void rrdset_next(RRDSET *st)
{
unsigned long long microseconds = 0;
if(likely(st->last_collected_time.tv_sec)) {
struct timeval now;
gettimeofday(&now, NULL);
microseconds = usecdiff(&now, &st->last_collected_time);
}
// prevent infinite loop
else microseconds = st->update_every * 1000000ULL;
rrdset_next_usec(st, microseconds);
}
int unit_test(long delay, long shift)
{
static int repeat = 0;
repeat++;
char name[101];
snprintfz(name, 100, "unittest-%d-%ld-%ld", repeat, delay, shift);
//debug_flags = 0xffffffff;
rrd_memory_mode = RRD_MEMORY_MODE_RAM;
rrd_update_every = 1;
int do_abs = 1;
int do_inc = 1;
int do_abst = 0;
int do_absi = 0;
RRDSET *st = rrdset_create("netdata", name, name, "netdata", NULL, "Unit Testing", "a value", 1, 1, RRDSET_TYPE_LINE);
st->debug = 1;
RRDDIM *rdabs = NULL;
RRDDIM *rdinc = NULL;
RRDDIM *rdabst = NULL;
RRDDIM *rdabsi = NULL;
if(do_abs) rdabs = rrddim_add(st, "absolute", "absolute", 1, 1, RRDDIM_ABSOLUTE);
if(do_inc) rdinc = rrddim_add(st, "incremental", "incremental", 1, 1, RRDDIM_INCREMENTAL);
if(do_abst) rdabst = rrddim_add(st, "percentage-of-absolute-row", "percentage-of-absolute-row", 1, 1, RRDDIM_PCENT_OVER_ROW_TOTAL);
if(do_absi) rdabsi = rrddim_add(st, "percentage-of-incremental-row", "percentage-of-incremental-row", 1, 1, RRDDIM_PCENT_OVER_DIFF_TOTAL);
long increment = 1000;
collected_number i = 0;
unsigned long c, dimensions = 0;
RRDDIM *rd;
for(rd = st->dimensions ; rd ; rd = rd->next) dimensions++;
for(c = 0; c < 20 ;c++) {
i += increment;
fprintf(stderr, "\n\nLOOP = %lu, DELAY = %ld, VALUE = " COLLECTED_NUMBER_FORMAT "\n", c, delay, i);
if(c) {
rrdset_next_usec(st, delay);
}
if(do_abs) rrddim_set(st, "absolute", i);
if(do_inc) rrddim_set(st, "incremental", i);
if(do_abst) rrddim_set(st, "percentage-of-absolute-row", i);
if(do_absi) rrddim_set(st, "percentage-of-incremental-row", i);
if(!c) {
gettimeofday(&st->last_collected_time, NULL);
st->last_collected_time.tv_usec = shift;
}
// prevent it from deleting the dimensions
for(rd = st->dimensions ; rd ; rd = rd->next)
rd->last_collected_time.tv_sec = st->last_collected_time.tv_sec;
rrdset_done(st);
}
unsigned long oincrement = increment;
increment = increment * st->update_every * 1000000 / delay;
fprintf(stderr, "\n\nORIGINAL INCREMENT: %lu, INCREMENT %ld, DELAY %ld, SHIFT %ld\n", oincrement * 10, increment * 10, delay, shift);
int ret = 0;
storage_number sn;
calculated_number cn, v;
for(c = 0 ; c < st->counter ; c++) {
fprintf(stderr, "\nPOSITION: c = %lu, EXPECTED VALUE %lu\n", c, (oincrement + c * increment + increment * (1000000 - shift) / 1000000 )* 10);
for(rd = st->dimensions ; rd ; rd = rd->next) {
sn = rd->values[c];
cn = unpack_storage_number(sn);
fprintf(stderr, "\t %s " CALCULATED_NUMBER_FORMAT " (PACKED AS " STORAGE_NUMBER_FORMAT ") -> ", rd->id, cn, sn);
if(rd == rdabs) v =
( oincrement
// + (increment * (1000000 - shift) / 1000000)
+ (c + 1) * increment
);
else if(rd == rdinc) v = (c?(increment):(increment * (1000000 - shift) / 1000000));
else if(rd == rdabst) v = oincrement / dimensions / 10;
else if(rd == rdabsi) v = oincrement / dimensions / 10;
else v = 0;
if(v == cn) fprintf(stderr, "passed.\n");
else {
fprintf(stderr, "ERROR! (expected " CALCULATED_NUMBER_FORMAT ")\n", v);
ret = 1;
}
}
}
if(ret)
fprintf(stderr, "\n\nUNIT TEST(%ld, %ld) FAILED\n\n", delay, shift);
return ret;
}
int run_test(struct test *test)
{
fprintf(stderr, "\nRunning test '%s':\n%s\n", test->name, test->description);
rrd_memory_mode = RRD_MEMORY_MODE_RAM;
rrd_update_every = test->update_every;
char name[101];
snprintfz(name, 100, "unittest-%s", test->name);
// create the chart
RRDSET *st = rrdset_create("netdata", name, name, "netdata", NULL, "Unit Testing", "a value", 1, 1, RRDSET_TYPE_LINE);
RRDDIM *rd = rrddim_add(st, "dim1", NULL, test->multiplier, test->divisor, test->algorithm);
RRDDIM *rd2 = NULL;
if(test->feed2)
rd2 = rrddim_add(st, "dim2", NULL, test->multiplier, test->divisor, test->algorithm);
st->debug = 1;
// feed it with the test data
unsigned long c;
for(c = 0; c < test->feed_entries; c++) {
if(debug_flags) fprintf(stderr, "\n\n");
if(c) {
fprintf(stderr, " > %s: feeding position %lu, after %llu microseconds\n", test->name, c+1, test->feed[c].microseconds);
rrdset_next_usec(st, test->feed[c].microseconds);
}
else {
fprintf(stderr, " > %s: feeding position %lu\n", test->name, c+1);
}
fprintf(stderr, " >> %s with value " COLLECTED_NUMBER_FORMAT "\n", rd->name, test->feed[c].value);
rrddim_set(st, "dim1", test->feed[c].value);
if(rd2) {
fprintf(stderr, " >> %s with value " COLLECTED_NUMBER_FORMAT "\n", rd2->name, test->feed2[c]);
rrddim_set(st, "dim2", test->feed2[c]);
}
rrdset_done(st);
// align the first entry to second boundary
if(!c) {
fprintf(stderr, " > %s: fixing first collection time to be %llu microseconds to second boundary\n", test->name, test->feed[c].microseconds);
rd->last_collected_time.tv_usec = st->last_collected_time.tv_usec = st->last_updated.tv_usec = test->feed[c].microseconds;
}
}
// check the result
int errors = 0;
if(st->counter != test->result_entries) {
fprintf(stderr, " %s stored %lu entries, but we were expecting %lu, ### E R R O R ###\n", test->name, st->counter, test->result_entries);
errors++;
}
unsigned long max = (st->counter < test->result_entries)?st->counter:test->result_entries;
for(c = 0 ; c < max ; c++) {
calculated_number v = unpack_storage_number(rd->values[c]);
calculated_number n = test->results[c];
int same = (roundl(v * 10000000.0) == roundl(n * 10000000.0))?1:0;
fprintf(stderr, " %s/%s: checking position %lu, expecting value " CALCULATED_NUMBER_FORMAT ", found " CALCULATED_NUMBER_FORMAT ", %s\n", test->name, rd->name, c+1, n, v, (same)?"OK":"### E R R O R ###");
if(!same) errors++;
if(rd2) {
v = unpack_storage_number(rd2->values[c]);
n = test->results2[c];
same = (roundl(v * 10000000.0) == roundl(n * 10000000.0))?1:0;
fprintf(stderr, " %s/%s: checking position %lu, expecting value " CALCULATED_NUMBER_FORMAT ", found " CALCULATED_NUMBER_FORMAT ", %s\n", test->name, rd2->name, c+1, n, v, (same)?"OK":"### E R R O R ###");
if(!same) errors++;
}
}
return errors;
}
void *checks_main(void *ptr)
{
if(ptr) { ; }
info("CHECKS thread created with task id %d", gettid());
if(pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED, NULL) != 0)
error("Cannot set pthread cancel type to DEFERRED.");
if(pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL) != 0)
error("Cannot set pthread cancel state to ENABLE.");
unsigned long long usec = 0, susec = rrd_update_every * 1000000ULL, loop_usec = 0, total_susec = 0;
struct timeval now, last, loop;
RRDSET *check1, *check2, *check3, *apps_cpu = NULL;
check1 = rrdset_create("netdata", "check1", NULL, "netdata", "Caller gives microseconds", "a million !", 99999, rrd_update_every, RRDSET_TYPE_LINE);
rrddim_add(check1, "absolute", NULL, -1, 1, RRDDIM_ABSOLUTE);
rrddim_add(check1, "incremental", NULL, 1, 1, RRDDIM_INCREMENTAL);
check2 = rrdset_create("netdata", "check2", NULL, "netdata", "Netdata calcs microseconds", "a million !", 99999, rrd_update_every, RRDSET_TYPE_LINE);
rrddim_add(check2, "absolute", NULL, -1, 1, RRDDIM_ABSOLUTE);
rrddim_add(check2, "incremental", NULL, 1, 1, RRDDIM_INCREMENTAL);
check3 = rrdset_create("netdata", "checkdt", NULL, "netdata", "Clock difference", "microseconds diff", 99999, rrd_update_every, RRDSET_TYPE_LINE);
rrddim_add(check3, "caller", NULL, 1, 1, RRDDIM_ABSOLUTE);
rrddim_add(check3, "netdata", NULL, 1, 1, RRDDIM_ABSOLUTE);
rrddim_add(check3, "apps.plugin", NULL, 1, 1, RRDDIM_ABSOLUTE);
gettimeofday(&last, NULL);
while(1) {
usleep(susec);
// find the time to sleep in order to wait exactly update_every seconds
gettimeofday(&now, NULL);
loop_usec = usecdiff(&now, &last);
usec = loop_usec - susec;
debug(D_PROCNETDEV_LOOP, "CHECK: last loop took %llu usec (worked for %llu, sleeped for %llu).", loop_usec, usec, susec);
if(usec < (rrd_update_every * 1000000ULL / 2ULL)) susec = (rrd_update_every * 1000000ULL) - usec;
else susec = rrd_update_every * 1000000ULL / 2ULL;
// --------------------------------------------------------------------
// Calculate loop time
last.tv_sec = now.tv_sec;
last.tv_usec = now.tv_usec;
total_susec += loop_usec;
// --------------------------------------------------------------------
// check chart 1
if(check1->counter_done) rrdset_next_usec(check1, loop_usec);
rrddim_set(check1, "absolute", 1000000);
rrddim_set(check1, "incremental", total_susec);
rrdset_done(check1);
// --------------------------------------------------------------------
// check chart 2
if(check2->counter_done) rrdset_next(check2);
rrddim_set(check2, "absolute", 1000000);
rrddim_set(check2, "incremental", total_susec);
rrdset_done(check2);
// --------------------------------------------------------------------
// check chart 3
if(!apps_cpu) apps_cpu = rrdset_find("apps.cpu");
if(check3->counter_done) rrdset_next_usec(check3, loop_usec);
gettimeofday(&loop, NULL);
rrddim_set(check3, "caller", (long long)usecdiff(&loop, &check1->last_collected_time));
rrddim_set(check3, "netdata", (long long)usecdiff(&loop, &check2->last_collected_time));
if(apps_cpu) rrddim_set(check3, "apps.plugin", (long long)usecdiff(&loop, &apps_cpu->last_collected_time));
rrdset_done(check3);
}
return NULL;
}
void *pluginsd_worker_thread(void *arg)
{
struct plugind *cd = (struct plugind *)arg;
char line[PLUGINSD_LINE_MAX + 1];
#ifdef DETACH_PLUGINS_FROM_NETDATA
unsigned long long usec = 0, susec = 0;
struct timeval last = {0, 0} , now = {0, 0};
#endif
char *words[MAX_WORDS] = { NULL };
uint32_t SET_HASH = simple_hash("SET");
uint32_t BEGIN_HASH = simple_hash("BEGIN");
uint32_t END_HASH = simple_hash("END");
uint32_t FLUSH_HASH = simple_hash("FLUSH");
uint32_t CHART_HASH = simple_hash("CHART");
uint32_t DIMENSION_HASH = simple_hash("DIMENSION");
uint32_t DISABLE_HASH = simple_hash("DISABLE");
#ifdef DETACH_PLUGINS_FROM_NETDATA
uint32_t MYPID_HASH = simple_hash("MYPID");
uint32_t STOPPING_WAKE_ME_UP_PLEASE_HASH = simple_hash("STOPPING_WAKE_ME_UP_PLEASE");
#endif
while(likely(1)) {
if(unlikely(netdata_exit)) break;
FILE *fp = mypopen(cd->cmd, &cd->pid);
if(unlikely(!fp)) {
error("Cannot popen(\"%s\", \"r\").", cd->cmd);
break;
}
info("PLUGINSD: '%s' running on pid %d", cd->fullfilename, cd->pid);
RRDSET *st = NULL;
unsigned long long count = 0;
char *s;
uint32_t hash;
while(likely(fgets(line, PLUGINSD_LINE_MAX, fp) != NULL)) {
if(unlikely(netdata_exit)) break;
line[PLUGINSD_LINE_MAX] = '\0';
// debug(D_PLUGINSD, "PLUGINSD: %s: %s", cd->filename, line);
int w = pluginsd_split_words(line, words, MAX_WORDS);
s = words[0];
if(unlikely(!s || !*s || !w)) {
// debug(D_PLUGINSD, "PLUGINSD: empty line");
continue;
}
// debug(D_PLUGINSD, "PLUGINSD: words 0='%s' 1='%s' 2='%s' 3='%s' 4='%s' 5='%s' 6='%s' 7='%s' 8='%s' 9='%s'", words[0], words[1], words[2], words[3], words[4], words[5], words[6], words[7], words[8], words[9]);
hash = simple_hash(s);
if(likely(hash == SET_HASH && !strcmp(s, "SET"))) {
char *dimension = words[1];
char *value = words[2];
if(unlikely(!dimension || !*dimension)) {
error("PLUGINSD: '%s' is requesting a SET on chart '%s', without a dimension. Disabling it.", cd->fullfilename, st->id);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
if(unlikely(!value || !*value)) value = NULL;
if(unlikely(!st)) {
error("PLUGINSD: '%s' is requesting a SET on dimension %s with value %s, without a BEGIN. Disabling it.", cd->fullfilename, dimension, value?value:"<nothing>");
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
if(unlikely(st->debug)) debug(D_PLUGINSD, "PLUGINSD: '%s' is setting dimension %s/%s to %s", cd->fullfilename, st->id, dimension, value?value:"<nothing>");
if(value) rrddim_set(st, dimension, atoll(value));
count++;
}
else if(likely(hash == BEGIN_HASH && !strcmp(s, "BEGIN"))) {
char *id = words[1];
char *microseconds_txt = words[2];
if(unlikely(!id)) {
error("PLUGINSD: '%s' is requesting a BEGIN without a chart id. Disabling it.", cd->fullfilename);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
st = rrdset_find(id);
if(unlikely(!st)) {
error("PLUGINSD: '%s' is requesting a BEGIN on chart '%s', which does not exist. Disabling it.", cd->fullfilename, id);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
if(likely(st->counter_done)) {
unsigned long long microseconds = 0;
if(microseconds_txt && *microseconds_txt) microseconds = strtoull(microseconds_txt, NULL, 10);
if(microseconds) rrdset_next_usec(st, microseconds);
else rrdset_next_plugins(st);
}
}
else if(likely(hash == END_HASH && !strcmp(s, "END"))) {
if(unlikely(!st)) {
error("PLUGINSD: '%s' is requesting an END, without a BEGIN. Disabling it.", cd->fullfilename);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
if(unlikely(st->debug)) debug(D_PLUGINSD, "PLUGINSD: '%s' is requesting a END on chart %s", cd->fullfilename, st->id);
rrdset_done(st);
st = NULL;
}
else if(likely(hash == FLUSH_HASH && !strcmp(s, "FLUSH"))) {
debug(D_PLUGINSD, "PLUGINSD: '%s' is requesting a FLUSH", cd->fullfilename);
st = NULL;
}
else if(likely(hash == CHART_HASH && !strcmp(s, "CHART"))) {
st = NULL;
char *type = words[1];
char *id = NULL;
if(likely(type)) {
id = strchr(type, '.');
if(likely(id)) { *id = '\0'; id++; }
}
char *name = words[2];
char *title = words[3];
char *units = words[4];
char *family = words[5];
char *category = words[6];
char *chart = words[7];
char *priority_s = words[8];
char *update_every_s = words[9];
if(unlikely(!type || !*type || !id || !*id)) {
error("PLUGINSD: '%s' is requesting a CHART, without a type.id. Disabling it.", cd->fullfilename);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
int priority = 1000;
if(likely(priority_s)) priority = atoi(priority_s);
int update_every = cd->update_every;
if(likely(update_every_s)) update_every = atoi(update_every_s);
if(unlikely(!update_every)) update_every = cd->update_every;
int chart_type = RRDSET_TYPE_LINE;
if(unlikely(chart)) chart_type = rrdset_type_id(chart);
if(unlikely(!name || !*name)) name = NULL;
if(unlikely(!family || !*family)) family = id;
if(unlikely(!category || !*category)) category = type;
st = rrdset_find_bytype(type, id);
if(unlikely(!st)) {
debug(D_PLUGINSD, "PLUGINSD: Creating chart type='%s', id='%s', name='%s', family='%s', category='%s', chart='%s', priority=%d, update_every=%d"
, type, id
, name?name:""
, family?family:""
, category?category:""
, rrdset_type_name(chart_type)
, priority
, update_every
);
st = rrdset_create(type, id, name, family, title, units, priority, update_every, chart_type);
cd->update_every = update_every;
if(unlikely(strcmp(category, "none") == 0)) st->isdetail = 1;
}
else debug(D_PLUGINSD, "PLUGINSD: Chart '%s' already exists. Not adding it again.", st->id);
}
else if(likely(hash == DIMENSION_HASH && !strcmp(s, "DIMENSION"))) {
char *id = words[1];
char *name = words[2];
char *algorithm = words[3];
char *multiplier_s = words[4];
char *divisor_s = words[5];
char *options = words[6];
if(unlikely(!id || !*id)) {
error("PLUGINSD: '%s' is requesting a DIMENSION, without an id. Disabling it.", cd->fullfilename);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
if(unlikely(!st)) {
error("PLUGINSD: '%s' is requesting a DIMENSION, without a CHART. Disabling it.", cd->fullfilename);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
long multiplier = 1;
if(multiplier_s && *multiplier_s) multiplier = atol(multiplier_s);
if(unlikely(!multiplier)) multiplier = 1;
long divisor = 1;
if(likely(divisor_s && *divisor_s)) divisor = atol(divisor_s);
if(unlikely(!divisor)) divisor = 1;
if(unlikely(!algorithm || !*algorithm)) algorithm = "absolute";
if(unlikely(st->debug)) debug(D_PLUGINSD, "PLUGINSD: Creating dimension in chart %s, id='%s', name='%s', algorithm='%s', multiplier=%ld, divisor=%ld, hidden='%s'"
, st->id
, id
, name?name:""
, rrddim_algorithm_name(rrddim_algorithm_id(algorithm))
, multiplier
, divisor
, options?options:""
);
RRDDIM *rd = rrddim_find(st, id);
if(unlikely(!rd)) {
rd = rrddim_add(st, id, name, multiplier, divisor, rrddim_algorithm_id(algorithm));
rd->flags = 0x00000000;
if(options && *options) {
if(strstr(options, "hidden") != NULL) rd->flags |= RRDDIM_FLAG_HIDDEN;
if(strstr(options, "noreset") != NULL) rd->flags |= RRDDIM_FLAG_DONT_DETECT_RESETS_OR_OVERFLOWS;
if(strstr(options, "nooverflow") != NULL) rd->flags |= RRDDIM_FLAG_DONT_DETECT_RESETS_OR_OVERFLOWS;
}
}
else if(unlikely(st->debug)) debug(D_PLUGINSD, "PLUGINSD: dimension %s/%s already exists. Not adding it again.", st->id, id);
}
else if(unlikely(hash == DISABLE_HASH && !strcmp(s, "DISABLE"))) {
error("PLUGINSD: '%s' called DISABLE. Disabling it.", cd->fullfilename);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
#ifdef DETACH_PLUGINS_FROM_NETDATA
else if(likely(hash == MYPID_HASH && !strcmp(s, "MYPID"))) {
char *pid_s = words[1];
pid_t pid = atol(pid_s);
if(likely(pid)) cd->pid = pid;
debug(D_PLUGINSD, "PLUGINSD: %s is on pid %d", cd->id, cd->pid);
}
else if(likely(hash == STOPPING_WAKE_ME_UP_PLEASE_HASH && !strcmp(s, "STOPPING_WAKE_ME_UP_PLEASE"))) {
error("PLUGINSD: '%s' (pid %d) called STOPPING_WAKE_ME_UP_PLEASE.", cd->fullfilename, cd->pid);
gettimeofday(&now, NULL);
if(unlikely(!usec && !susec)) {
// our first run
susec = cd->rrd_update_every * 1000000ULL;
}
else {
// second+ run
usec = usecdiff(&now, &last) - susec;
error("PLUGINSD: %s last loop took %llu usec (worked for %llu, sleeped for %llu).\n", cd->fullfilename, usec + susec, usec, susec);
if(unlikely(usec < (rrd_update_every * 1000000ULL / 2ULL))) susec = (rrd_update_every * 1000000ULL) - usec;
else susec = rrd_update_every * 1000000ULL / 2ULL;
}
error("PLUGINSD: %s sleeping for %llu. Will kill with SIGCONT pid %d to wake it up.\n", cd->fullfilename, susec, cd->pid);
usleep(susec);
killpid(cd->pid, SIGCONT);
bcopy(&now, &last, sizeof(struct timeval));
break;
}
#endif
else {
error("PLUGINSD: '%s' is sending command '%s' which is not known by netdata. Disabling it.", cd->fullfilename, s);
cd->enabled = 0;
killpid(cd->pid, SIGTERM);
break;
}
}
info("PLUGINSD: '%s' on pid %d stopped.", cd->fullfilename, cd->pid);
// fgets() failed or loop broke
int code = mypclose(fp, cd->pid);
if(code == 1 || code == 127) {
// 1 = DISABLE
// 127 = cannot even run it
error("PLUGINSD: '%s' (pid %d) exited with code %d. Disabling it.", cd->fullfilename, cd->pid, code);
cd->enabled = 0;
}
if(netdata_exit) {
cd->pid = 0;
cd->enabled = 0;
cd->obsolete = 1;
return NULL;
}
if(unlikely(!count && cd->enabled)) {
error("PLUGINSD: '%s' (pid %d) does not generate usefull output. Waiting a bit before starting it again.", cd->fullfilename, cd->pid);
sleep((unsigned int) (cd->update_every * 10));
}
cd->pid = 0;
if(likely(cd->enabled)) sleep((unsigned int) cd->update_every);
else break;
}
cd->obsolete = 1;
return NULL;
}
int do_proc_diskstats(int update_every, unsigned long long dt) {
static procfile *ff = NULL;
static char path_to_get_hw_sector_size[FILENAME_MAX + 1] = "";
static int enable_new_disks = -1;
static int do_io = -1, do_ops = -1, do_mops = -1, do_iotime = -1, do_qops = -1, do_util = -1, do_backlog = -1;
if(enable_new_disks == -1) enable_new_disks = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "enable new disks detected at runtime", CONFIG_ONDEMAND_ONDEMAND);
if(do_io == -1) do_io = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "bandwidth for all disks", CONFIG_ONDEMAND_ONDEMAND);
if(do_ops == -1) do_ops = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "operations for all disks", CONFIG_ONDEMAND_ONDEMAND);
if(do_mops == -1) do_mops = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "merged operations for all disks", CONFIG_ONDEMAND_ONDEMAND);
if(do_iotime == -1) do_iotime = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "i/o time for all disks", CONFIG_ONDEMAND_ONDEMAND);
if(do_qops == -1) do_qops = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "queued operations for all disks", CONFIG_ONDEMAND_ONDEMAND);
if(do_util == -1) do_util = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "utilization percentage for all disks", CONFIG_ONDEMAND_ONDEMAND);
if(do_backlog == -1)do_backlog = config_get_boolean_ondemand("plugin:proc:/proc/diskstats", "backlog for all disks", CONFIG_ONDEMAND_ONDEMAND);
if(!ff) {
char filename[FILENAME_MAX + 1];
snprintf(filename, FILENAME_MAX, "%s%s", global_host_prefix, "/proc/diskstats");
ff = procfile_open(config_get("plugin:proc:/proc/diskstats", "filename to monitor", filename), " \t", PROCFILE_FLAG_DEFAULT);
}
if(!ff) return 1;
if(!path_to_get_hw_sector_size[0]) {
char filename[FILENAME_MAX + 1];
snprintf(filename, FILENAME_MAX, "%s%s", global_host_prefix, "/sys/block/%s/queue/hw_sector_size");
snprintf(path_to_get_hw_sector_size, FILENAME_MAX, "%s%s", global_host_prefix, config_get("plugin:proc:/proc/diskstats", "path to get h/w sector size", filename));
}
ff = procfile_readall(ff);
if(!ff) return 0; // we return 0, so that we will retry to open it next time
uint32_t lines = procfile_lines(ff), l;
uint32_t words;
for(l = 0; l < lines ;l++) {
char *disk;
unsigned long long major = 0, minor = 0,
reads = 0, mreads = 0, readsectors = 0, readms = 0,
writes = 0, mwrites = 0, writesectors = 0, writems = 0,
queued_ios = 0, busy_ms = 0, backlog_ms = 0;
unsigned long long last_reads = 0, last_readsectors = 0, last_readms = 0,
last_writes = 0, last_writesectors = 0, last_writems = 0,
last_busy_ms = 0;
words = procfile_linewords(ff, l);
if(words < 14) continue;
major = strtoull(procfile_lineword(ff, l, 0), NULL, 10);
minor = strtoull(procfile_lineword(ff, l, 1), NULL, 10);
disk = procfile_lineword(ff, l, 2);
// # of reads completed # of writes completed
// This is the total number of reads or writes completed successfully.
reads = strtoull(procfile_lineword(ff, l, 3), NULL, 10); // rd_ios
writes = strtoull(procfile_lineword(ff, l, 7), NULL, 10); // wr_ios
// # of reads merged # of writes merged
// Reads and writes which are adjacent to each other may be merged for
// efficiency. Thus two 4K reads may become one 8K read before it is
// ultimately handed to the disk, and so it will be counted (and queued)
mreads = strtoull(procfile_lineword(ff, l, 4), NULL, 10); // rd_merges_or_rd_sec
mwrites = strtoull(procfile_lineword(ff, l, 8), NULL, 10); // wr_merges
// # of sectors read # of sectors written
// This is the total number of sectors read or written successfully.
readsectors = strtoull(procfile_lineword(ff, l, 5), NULL, 10); // rd_sec_or_wr_ios
writesectors = strtoull(procfile_lineword(ff, l, 9), NULL, 10); // wr_sec
// # of milliseconds spent reading # of milliseconds spent writing
// This is the total number of milliseconds spent by all reads or writes (as
// measured from __make_request() to end_that_request_last()).
readms = strtoull(procfile_lineword(ff, l, 6), NULL, 10); // rd_ticks_or_wr_sec
writems = strtoull(procfile_lineword(ff, l, 10), NULL, 10); // wr_ticks
// # of I/Os currently in progress
// The only field that should go to zero. Incremented as requests are
// given to appropriate struct request_queue and decremented as they finish.
queued_ios = strtoull(procfile_lineword(ff, l, 11), NULL, 10); // ios_pgr
// # of milliseconds spent doing I/Os
// This field increases so long as field queued_ios is nonzero.
busy_ms = strtoull(procfile_lineword(ff, l, 12), NULL, 10); // tot_ticks
// weighted # of milliseconds spent doing I/Os
// This field is incremented at each I/O start, I/O completion, I/O
// merge, or read of these stats by the number of I/Os in progress
// (field queued_ios) times the number of milliseconds spent doing I/O since the
// last update of this field. This can provide an easy measure of both
// I/O completion time and the backlog that may be accumulating.
backlog_ms = strtoull(procfile_lineword(ff, l, 13), NULL, 10); // rq_ticks
int def_enabled = 0;
// remove slashes from disk names
char *s;
for(s = disk; *s ;s++) if(*s == '/') *s = '_';
switch(major) {
case 9: // MDs
case 43: // network block
case 144: // nfs
case 145: // nfs
case 146: // nfs
case 199: // veritas
case 201: // veritas
case 251: // dm
def_enabled = enable_new_disks;
break;
case 48: // RAID
case 49: // RAID
case 50: // RAID
case 51: // RAID
case 52: // RAID
case 53: // RAID
case 54: // RAID
case 55: // RAID
case 112: // RAID
case 136: // RAID
case 137: // RAID
case 138: // RAID
case 139: // RAID
case 140: // RAID
case 141: // RAID
case 142: // RAID
case 143: // RAID
case 179: // MMC
case 180: // USB
if(minor % 8) def_enabled = 0; // partitions
else def_enabled = enable_new_disks;
break;
case 8: // scsi disks
case 65: // scsi disks
case 66: // scsi disks
case 67: // scsi disks
case 68: // scsi disks
case 69: // scsi disks
case 70: // scsi disks
case 71: // scsi disks
case 72: // scsi disks
case 73: // scsi disks
case 74: // scsi disks
case 75: // scsi disks
case 76: // scsi disks
case 77: // scsi disks
case 78: // scsi disks
case 79: // scsi disks
case 80: // i2o
case 81: // i2o
case 82: // i2o
case 83: // i2o
case 84: // i2o
case 85: // i2o
case 86: // i2o
case 87: // i2o
case 101: // hyperdisk
case 102: // compressed
case 104: // scsi
case 105: // scsi
case 106: // scsi
case 107: // scsi
case 108: // scsi
case 109: // scsi
case 110: // scsi
case 111: // scsi
case 114: // bios raid
case 116: // ram board
case 128: // scsi
case 129: // scsi
case 130: // scsi
case 131: // scsi
case 132: // scsi
case 133: // scsi
case 134: // scsi
case 135: // scsi
case 153: // raid
case 202: // xen
case 253: // virtio
case 256: // flash
case 257: // flash
case 259: // nvme0n1 issue #119
if(minor % 16) def_enabled = 0; // partitions
else def_enabled = enable_new_disks;
break;
case 160: // raid
case 161: // raid
if(minor % 32) def_enabled = 0; // partitions
else def_enabled = enable_new_disks;
break;
case 3: // ide
case 13: // 8bit ide
case 22: // ide
case 33: // ide
case 34: // ide
case 56: // ide
case 57: // ide
case 88: // ide
case 89: // ide
case 90: // ide
case 91: // ide
if(minor % 64) def_enabled = 0; // partitions
else def_enabled = enable_new_disks;
break;
case 252: // zram
def_enabled = 0;
break;
default:
def_enabled = 0;
break;
}
int ddo_io = do_io, ddo_ops = do_ops, ddo_mops = do_mops, ddo_iotime = do_iotime, ddo_qops = do_qops, ddo_util = do_util, ddo_backlog = do_backlog;
// check which charts are enabled for this disk
{
char var_name[4096 + 1];
snprintf(var_name, 4096, "plugin:proc:/proc/diskstats:%s", disk);
def_enabled = config_get_boolean_ondemand(var_name, "enabled", def_enabled);
if(def_enabled == CONFIG_ONDEMAND_NO) continue;
if(def_enabled == CONFIG_ONDEMAND_ONDEMAND && !reads && !writes) continue;
ddo_io = config_get_boolean_ondemand(var_name, "bandwidth", ddo_io);
ddo_ops = config_get_boolean_ondemand(var_name, "operations", ddo_ops);
ddo_mops = config_get_boolean_ondemand(var_name, "merged operations", ddo_mops);
ddo_iotime = config_get_boolean_ondemand(var_name, "i/o time", ddo_iotime);
ddo_qops = config_get_boolean_ondemand(var_name, "queued operations", ddo_qops);
ddo_util = config_get_boolean_ondemand(var_name, "utilization percentage", ddo_util);
ddo_backlog = config_get_boolean_ondemand(var_name, "backlog", ddo_backlog);
// by default, do not add charts that do not have values
if(ddo_io == CONFIG_ONDEMAND_ONDEMAND && !reads && !writes) ddo_io = 0;
if(ddo_mops == CONFIG_ONDEMAND_ONDEMAND && mreads == 0 && mwrites == 0) ddo_mops = 0;
if(ddo_iotime == CONFIG_ONDEMAND_ONDEMAND && readms == 0 && writems == 0) ddo_iotime = 0;
if(ddo_util == CONFIG_ONDEMAND_ONDEMAND && busy_ms == 0) ddo_util = 0;
if(ddo_backlog == CONFIG_ONDEMAND_ONDEMAND && backlog_ms == 0) ddo_backlog = 0;
if(ddo_qops == CONFIG_ONDEMAND_ONDEMAND && backlog_ms == 0) ddo_qops = 0;
// for absolute values, we need to switch the setting to 'yes'
// to allow it refresh from now on
if(ddo_qops == CONFIG_ONDEMAND_ONDEMAND) config_set(var_name, "queued operations", "yes");
}
RRDSET *st;
// --------------------------------------------------------------------
int sector_size = 512;
if(ddo_io) {
st = rrdset_find_bytype(RRD_TYPE_DISK, disk);
if(!st) {
char tf[FILENAME_MAX + 1], *t;
char ssfilename[FILENAME_MAX + 1];
strncpy(tf, disk, FILENAME_MAX);
tf[FILENAME_MAX] = '\0';
// replace all / with !
while((t = strchr(tf, '/'))) *t = '!';
snprintf(ssfilename, FILENAME_MAX, path_to_get_hw_sector_size, tf);
FILE *fpss = fopen(ssfilename, "r");
if(fpss) {
char ssbuffer[1025];
char *tmp = fgets(ssbuffer, 1024, fpss);
if(tmp) {
sector_size = atoi(tmp);
if(sector_size <= 0) {
error("Invalid sector size %d for device %s in %s. Assuming 512.", sector_size, disk, ssfilename);
sector_size = 512;
}
}
else error("Cannot read data for sector size for device %s from %s. Assuming 512.", disk, ssfilename);
fclose(fpss);
}
else error("Cannot read sector size for device %s from %s. Assuming 512.", disk, ssfilename);
st = rrdset_create(RRD_TYPE_DISK, disk, NULL, disk, "disk.io", "Disk I/O Bandwidth", "kilobytes/s", 2000, update_every, RRDSET_TYPE_AREA);
rrddim_add(st, "reads", NULL, sector_size, 1024, RRDDIM_INCREMENTAL);
rrddim_add(st, "writes", NULL, sector_size * -1, 1024, RRDDIM_INCREMENTAL);
}
else rrdset_next_usec(st, dt);
last_readsectors = rrddim_set(st, "reads", readsectors);
last_writesectors = rrddim_set(st, "writes", writesectors);
rrdset_done(st);
}
// --------------------------------------------------------------------
if(ddo_ops) {
st = rrdset_find_bytype("disk_ops", disk);
if(!st) {
st = rrdset_create("disk_ops", disk, NULL, disk, "disk.ops", "Disk Completed I/O Operations", "operations/s", 2001, update_every, RRDSET_TYPE_LINE);
st->isdetail = 1;
rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_INCREMENTAL);
rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_INCREMENTAL);
}
else rrdset_next_usec(st, dt);
last_reads = rrddim_set(st, "reads", reads);
last_writes = rrddim_set(st, "writes", writes);
rrdset_done(st);
}
// --------------------------------------------------------------------
if(ddo_qops) {
st = rrdset_find_bytype("disk_qops", disk);
if(!st) {
st = rrdset_create("disk_qops", disk, NULL, disk, "disk.qops", "Disk Current I/O Operations", "operations", 2002, update_every, RRDSET_TYPE_LINE);
st->isdetail = 1;
rrddim_add(st, "operations", NULL, 1, 1, RRDDIM_ABSOLUTE);
}
else rrdset_next_usec(st, dt);
rrddim_set(st, "operations", queued_ios);
rrdset_done(st);
}
// --------------------------------------------------------------------
if(ddo_backlog) {
st = rrdset_find_bytype("disk_backlog", disk);
if(!st) {
st = rrdset_create("disk_backlog", disk, NULL, disk, "disk.backlog", "Disk Backlog", "backlog (ms)", 2003, update_every, RRDSET_TYPE_AREA);
st->isdetail = 1;
rrddim_add(st, "backlog", NULL, 1, 10, RRDDIM_INCREMENTAL);
}
else rrdset_next_usec(st, dt);
rrddim_set(st, "backlog", backlog_ms);
rrdset_done(st);
}
// --------------------------------------------------------------------
if(ddo_util) {
st = rrdset_find_bytype("disk_util", disk);
if(!st) {
st = rrdset_create("disk_util", disk, NULL, disk, "disk.util", "Disk Utilization Time", "% of time working", 2004, update_every, RRDSET_TYPE_AREA);
st->isdetail = 1;
rrddim_add(st, "utilization", NULL, 1, 10, RRDDIM_INCREMENTAL);
}
else rrdset_next_usec(st, dt);
last_busy_ms = rrddim_set(st, "utilization", busy_ms);
rrdset_done(st);
}
// --------------------------------------------------------------------
if(ddo_mops) {
st = rrdset_find_bytype("disk_mops", disk);
if(!st) {
st = rrdset_create("disk_mops", disk, NULL, disk, "disk.mops", "Disk Merged Operations", "merged operations/s", 2021, update_every, RRDSET_TYPE_LINE);
st->isdetail = 1;
rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_INCREMENTAL);
rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_INCREMENTAL);
}
else rrdset_next_usec(st, dt);
rrddim_set(st, "reads", mreads);
rrddim_set(st, "writes", mwrites);
rrdset_done(st);
}
// --------------------------------------------------------------------
if(ddo_iotime) {
st = rrdset_find_bytype("disk_iotime", disk);
if(!st) {
st = rrdset_create("disk_iotime", disk, NULL, disk, "disk.iotime", "Disk Total I/O Time", "milliseconds/s", 2022, update_every, RRDSET_TYPE_LINE);
st->isdetail = 1;
rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_INCREMENTAL);
rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_INCREMENTAL);
}
else rrdset_next_usec(st, dt);
last_readms = rrddim_set(st, "reads", readms);
last_writems = rrddim_set(st, "writes", writems);
rrdset_done(st);
}
// --------------------------------------------------------------------
// calculate differential charts
// only if this is not the first time we run
if(dt) {
if(ddo_iotime && ddo_ops) {
st = rrdset_find_bytype("disk_await", disk);
if(!st) {
st = rrdset_create("disk_await", disk, NULL, disk, "disk.await", "Average Completed I/O Operation Time", "ms per operation", 2005, update_every, RRDSET_TYPE_LINE);
st->isdetail = 1;
rrddim_add(st, "reads", NULL, 1, 1, RRDDIM_ABSOLUTE);
rrddim_add(st, "writes", NULL, -1, 1, RRDDIM_ABSOLUTE);
}
else rrdset_next_usec(st, dt);
rrddim_set(st, "reads", (reads - last_reads) ? (readms - last_readms) / (reads - last_reads) : 0);
rrddim_set(st, "writes", (writes - last_writes) ? (writems - last_writems) / (writes - last_writes) : 0);
rrdset_done(st);
}
if(ddo_io && ddo_ops) {
st = rrdset_find_bytype("disk_avgsz", disk);
if(!st) {
st = rrdset_create("disk_avgsz", disk, NULL, disk, "disk.avgsz", "Average Completed I/O Operation Bandwidth", "kilobytes per operation", 2006, update_every, RRDSET_TYPE_AREA);
st->isdetail = 1;
rrddim_add(st, "reads", NULL, sector_size, 1024, RRDDIM_ABSOLUTE);
rrddim_add(st, "writes", NULL, -sector_size, 1024, RRDDIM_ABSOLUTE);
}
else rrdset_next_usec(st, dt);
rrddim_set(st, "reads", (reads - last_reads) ? (readsectors - last_readsectors) / (reads - last_reads) : 0);
rrddim_set(st, "writes", (writes - last_writes) ? (writesectors - last_writesectors) / (writes - last_writes) : 0);
rrdset_done(st);
}
if(ddo_util && ddo_ops) {
st = rrdset_find_bytype("disk_svctm", disk);
if(!st) {
st = rrdset_create("disk_svctm", disk, NULL, disk, "disk.svctm", "Average Service Time", "ms per operation", 2007, update_every, RRDSET_TYPE_LINE);
st->isdetail = 1;
rrddim_add(st, "svctm", NULL, 1, 1, RRDDIM_ABSOLUTE);
}
else rrdset_next_usec(st, dt);
rrddim_set(st, "svctm", ((reads - last_reads) + (writes - last_writes)) ? (busy_ms - last_busy_ms) / ((reads - last_reads) + (writes - last_writes)) : 0);
rrdset_done(st);
}
}
}
return 0;
}