_io_storage_bytes_written -= p->last_io_storage_bytes_written;
w->fix_io_cancelled_write_bytes -= p->last_io_cancelled_write_bytes;
// ---
- o->fix_cutime += p->last_cutime;
- o->fix_cstime += p->last_cstime;
- o->fix_cminflt += p->last_cminflt;
- o->fix_cmajflt += p->last_cmajflt;
+ o->fix_cutime += cutime;
+ o->fix_cstime += cstime;
+ o->fix_cminflt += cminflt;
+ o->fix_cmajflt += cmajflt;
- o->fix_utime += p->last_utime;
- o->fix_stime += p->last_stime;
+ o->fix_utime += p->last_utime;
+ o->fix_stime += p->last_stime;
o->fix_minflt += p->last_minflt;
o->fix_majflt += p->last_majflt;
- o->fix_io_logical_bytes_read += p->last_io_logical_bytes_read;
+ o->fix_io_logical_bytes_read += p->last_io_logical_bytes_read;
o->fix_io_logical_bytes_written += p->last_io_logical_bytes_written;
- o->fix_io_read_calls += p->last_io_read_calls;
- o->fix_io_write_calls += p->last_io_write_calls;
- o->fix_io_storage_bytes_read += p->last_io_storage_bytes_read;
+ o->fix_io_read_calls += p->last_io_read_calls;
+ o->fix_io_write_calls += p->last_io_write_calls;
+ o->fix_io_storage_bytes_read += p->last_io_storage_bytes_read;
o->fix_io_storage_bytes_written += p->last_io_storage_bytes_written;
o->fix_io_cancelled_write_bytes += p->last_io_cancelled_write_bytes;
}
// since the process has exited, the user
// will see a drop in our charts, because the incremental
- // values of this process will not be there
+ // values of this process will not be there from now on
// add them to the fix_* values and they will be added to
// the reported values, so that the report goes steady
- w->fix_minflt += p->minflt;
- w->fix_majflt += p->majflt;
- w->fix_utime += p->utime;
- w->fix_stime += p->stime;
- w->fix_cminflt += p->cminflt;
- w->fix_cmajflt += p->cmajflt;
- w->fix_cutime += p->cutime;
- w->fix_cstime += p->cstime;
-
- w->fix_io_logical_bytes_read += p->io_logical_bytes_read;
- w->fix_io_logical_bytes_written += p->io_logical_bytes_written;
- w->fix_io_read_calls += p->io_read_calls;
- w->fix_io_write_calls += p->io_write_calls;
- w->fix_io_storage_bytes_read += p->io_storage_bytes_read;
- w->fix_io_storage_bytes_written += p->io_storage_bytes_written;
- w->fix_io_cancelled_write_bytes += p->io_cancelled_write_bytes;
+
+ w->fix_minflt += p->last_minflt;
+ w->fix_majflt += p->last_majflt;
+ w->fix_utime += p->last_utime;
+ w->fix_stime += p->last_stime;
+
+ w->fix_cminflt += (p->last_cminflt - p->last_fix_cminflt);
+ w->fix_cmajflt += (p->last_cmajflt - p->last_fix_cmajflt);
+ w->fix_cutime += (p->last_cutime - p->last_fix_cutime);
+ w->fix_cstime += (p->last_cstime - p->last_fix_cstime);
+
+ w->fix_io_logical_bytes_read += p->last_io_logical_bytes_read;
+ w->fix_io_logical_bytes_written += p->last_io_logical_bytes_written;
+ w->fix_io_read_calls += p->last_io_read_calls;
+ w->fix_io_write_calls += p->last_io_write_calls;
+ w->fix_io_storage_bytes_read += p->last_io_storage_bytes_read;
+ w->fix_io_storage_bytes_written += p->last_io_storage_bytes_written;
+ w->fix_io_cancelled_write_bytes += p->last_io_cancelled_write_bytes;
}
+ //if((long long)w->cutime + w->fix_cutime < 0)
+ // 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",
+ // 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);
}
void count_targets_fds(struct target *root) {
link_all_processes_to_their_parents();
apply_apps_groups_targets_inheritance();
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- aggregate_children_to_parents();
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
zero_all_targets(users_root_target);
zero_all_targets(groups_root_target);
apps_groups_targets = zero_all_targets(apps_groups_root_target);
-#ifdef AGGREGATE_CHILDREN_TO_PARENTS
- if(debug)
- debug_childrens_aggregations(0, 1);
-#endif /* AGGREGATE_CHILDREN_TO_PARENTS */
-
// this has to be done, before the cleanup
struct pid_stat *p = NULL;
struct target *w = NULL, *o = NULL;
w = p->user_target;
else {
if(unlikely(debug && p->user_target))
- fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched user from %d (%s) to %d.\n", p->pid, p->comm, p->user_target->uid, p->user_target->name, p->uid);
+ 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);
w = p->user_target = get_users_target(p->uid);
}
w = p->group_target;
else {
if(unlikely(debug && p->group_target))
- fprintf(stderr, "apps.plugin: \t\tpid %d (%s) switched group from %d (%s) to %d.\n", p->pid, p->comm, p->group_target->gid, p->group_target->name, p->gid);
+ 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);
w = p->group_target = get_groups_target(p->gid);
}
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->stime + w->fix_utime + w->fix_stime);
+ 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));
}
fprintf(stdout, "END\n");
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->fix_utime);
+ fprintf(stdout, "SET %s = %llu\n", w->name, w->utime + w->fix_utime + (include_exited_childs?(w->cutime + w->fix_cutime):0));
}
fprintf(stdout, "END\n");
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "SET %s = %llu\n", w->name, w->stime + w->fix_stime);
+ fprintf(stdout, "SET %s = %llu\n", w->name, w->stime + w->fix_stime + (include_exited_childs?(w->cstime + w->fix_cstime):0));
}
fprintf(stdout, "END\n");
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "SET %s = %llu\n", w->name, w->minflt + w->fix_minflt);
+ fprintf(stdout, "SET %s = %llu\n", w->name, w->minflt + w->fix_minflt + (include_exited_childs?(w->cminflt + w->fix_cminflt):0));
}
fprintf(stdout, "END\n");
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "SET %s = %llu\n", w->name, w->majflt + w->fix_majflt);
+ fprintf(stdout, "SET %s = %llu\n", w->name, w->majflt + w->fix_majflt + (include_exited_childs?(w->cmajflt + w->fix_cmajflt):0));
}
fprintf(stdout, "END\n");
// we have something new to show
// update the charts
- fprintf(stdout, "CHART %s.cpu '' '%s CPU Time (%ld%% = %ld core%s)' 'cpu time %%' cpu %s.cpu stacked 20001 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
+ 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);
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "DIMENSION %s '' incremental 100 %llu %s\n", w->name, Hertz, w->hidden ? "hidden,noreset" : "noreset");
+ fprintf(stdout, "DIMENSION %s '' incremental 100 %u %s\n", w->name, hz, w->hidden ? "hidden,noreset" : "noreset");
}
fprintf(stdout, "CHART %s.mem '' '%s Dedicated Memory (w/o shared)' 'MB' mem %s.mem stacked 20003 %d\n", type, title, type, update_every);
fprintf(stdout, "DIMENSION %s '' absolute 1 1 noreset\n", w->name);
}
- fprintf(stdout, "CHART %s.cpu_user '' '%s CPU User Time (%ld%% = %ld core%s)' 'cpu time %%' cpu %s.cpu_user stacked 20020 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
+ 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);
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "DIMENSION %s '' incremental 100 %llu noreset\n", w->name, Hertz * processors);
+ fprintf(stdout, "DIMENSION %s '' incremental 100 %u noreset\n", w->name, hz);
}
- fprintf(stdout, "CHART %s.cpu_system '' '%s CPU System Time (%ld%% = %ld core%s)' 'cpu time %%' cpu %s.cpu_system stacked 20021 %d\n", type, title, (processors * 100), processors, (processors>1)?"s":"", type, update_every);
+ 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);
for (w = root; w ; w = w->next) {
if(w->target || (!w->processes && !w->exposed)) continue;
- fprintf(stdout, "DIMENSION %s '' incremental 100 %llu noreset\n", w->name, Hertz * processors);
+ fprintf(stdout, "DIMENSION %s '' incremental 100 %u noreset\n", w->name, hz);
}
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);
continue;
}
+ if(strcmp("no-childs", argv[i]) == 0) {
+ include_exited_childs = 0;
+ continue;
+ }
+
+ if(strcmp("with-childs", argv[i]) == 0) {
+ include_exited_childs = 1;
+ continue;
+ }
+
if(!name) {
name = argv[i];
continue;
}
}
-unsigned long long sutime() {
- struct timeval now;
- gettimeofday(&now, NULL);
- return now.tv_sec * 1000000ULL + now.tv_usec;
-}
-
int main(int argc, char **argv)
{
// debug_flags = D_PROCFILE;
// disable syslog for apps.plugin
error_log_syslog = 0;
+ // set errors flood protection to 100 logs per hour
+ error_log_errors_per_period = 100;
+ error_log_throttle_period = 3600;
+
host_prefix = getenv("NETDATA_HOST_PREFIX");
if(host_prefix == NULL) {
info("NETDATA_HOST_PREFIX is not passed from netdata");
}
else info("Found NETDATA_CONFIG_DIR='%s'", config_dir);
- info("starting...");
+#ifdef NETDATA_INTERNAL_CHECKS
+ if(debug_flags != 0) {
+ struct rlimit rl = { RLIM_INFINITY, RLIM_INFINITY };
+ if(setrlimit(RLIMIT_CORE, &rl) != 0)
+ info("Cannot request unlimited core dumps for debugging... Proceeding anyway...");
+ prctl(PR_SET_DUMPABLE, 1, 0, 0, 0);
+ }
+#endif /* NETDATA_INTERNAL_CHECKS */
procfile_adaptive_initial_allocation = 1;
time_t started_t = time(NULL);
time_t current_t;
- Hertz = get_system_hertz();
+ get_HZ();
pid_max = get_system_pid_max();
processors = get_system_cpus();
exit(1);
}
- fprintf(stdout, "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %d\n", update_every);
- fprintf(stdout, "DIMENSION user '' incremental 1 %d\n", 1000);
- fprintf(stdout, "DIMENSION system '' incremental 1 %d\n", 1000);
-
- fprintf(stdout, "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %d\n", update_every);
- fprintf(stdout, "DIMENSION files '' incremental 1 1\n");
- fprintf(stdout, "DIMENSION pids '' absolute 1 1\n");
- fprintf(stdout, "DIMENSION fds '' absolute 1 1\n");
- fprintf(stdout, "DIMENSION targets '' absolute 1 1\n");
-
+ fprintf(stdout, "CHART netdata.apps_cpu '' 'Apps Plugin CPU' 'milliseconds/s' apps.plugin netdata.apps_cpu stacked 140000 %1$d\n"
+ "DIMENSION user '' incremental 1 1000\n"
+ "DIMENSION system '' incremental 1 1000\n"
+ "CHART netdata.apps_files '' 'Apps Plugin Files' 'files/s' apps.plugin netdata.apps_files line 140001 %1$d\n"
+ "DIMENSION files '' incremental 1 1\n"
+ "DIMENSION pids '' absolute 1 1\n"
+ "DIMENSION fds '' absolute 1 1\n"
+ "DIMENSION targets '' absolute 1 1\n", update_every);
#ifndef PROFILING_MODE
unsigned long long sunext = (time(NULL) - (time(NULL) % update_every) + update_every) * 1000000ULL;
for(;1; counter++) {
#ifndef PROFILING_MODE
// delay until it is our time to run
- while((sunow = sutime()) < sunext)
- usleep((useconds_t)(sunext - sunow));
+ while((sunow = timems()) < sunext)
+ usecsleep(sunext - sunow);
// find the next time we need to run
- while(sutime() > sunext)
+ while(timems() > sunext)
sunext += update_every * 1000000ULL;
#endif /* PROFILING_MODE */
send_collected_data_to_netdata(users_root_target, "users", dt);
send_collected_data_to_netdata(groups_root_target, "groups", dt);
- if(debug) fprintf(stderr, "apps.plugin: done Loop No %llu\n", counter);
- fflush(NULL);
+ if(unlikely(debug))
+ fprintf(stderr, "apps.plugin: done Loop No %llu\n", counter);
current_t = time(NULL);
projects / netdata.git / blobdiff