1 /**
2 * collectd - src/threshold.c
3 * Copyright (C) 2007-2010 Florian Forster
4 * Copyright (C) 2008-2009 Sebastian Harl
5 * Copyright (C) 2009 Andrés J. Díaz
6 *
7 * This program is free software; you can redistribute it and/or modify it
8 * under the terms of the GNU General Public License as published by the
9 * Free Software Foundation; only version 2 of the License is applicable.
10 *
11 * This program is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with this program; if not, write to the Free Software Foundation, Inc.,
18 * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 *
20 * Author:
21 * Florian octo Forster <octo at collectd.org>
22 * Sebastian Harl <sh at tokkee.org>
23 * Andrés J. Díaz <ajdiaz at connectical.com>
24 **/
26 #include "collectd.h"
27 #include "common.h"
28 #include "plugin.h"
29 #include "utils_avltree.h"
30 #include "utils_cache.h"
31 #include "utils_threshold.h"
33 /*
34 * Threshold management
35 * ====================
36 * The following functions add, delete, search, etc. configured thresholds to
37 * the underlying AVL trees.
38 */
40 /*
41 * int ut_threshold_add
42 *
43 * Adds a threshold configuration to the list of thresholds. The threshold_t
44 * structure is copied and may be destroyed after this call. Returns zero on
45 * success, non-zero otherwise.
46 */
47 static int ut_threshold_add (const threshold_t *th)
48 { /* {{{ */
49 char name[6 * DATA_MAX_NAME_LEN];
50 char *name_copy;
51 threshold_t *th_copy;
52 threshold_t *th_ptr;
53 int status = 0;
55 if (format_name (name, sizeof (name), th->host,
56 th->plugin, th->plugin_instance,
57 th->type, th->type_instance) != 0)
58 {
59 ERROR ("ut_threshold_add: format_name failed.");
60 return (-1);
61 }
63 name_copy = strdup (name);
64 if (name_copy == NULL)
65 {
66 ERROR ("ut_threshold_add: strdup failed.");
67 return (-1);
68 }
70 th_copy = malloc (sizeof (*th_copy));
71 if (th_copy == NULL)
72 {
73 sfree (name_copy);
74 ERROR ("ut_threshold_add: malloc failed.");
75 return (-1);
76 }
77 memcpy (th_copy, th, sizeof (threshold_t));
79 DEBUG ("ut_threshold_add: Adding entry `%s'", name);
81 pthread_mutex_lock (&threshold_lock);
83 th_ptr = threshold_get (th->host, th->plugin, th->plugin_instance,
84 th->type, th->type_instance);
86 while ((th_ptr != NULL) && (th_ptr->next != NULL))
87 th_ptr = th_ptr->next;
89 if (th_ptr == NULL) /* no such threshold yet */
90 {
91 status = c_avl_insert (threshold_tree, name_copy, th_copy);
92 }
93 else /* th_ptr points to the last threshold in the list */
94 {
95 th_ptr->next = th_copy;
96 /* name_copy isn't needed */
97 sfree (name_copy);
98 }
100 pthread_mutex_unlock (&threshold_lock);
102 if (status != 0)
103 {
104 ERROR ("ut_threshold_add: c_avl_insert (%s) failed.", name);
105 sfree (name_copy);
106 sfree (th_copy);
107 }
109 return (status);
110 } /* }}} int ut_threshold_add */
112 /*
113 * Configuration
114 * =============
115 * The following approximately two hundred functions are used to handle the
116 * configuration and fill the threshold list.
117 * {{{ */
118 static int ut_config_type_datasource (threshold_t *th, oconfig_item_t *ci)
119 {
120 if ((ci->values_num != 1)
121 || (ci->values[0].type != OCONFIG_TYPE_STRING))
122 {
123 WARNING ("threshold values: The `DataSource' option needs exactly one "
124 "string argument.");
125 return (-1);
126 }
128 sstrncpy (th->data_source, ci->values[0].value.string,
129 sizeof (th->data_source));
131 return (0);
132 } /* int ut_config_type_datasource */
134 static int ut_config_type_instance (threshold_t *th, oconfig_item_t *ci)
135 {
136 if ((ci->values_num != 1)
137 || (ci->values[0].type != OCONFIG_TYPE_STRING))
138 {
139 WARNING ("threshold values: The `Instance' option needs exactly one "
140 "string argument.");
141 return (-1);
142 }
144 sstrncpy (th->type_instance, ci->values[0].value.string,
145 sizeof (th->type_instance));
147 return (0);
148 } /* int ut_config_type_instance */
150 static int ut_config_type_max (threshold_t *th, oconfig_item_t *ci)
151 {
152 if ((ci->values_num != 1)
153 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
154 {
155 WARNING ("threshold values: The `%s' option needs exactly one "
156 "number argument.", ci->key);
157 return (-1);
158 }
160 if (strcasecmp (ci->key, "WarningMax") == 0)
161 th->warning_max = ci->values[0].value.number;
162 else
163 th->failure_max = ci->values[0].value.number;
165 return (0);
166 } /* int ut_config_type_max */
168 static int ut_config_type_min (threshold_t *th, oconfig_item_t *ci)
169 {
170 if ((ci->values_num != 1)
171 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
172 {
173 WARNING ("threshold values: The `%s' option needs exactly one "
174 "number argument.", ci->key);
175 return (-1);
176 }
178 if (strcasecmp (ci->key, "WarningMin") == 0)
179 th->warning_min = ci->values[0].value.number;
180 else
181 th->failure_min = ci->values[0].value.number;
183 return (0);
184 } /* int ut_config_type_min */
186 static int ut_config_type_hits (threshold_t *th, oconfig_item_t *ci)
187 {
188 if ((ci->values_num != 1)
189 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
190 {
191 WARNING ("threshold values: The `%s' option needs exactly one "
192 "number argument.", ci->key);
193 return (-1);
194 }
196 th->hits = ci->values[0].value.number;
198 return (0);
199 } /* int ut_config_type_hits */
201 static int ut_config_type_hysteresis (threshold_t *th, oconfig_item_t *ci)
202 {
203 if ((ci->values_num != 1)
204 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
205 {
206 WARNING ("threshold values: The `%s' option needs exactly one "
207 "number argument.", ci->key);
208 return (-1);
209 }
211 th->hysteresis = ci->values[0].value.number;
213 return (0);
214 } /* int ut_config_type_hysteresis */
216 static int ut_config_type (const threshold_t *th_orig, oconfig_item_t *ci)
217 {
218 int i;
219 threshold_t th;
220 int status = 0;
222 if ((ci->values_num != 1)
223 || (ci->values[0].type != OCONFIG_TYPE_STRING))
224 {
225 WARNING ("threshold values: The `Type' block needs exactly one string "
226 "argument.");
227 return (-1);
228 }
230 if (ci->children_num < 1)
231 {
232 WARNING ("threshold values: The `Type' block needs at least one option.");
233 return (-1);
234 }
236 memcpy (&th, th_orig, sizeof (th));
237 sstrncpy (th.type, ci->values[0].value.string, sizeof (th.type));
239 th.warning_min = NAN;
240 th.warning_max = NAN;
241 th.failure_min = NAN;
242 th.failure_max = NAN;
243 th.hits = 0;
244 th.hysteresis = 0;
245 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
247 for (i = 0; i < ci->children_num; i++)
248 {
249 oconfig_item_t *option = ci->children + i;
251 if (strcasecmp ("Instance", option->key) == 0)
252 status = ut_config_type_instance (&th, option);
253 else if (strcasecmp ("DataSource", option->key) == 0)
254 status = ut_config_type_datasource (&th, option);
255 else if ((strcasecmp ("WarningMax", option->key) == 0)
256 || (strcasecmp ("FailureMax", option->key) == 0))
257 status = ut_config_type_max (&th, option);
258 else if ((strcasecmp ("WarningMin", option->key) == 0)
259 || (strcasecmp ("FailureMin", option->key) == 0))
260 status = ut_config_type_min (&th, option);
261 else if (strcasecmp ("Interesting", option->key) == 0)
262 status = cf_util_get_flag (option, &th.flags, UT_FLAG_INTERESTING);
263 else if (strcasecmp ("Invert", option->key) == 0)
264 status = cf_util_get_flag (option, &th.flags, UT_FLAG_INVERT);
265 else if (strcasecmp ("Persist", option->key) == 0)
266 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERSIST);
267 else if (strcasecmp ("PersistOK", option->key) == 0)
268 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERSIST_OK);
269 else if (strcasecmp ("Percentage", option->key) == 0)
270 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERCENTAGE);
271 else if (strcasecmp ("Hits", option->key) == 0)
272 status = ut_config_type_hits (&th, option);
273 else if (strcasecmp ("Hysteresis", option->key) == 0)
274 status = ut_config_type_hysteresis (&th, option);
275 else
276 {
277 WARNING ("threshold values: Option `%s' not allowed inside a `Type' "
278 "block.", option->key);
279 status = -1;
280 }
282 if (status != 0)
283 break;
284 }
286 if (status == 0)
287 {
288 status = ut_threshold_add (&th);
289 }
291 return (status);
292 } /* int ut_config_type */
294 static int ut_config_plugin_instance (threshold_t *th, oconfig_item_t *ci)
295 {
296 if ((ci->values_num != 1)
297 || (ci->values[0].type != OCONFIG_TYPE_STRING))
298 {
299 WARNING ("threshold values: The `Instance' option needs exactly one "
300 "string argument.");
301 return (-1);
302 }
304 sstrncpy (th->plugin_instance, ci->values[0].value.string,
305 sizeof (th->plugin_instance));
307 return (0);
308 } /* int ut_config_plugin_instance */
310 static int ut_config_plugin (const threshold_t *th_orig, oconfig_item_t *ci)
311 {
312 int i;
313 threshold_t th;
314 int status = 0;
316 if ((ci->values_num != 1)
317 || (ci->values[0].type != OCONFIG_TYPE_STRING))
318 {
319 WARNING ("threshold values: The `Plugin' block needs exactly one string "
320 "argument.");
321 return (-1);
322 }
324 if (ci->children_num < 1)
325 {
326 WARNING ("threshold values: The `Plugin' block needs at least one nested "
327 "block.");
328 return (-1);
329 }
331 memcpy (&th, th_orig, sizeof (th));
332 sstrncpy (th.plugin, ci->values[0].value.string, sizeof (th.plugin));
334 for (i = 0; i < ci->children_num; i++)
335 {
336 oconfig_item_t *option = ci->children + i;
338 if (strcasecmp ("Type", option->key) == 0)
339 status = ut_config_type (&th, option);
340 else if (strcasecmp ("Instance", option->key) == 0)
341 status = ut_config_plugin_instance (&th, option);
342 else
343 {
344 WARNING ("threshold values: Option `%s' not allowed inside a `Plugin' "
345 "block.", option->key);
346 status = -1;
347 }
349 if (status != 0)
350 break;
351 }
353 return (status);
354 } /* int ut_config_plugin */
356 static int ut_config_host (const threshold_t *th_orig, oconfig_item_t *ci)
357 {
358 int i;
359 threshold_t th;
360 int status = 0;
362 if ((ci->values_num != 1)
363 || (ci->values[0].type != OCONFIG_TYPE_STRING))
364 {
365 WARNING ("threshold values: The `Host' block needs exactly one string "
366 "argument.");
367 return (-1);
368 }
370 if (ci->children_num < 1)
371 {
372 WARNING ("threshold values: The `Host' block needs at least one nested "
373 "block.");
374 return (-1);
375 }
377 memcpy (&th, th_orig, sizeof (th));
378 sstrncpy (th.host, ci->values[0].value.string, sizeof (th.host));
380 for (i = 0; i < ci->children_num; i++)
381 {
382 oconfig_item_t *option = ci->children + i;
384 if (strcasecmp ("Type", option->key) == 0)
385 status = ut_config_type (&th, option);
386 else if (strcasecmp ("Plugin", option->key) == 0)
387 status = ut_config_plugin (&th, option);
388 else
389 {
390 WARNING ("threshold values: Option `%s' not allowed inside a `Host' "
391 "block.", option->key);
392 status = -1;
393 }
395 if (status != 0)
396 break;
397 }
399 return (status);
400 } /* int ut_config_host */
401 /*
402 * End of the functions used to configure threshold values.
403 */
404 /* }}} */
406 /*
407 * int ut_report_state
408 *
409 * Checks if the `state' differs from the old state and creates a notification
410 * if appropriate.
411 * Does not fail.
412 */
413 static int ut_report_state (const data_set_t *ds,
414 const value_list_t *vl,
415 const threshold_t *th,
416 const gauge_t *values,
417 int ds_index,
418 int state)
419 { /* {{{ */
420 int state_old;
421 notification_t n;
423 char *buf;
424 size_t bufsize;
426 int status;
428 /* Check if hits matched */
429 if ( (th->hits != 0) )
430 {
431 int hits = uc_get_hits(ds,vl);
432 /* STATE_OKAY resets hits unless PERSIST_OK flag is set. Hits resets if
433 * threshold is hit. */
434 if ( ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) ) || (hits > th->hits) )
435 {
436 DEBUG("ut_report_state: reset uc_get_hits = 0");
437 uc_set_hits(ds,vl,0); /* reset hit counter and notify */
438 } else {
439 DEBUG("ut_report_state: th->hits = %d, uc_get_hits = %d",th->hits,uc_get_hits(ds,vl));
440 (void) uc_inc_hits(ds,vl,1); /* increase hit counter */
441 return (0);
442 }
443 } /* end check hits */
445 state_old = uc_get_state (ds, vl);
447 /* If the state didn't change, report if `persistent' is specified. If the
448 * state is `okay', then only report if `persist_ok` flag is set. */
449 if (state == state_old)
450 {
451 if ((th->flags & UT_FLAG_PERSIST) == 0)
452 return (0);
453 else if ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) )
454 return (0);
455 }
457 if (state != state_old)
458 uc_set_state (ds, vl, state);
460 NOTIFICATION_INIT_VL (&n, vl);
462 buf = n.message;
463 bufsize = sizeof (n.message);
465 if (state == STATE_OKAY)
466 n.severity = NOTIF_OKAY;
467 else if (state == STATE_WARNING)
468 n.severity = NOTIF_WARNING;
469 else
470 n.severity = NOTIF_FAILURE;
472 n.time = vl->time;
474 status = ssnprintf (buf, bufsize, "Host %s, plugin %s",
475 vl->host, vl->plugin);
476 buf += status;
477 bufsize -= status;
479 if (vl->plugin_instance[0] != '\0')
480 {
481 status = ssnprintf (buf, bufsize, " (instance %s)",
482 vl->plugin_instance);
483 buf += status;
484 bufsize -= status;
485 }
487 status = ssnprintf (buf, bufsize, " type %s", vl->type);
488 buf += status;
489 bufsize -= status;
491 if (vl->type_instance[0] != '\0')
492 {
493 status = ssnprintf (buf, bufsize, " (instance %s)",
494 vl->type_instance);
495 buf += status;
496 bufsize -= status;
497 }
499 plugin_notification_meta_add_string (&n, "DataSource",
500 ds->ds[ds_index].name);
501 plugin_notification_meta_add_double (&n, "CurrentValue", values[ds_index]);
502 plugin_notification_meta_add_double (&n, "WarningMin", th->warning_min);
503 plugin_notification_meta_add_double (&n, "WarningMax", th->warning_max);
504 plugin_notification_meta_add_double (&n, "FailureMin", th->failure_min);
505 plugin_notification_meta_add_double (&n, "FailureMax", th->failure_max);
507 /* Send an okay notification */
508 if (state == STATE_OKAY)
509 {
510 if (state_old == STATE_MISSING)
511 ssnprintf (buf, bufsize, ": Value is no longer missing.");
512 else
513 ssnprintf (buf, bufsize,
514 ": All data sources are within range again. "
515 "Current value of \"%s\" is %f.",
516 ds->ds[ds_index].name, values[ds_index]);
517 }
518 else
519 {
520 double min;
521 double max;
523 min = (state == STATE_ERROR) ? th->failure_min : th->warning_min;
524 max = (state == STATE_ERROR) ? th->failure_max : th->warning_max;
526 if (th->flags & UT_FLAG_INVERT)
527 {
528 if (!isnan (min) && !isnan (max))
529 {
530 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
531 "%f. That is within the %s region of %f%s and %f%s.",
532 ds->ds[ds_index].name, values[ds_index],
533 (state == STATE_ERROR) ? "failure" : "warning",
534 min, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "",
535 max, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
536 }
537 else
538 {
539 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
540 "%f. That is %s the %s threshold of %f%s.",
541 ds->ds[ds_index].name, values[ds_index],
542 isnan (min) ? "below" : "above",
543 (state == STATE_ERROR) ? "failure" : "warning",
544 isnan (min) ? max : min,
545 ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
546 }
547 }
548 else if (th->flags & UT_FLAG_PERCENTAGE)
549 {
550 gauge_t value;
551 gauge_t sum;
552 size_t i;
554 sum = 0.0;
555 for (i = 0; i < vl->values_len; i++)
556 {
557 if (isnan (values[i]))
558 continue;
560 sum += values[i];
561 }
563 if (sum == 0.0)
564 value = NAN;
565 else
566 value = 100.0 * values[ds_index] / sum;
568 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
569 "%g (%.2f%%). That is %s the %s threshold of %.2f%%.",
570 ds->ds[ds_index].name, values[ds_index], value,
571 (value < min) ? "below" : "above",
572 (state == STATE_ERROR) ? "failure" : "warning",
573 (value < min) ? min : max);
574 }
575 else /* is not inverted */
576 {
577 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
578 "%f. That is %s the %s threshold of %f.",
579 ds->ds[ds_index].name, values[ds_index],
580 (values[ds_index] < min) ? "below" : "above",
581 (state == STATE_ERROR) ? "failure" : "warning",
582 (values[ds_index] < min) ? min : max);
583 }
584 }
586 plugin_dispatch_notification (&n);
588 plugin_notification_meta_free (n.meta);
589 return (0);
590 } /* }}} int ut_report_state */
592 /*
593 * int ut_check_one_data_source
594 *
595 * Checks one data source against the given threshold configuration. If the
596 * `DataSource' option is set in the threshold, and the name does NOT match,
597 * `okay' is returned. If the threshold does match, its failure and warning
598 * min and max values are checked and `failure' or `warning' is returned if
599 * appropriate.
600 * Does not fail.
601 */
602 static int ut_check_one_data_source (const data_set_t *ds,
603 const value_list_t __attribute__((unused)) *vl,
604 const threshold_t *th,
605 const gauge_t *values,
606 int ds_index)
607 { /* {{{ */
608 const char *ds_name;
609 int is_warning = 0;
610 int is_failure = 0;
611 int prev_state = STATE_OKAY;
613 /* check if this threshold applies to this data source */
614 if (ds != NULL)
615 {
616 ds_name = ds->ds[ds_index].name;
617 if ((th->data_source[0] != 0)
618 && (strcmp (ds_name, th->data_source) != 0))
619 return (STATE_OKAY);
620 }
622 if ((th->flags & UT_FLAG_INVERT) != 0)
623 {
624 is_warning--;
625 is_failure--;
626 }
628 /* XXX: This is an experimental code, not optimized, not fast, not reliable,
629 * and probably, do not work as you expect. Enjoy! :D */
630 if (th->hysteresis > 0)
631 {
632 prev_state = uc_get_state(ds,vl);
633 /* The purpose of hysteresis is elliminating flapping state when the value
634 * oscilates around the thresholds. In other words, what is important is
635 * the previous state; if the new value would trigger a transition, make
636 * sure that we artificially widen the range which is considered to apply
637 * for the previous state, and only trigger the notification if the value
638 * is outside of this expanded range.
639 *
640 * There is no hysteresis for the OKAY state.
641 * */
642 gauge_t hysteresis_for_warning = 0, hysteresis_for_failure = 0;
643 switch (prev_state)
644 {
645 case STATE_ERROR:
646 hysteresis_for_failure = th->hysteresis;
647 break;
648 case STATE_WARNING:
649 hysteresis_for_warning = th->hysteresis;
650 break;
651 case STATE_OKAY:
652 /* do nothing -- the hysteresis only applies to the non-normal states */
653 break;
654 }
656 if ((!isnan (th->failure_min) && (th->failure_min + hysteresis_for_failure > values[ds_index]))
657 || (!isnan (th->failure_max) && (th->failure_max - hysteresis_for_failure < values[ds_index])))
658 is_failure++;
660 if ((!isnan (th->warning_min) && (th->warning_min + hysteresis_for_warning > values[ds_index]))
661 || (!isnan (th->warning_max) && (th->warning_max - hysteresis_for_warning < values[ds_index])))
662 is_warning++;
664 }
665 else { /* no hysteresis */
666 if ((!isnan (th->failure_min) && (th->failure_min > values[ds_index]))
667 || (!isnan (th->failure_max) && (th->failure_max < values[ds_index])))
668 is_failure++;
670 if ((!isnan (th->warning_min) && (th->warning_min > values[ds_index]))
671 || (!isnan (th->warning_max) && (th->warning_max < values[ds_index])))
672 is_warning++;
673 }
675 if (is_failure != 0)
676 return (STATE_ERROR);
678 if (is_warning != 0)
679 return (STATE_WARNING);
681 return (STATE_OKAY);
682 } /* }}} int ut_check_one_data_source */
684 /*
685 * int ut_check_one_threshold
686 *
687 * Checks all data sources of a value list against the given threshold, using
688 * the ut_check_one_data_source function above. Returns the worst status,
689 * which is `okay' if nothing has failed.
690 * Returns less than zero if the data set doesn't have any data sources.
691 */
692 static int ut_check_one_threshold (const data_set_t *ds,
693 const value_list_t *vl,
694 const threshold_t *th,
695 const gauge_t *values,
696 int *ret_ds_index)
697 { /* {{{ */
698 int ret = -1;
699 int ds_index = -1;
700 size_t i;
701 gauge_t values_copy[ds->ds_num];
703 memcpy (values_copy, values, sizeof (values_copy));
705 if ((th->flags & UT_FLAG_PERCENTAGE) != 0)
706 {
707 int num = 0;
708 gauge_t sum=0.0;
710 if (ds->ds_num == 1)
711 {
712 WARNING ("ut_check_one_threshold: The %s type has only one data "
713 "source, but you have configured to check this as a percentage. "
714 "That doesn't make much sense, because the percentage will always "
715 "be 100%%!", ds->type);
716 }
718 /* Prepare `sum' and `num'. */
719 for (i = 0; i < ds->ds_num; i++)
720 if (!isnan (values[i]))
721 {
722 num++;
723 sum += values[i];
724 }
726 if ((num == 0) /* All data sources are undefined. */
727 || (sum == 0.0)) /* Sum is zero, cannot calculate percentage. */
728 {
729 for (i = 0; i < ds->ds_num; i++)
730 values_copy[i] = NAN;
731 }
732 else /* We can actually calculate the percentage. */
733 {
734 for (i = 0; i < ds->ds_num; i++)
735 values_copy[i] = 100.0 * values[i] / sum;
736 }
737 } /* if (UT_FLAG_PERCENTAGE) */
739 for (i = 0; i < ds->ds_num; i++)
740 {
741 int status;
743 status = ut_check_one_data_source (ds, vl, th, values_copy, i);
744 if (ret < status)
745 {
746 ret = status;
747 ds_index = i;
748 }
749 } /* for (ds->ds_num) */
751 if (ret_ds_index != NULL)
752 *ret_ds_index = ds_index;
754 return (ret);
755 } /* }}} int ut_check_one_threshold */
757 /*
758 * int ut_check_threshold
759 *
760 * Gets a list of matching thresholds and searches for the worst status by one
761 * of the thresholds. Then reports that status using the ut_report_state
762 * function above.
763 * Returns zero on success and if no threshold has been configured. Returns
764 * less than zero on failure.
765 */
766 static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl,
767 __attribute__((unused)) user_data_t *ud)
768 { /* {{{ */
769 threshold_t *th;
770 gauge_t *values;
771 int status;
773 int worst_state = -1;
774 threshold_t *worst_th = NULL;
775 int worst_ds_index = -1;
777 if (threshold_tree == NULL)
778 return (0);
780 /* Is this lock really necessary? So far, thresholds are only inserted at
781 * startup. -octo */
782 pthread_mutex_lock (&threshold_lock);
783 th = threshold_search (vl);
784 pthread_mutex_unlock (&threshold_lock);
785 if (th == NULL)
786 return (0);
788 DEBUG ("ut_check_threshold: Found matching threshold(s)");
790 values = uc_get_rate (ds, vl);
791 if (values == NULL)
792 return (0);
794 while (th != NULL)
795 {
796 int ds_index = -1;
798 status = ut_check_one_threshold (ds, vl, th, values, &ds_index);
799 if (status < 0)
800 {
801 ERROR ("ut_check_threshold: ut_check_one_threshold failed.");
802 sfree (values);
803 return (-1);
804 }
806 if (worst_state < status)
807 {
808 worst_state = status;
809 worst_th = th;
810 worst_ds_index = ds_index;
811 }
813 th = th->next;
814 } /* while (th) */
816 status = ut_report_state (ds, vl, worst_th, values,
817 worst_ds_index, worst_state);
818 if (status != 0)
819 {
820 ERROR ("ut_check_threshold: ut_report_state failed.");
821 sfree (values);
822 return (-1);
823 }
825 sfree (values);
827 return (0);
828 } /* }}} int ut_check_threshold */
830 /*
831 * int ut_missing
832 *
833 * This function is called whenever a value goes "missing".
834 */
835 static int ut_missing (const value_list_t *vl,
836 __attribute__((unused)) user_data_t *ud)
837 { /* {{{ */
838 threshold_t *th;
839 cdtime_t missing_time;
840 char identifier[6 * DATA_MAX_NAME_LEN];
841 notification_t n;
842 cdtime_t now;
844 if (threshold_tree == NULL)
845 return (0);
847 th = threshold_search (vl);
848 /* dispatch notifications for "interesting" values only */
849 if ((th == NULL) || ((th->flags & UT_FLAG_INTERESTING) == 0))
850 return (0);
852 now = cdtime ();
853 missing_time = now - vl->time;
854 FORMAT_VL (identifier, sizeof (identifier), vl);
856 NOTIFICATION_INIT_VL (&n, vl);
857 ssnprintf (n.message, sizeof (n.message),
858 "%s has not been updated for %.3f seconds.",
859 identifier, CDTIME_T_TO_DOUBLE (missing_time));
860 n.time = now;
862 plugin_dispatch_notification (&n);
864 return (0);
865 } /* }}} int ut_missing */
867 static int ut_config (oconfig_item_t *ci)
868 { /* {{{ */
869 int i;
870 int status = 0;
871 int old_size = c_avl_size (threshold_tree);
873 threshold_t th;
875 if (threshold_tree == NULL)
876 {
877 threshold_tree = c_avl_create ((int (*) (const void *, const void *)) strcmp);
878 if (threshold_tree == NULL)
879 {
880 ERROR ("ut_config: c_avl_create failed.");
881 return (-1);
882 }
883 }
885 memset (&th, '\0', sizeof (th));
886 th.warning_min = NAN;
887 th.warning_max = NAN;
888 th.failure_min = NAN;
889 th.failure_max = NAN;
891 th.hits = 0;
892 th.hysteresis = 0;
893 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
895 for (i = 0; i < ci->children_num; i++)
896 {
897 oconfig_item_t *option = ci->children + i;
899 if (strcasecmp ("Type", option->key) == 0)
900 status = ut_config_type (&th, option);
901 else if (strcasecmp ("Plugin", option->key) == 0)
902 status = ut_config_plugin (&th, option);
903 else if (strcasecmp ("Host", option->key) == 0)
904 status = ut_config_host (&th, option);
905 else
906 {
907 WARNING ("threshold values: Option `%s' not allowed here.", option->key);
908 status = -1;
909 }
911 if (status != 0)
912 break;
913 }
915 /* register callbacks if this is the first time we see a valid config */
916 if ((old_size == 0) && (c_avl_size (threshold_tree) > 0))
917 {
918 plugin_register_missing ("threshold", ut_missing,
919 /* user data = */ NULL);
920 plugin_register_write ("threshold", ut_check_threshold,
921 /* user data = */ NULL);
922 }
924 return (status);
925 } /* }}} int um_config */
927 void module_register (void)
928 {
929 plugin_register_complex_config ("threshold", ut_config);
930 }
932 /* vim: set sw=2 ts=8 sts=2 tw=78 et fdm=marker : */