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"
32 #include <assert.h>
33 #include <pthread.h>
35 /*
36 * Private data structures
37 * {{{ */
38 #define UT_FLAG_INVERT 0x01
39 #define UT_FLAG_PERSIST 0x02
40 #define UT_FLAG_PERCENTAGE 0x04
41 #define UT_FLAG_INTERESTING 0x08
42 typedef struct threshold_s
43 {
44 char host[DATA_MAX_NAME_LEN];
45 char plugin[DATA_MAX_NAME_LEN];
46 char plugin_instance[DATA_MAX_NAME_LEN];
47 char type[DATA_MAX_NAME_LEN];
48 char type_instance[DATA_MAX_NAME_LEN];
49 char data_source[DATA_MAX_NAME_LEN];
50 gauge_t warning_min;
51 gauge_t warning_max;
52 gauge_t failure_min;
53 gauge_t failure_max;
54 gauge_t hysteresis;
55 unsigned int flags;
56 int hits;
57 struct threshold_s *next;
58 } threshold_t;
59 /* }}} */
61 /*
62 * Private (static) variables
63 * {{{ */
64 static c_avl_tree_t *threshold_tree = NULL;
65 static pthread_mutex_t threshold_lock = PTHREAD_MUTEX_INITIALIZER;
66 /* }}} */
68 /*
69 * Threshold management
70 * ====================
71 * The following functions add, delete, search, etc. configured thresholds to
72 * the underlying AVL trees.
73 */
74 /*
75 * threshold_t *threshold_get
76 *
77 * Retrieve one specific threshold configuration. For looking up a threshold
78 * matching a value_list_t, see "threshold_search" below. Returns NULL if the
79 * specified threshold doesn't exist.
80 */
81 static threshold_t *threshold_get (const char *hostname,
82 const char *plugin, const char *plugin_instance,
83 const char *type, const char *type_instance)
84 { /* {{{ */
85 char name[6 * DATA_MAX_NAME_LEN];
86 threshold_t *th = NULL;
88 format_name (name, sizeof (name),
89 (hostname == NULL) ? "" : hostname,
90 (plugin == NULL) ? "" : plugin, plugin_instance,
91 (type == NULL) ? "" : type, type_instance);
92 name[sizeof (name) - 1] = '\0';
94 if (c_avl_get (threshold_tree, name, (void *) &th) == 0)
95 return (th);
96 else
97 return (NULL);
98 } /* }}} threshold_t *threshold_get */
100 /*
101 * int ut_threshold_add
102 *
103 * Adds a threshold configuration to the list of thresholds. The threshold_t
104 * structure is copied and may be destroyed after this call. Returns zero on
105 * success, non-zero otherwise.
106 */
107 static int ut_threshold_add (const threshold_t *th)
108 { /* {{{ */
109 char name[6 * DATA_MAX_NAME_LEN];
110 char *name_copy;
111 threshold_t *th_copy;
112 threshold_t *th_ptr;
113 int status = 0;
115 if (format_name (name, sizeof (name), th->host,
116 th->plugin, th->plugin_instance,
117 th->type, th->type_instance) != 0)
118 {
119 ERROR ("ut_threshold_add: format_name failed.");
120 return (-1);
121 }
123 name_copy = strdup (name);
124 if (name_copy == NULL)
125 {
126 ERROR ("ut_threshold_add: strdup failed.");
127 return (-1);
128 }
130 th_copy = (threshold_t *) malloc (sizeof (threshold_t));
131 if (th_copy == NULL)
132 {
133 sfree (name_copy);
134 ERROR ("ut_threshold_add: malloc failed.");
135 return (-1);
136 }
137 memcpy (th_copy, th, sizeof (threshold_t));
138 th_ptr = NULL;
140 DEBUG ("ut_threshold_add: Adding entry `%s'", name);
142 pthread_mutex_lock (&threshold_lock);
144 th_ptr = threshold_get (th->host, th->plugin, th->plugin_instance,
145 th->type, th->type_instance);
147 while ((th_ptr != NULL) && (th_ptr->next != NULL))
148 th_ptr = th_ptr->next;
150 if (th_ptr == NULL) /* no such threshold yet */
151 {
152 status = c_avl_insert (threshold_tree, name_copy, th_copy);
153 }
154 else /* th_ptr points to the last threshold in the list */
155 {
156 th_ptr->next = th_copy;
157 /* name_copy isn't needed */
158 sfree (name_copy);
159 }
161 pthread_mutex_unlock (&threshold_lock);
163 if (status != 0)
164 {
165 ERROR ("ut_threshold_add: c_avl_insert (%s) failed.", name);
166 sfree (name_copy);
167 sfree (th_copy);
168 }
170 return (status);
171 } /* }}} int ut_threshold_add */
173 /*
174 * threshold_t *threshold_search
175 *
176 * Searches for a threshold configuration using all the possible variations of
177 * "Host", "Plugin" and "Type" blocks. Returns NULL if no threshold could be
178 * found.
179 * XXX: This is likely the least efficient function in collectd.
180 */
181 static threshold_t *threshold_search (const value_list_t *vl)
182 { /* {{{ */
183 threshold_t *th;
185 if ((th = threshold_get (vl->host, vl->plugin, vl->plugin_instance,
186 vl->type, vl->type_instance)) != NULL)
187 return (th);
188 else if ((th = threshold_get (vl->host, vl->plugin, vl->plugin_instance,
189 vl->type, NULL)) != NULL)
190 return (th);
191 else if ((th = threshold_get (vl->host, vl->plugin, NULL,
192 vl->type, vl->type_instance)) != NULL)
193 return (th);
194 else if ((th = threshold_get (vl->host, vl->plugin, NULL,
195 vl->type, NULL)) != NULL)
196 return (th);
197 else if ((th = threshold_get (vl->host, "", NULL,
198 vl->type, vl->type_instance)) != NULL)
199 return (th);
200 else if ((th = threshold_get (vl->host, "", NULL,
201 vl->type, NULL)) != NULL)
202 return (th);
203 else if ((th = threshold_get ("", vl->plugin, vl->plugin_instance,
204 vl->type, vl->type_instance)) != NULL)
205 return (th);
206 else if ((th = threshold_get ("", vl->plugin, vl->plugin_instance,
207 vl->type, NULL)) != NULL)
208 return (th);
209 else if ((th = threshold_get ("", vl->plugin, NULL,
210 vl->type, vl->type_instance)) != NULL)
211 return (th);
212 else if ((th = threshold_get ("", vl->plugin, NULL,
213 vl->type, NULL)) != NULL)
214 return (th);
215 else if ((th = threshold_get ("", "", NULL,
216 vl->type, vl->type_instance)) != NULL)
217 return (th);
218 else if ((th = threshold_get ("", "", NULL,
219 vl->type, NULL)) != NULL)
220 return (th);
222 return (NULL);
223 } /* }}} threshold_t *threshold_search */
225 /*
226 * Configuration
227 * =============
228 * The following approximately two hundred functions are used to handle the
229 * configuration and fill the threshold list.
230 * {{{ */
231 static int ut_config_type_datasource (threshold_t *th, oconfig_item_t *ci)
232 {
233 if ((ci->values_num != 1)
234 || (ci->values[0].type != OCONFIG_TYPE_STRING))
235 {
236 WARNING ("threshold values: The `DataSource' option needs exactly one "
237 "string argument.");
238 return (-1);
239 }
241 sstrncpy (th->data_source, ci->values[0].value.string,
242 sizeof (th->data_source));
244 return (0);
245 } /* int ut_config_type_datasource */
247 static int ut_config_type_instance (threshold_t *th, oconfig_item_t *ci)
248 {
249 if ((ci->values_num != 1)
250 || (ci->values[0].type != OCONFIG_TYPE_STRING))
251 {
252 WARNING ("threshold values: The `Instance' option needs exactly one "
253 "string argument.");
254 return (-1);
255 }
257 sstrncpy (th->type_instance, ci->values[0].value.string,
258 sizeof (th->type_instance));
260 return (0);
261 } /* int ut_config_type_instance */
263 static int ut_config_type_max (threshold_t *th, oconfig_item_t *ci)
264 {
265 if ((ci->values_num != 1)
266 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
267 {
268 WARNING ("threshold values: The `%s' option needs exactly one "
269 "number argument.", ci->key);
270 return (-1);
271 }
273 if (strcasecmp (ci->key, "WarningMax") == 0)
274 th->warning_max = ci->values[0].value.number;
275 else
276 th->failure_max = ci->values[0].value.number;
278 return (0);
279 } /* int ut_config_type_max */
281 static int ut_config_type_min (threshold_t *th, oconfig_item_t *ci)
282 {
283 if ((ci->values_num != 1)
284 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
285 {
286 WARNING ("threshold values: The `%s' option needs exactly one "
287 "number argument.", ci->key);
288 return (-1);
289 }
291 if (strcasecmp (ci->key, "WarningMin") == 0)
292 th->warning_min = ci->values[0].value.number;
293 else
294 th->failure_min = ci->values[0].value.number;
296 return (0);
297 } /* int ut_config_type_min */
299 static int ut_config_type_hits (threshold_t *th, oconfig_item_t *ci)
300 {
301 if ((ci->values_num != 1)
302 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
303 {
304 WARNING ("threshold values: The `%s' option needs exactly one "
305 "number argument.", ci->key);
306 return (-1);
307 }
309 th->hits = ci->values[0].value.number;
311 return (0);
312 } /* int ut_config_type_hits */
314 static int ut_config_type_hysteresis (threshold_t *th, oconfig_item_t *ci)
315 {
316 if ((ci->values_num != 1)
317 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
318 {
319 WARNING ("threshold values: The `%s' option needs exactly one "
320 "number argument.", ci->key);
321 return (-1);
322 }
324 th->hysteresis = ci->values[0].value.number;
326 return (0);
327 } /* int ut_config_type_hysteresis */
329 static int ut_config_type (const threshold_t *th_orig, oconfig_item_t *ci)
330 {
331 int i;
332 threshold_t th;
333 int status = 0;
335 if ((ci->values_num != 1)
336 || (ci->values[0].type != OCONFIG_TYPE_STRING))
337 {
338 WARNING ("threshold values: The `Type' block needs exactly one string "
339 "argument.");
340 return (-1);
341 }
343 if (ci->children_num < 1)
344 {
345 WARNING ("threshold values: The `Type' block needs at least one option.");
346 return (-1);
347 }
349 memcpy (&th, th_orig, sizeof (th));
350 sstrncpy (th.type, ci->values[0].value.string, sizeof (th.type));
352 th.warning_min = NAN;
353 th.warning_max = NAN;
354 th.failure_min = NAN;
355 th.failure_max = NAN;
356 th.hits = 0;
357 th.hysteresis = 0;
358 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
360 for (i = 0; i < ci->children_num; i++)
361 {
362 oconfig_item_t *option = ci->children + i;
363 status = 0;
365 if (strcasecmp ("Instance", option->key) == 0)
366 status = ut_config_type_instance (&th, option);
367 else if (strcasecmp ("DataSource", option->key) == 0)
368 status = ut_config_type_datasource (&th, option);
369 else if ((strcasecmp ("WarningMax", option->key) == 0)
370 || (strcasecmp ("FailureMax", option->key) == 0))
371 status = ut_config_type_max (&th, option);
372 else if ((strcasecmp ("WarningMin", option->key) == 0)
373 || (strcasecmp ("FailureMin", option->key) == 0))
374 status = ut_config_type_min (&th, option);
375 else if (strcasecmp ("Interesting", option->key) == 0)
376 status = cf_util_get_flag (option, &th.flags, UT_FLAG_INTERESTING);
377 else if (strcasecmp ("Invert", option->key) == 0)
378 status = cf_util_get_flag (option, &th.flags, UT_FLAG_INVERT);
379 else if (strcasecmp ("Persist", option->key) == 0)
380 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERSIST);
381 else if (strcasecmp ("Percentage", option->key) == 0)
382 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERCENTAGE);
383 else if (strcasecmp ("Hits", option->key) == 0)
384 status = ut_config_type_hits (&th, option);
385 else if (strcasecmp ("Hysteresis", option->key) == 0)
386 status = ut_config_type_hysteresis (&th, option);
387 else
388 {
389 WARNING ("threshold values: Option `%s' not allowed inside a `Type' "
390 "block.", option->key);
391 status = -1;
392 }
394 if (status != 0)
395 break;
396 }
398 if (status == 0)
399 {
400 status = ut_threshold_add (&th);
401 }
403 return (status);
404 } /* int ut_config_type */
406 static int ut_config_plugin_instance (threshold_t *th, oconfig_item_t *ci)
407 {
408 if ((ci->values_num != 1)
409 || (ci->values[0].type != OCONFIG_TYPE_STRING))
410 {
411 WARNING ("threshold values: The `Instance' option needs exactly one "
412 "string argument.");
413 return (-1);
414 }
416 sstrncpy (th->plugin_instance, ci->values[0].value.string,
417 sizeof (th->plugin_instance));
419 return (0);
420 } /* int ut_config_plugin_instance */
422 static int ut_config_plugin (const threshold_t *th_orig, oconfig_item_t *ci)
423 {
424 int i;
425 threshold_t th;
426 int status = 0;
428 if ((ci->values_num != 1)
429 || (ci->values[0].type != OCONFIG_TYPE_STRING))
430 {
431 WARNING ("threshold values: The `Plugin' block needs exactly one string "
432 "argument.");
433 return (-1);
434 }
436 if (ci->children_num < 1)
437 {
438 WARNING ("threshold values: The `Plugin' block needs at least one nested "
439 "block.");
440 return (-1);
441 }
443 memcpy (&th, th_orig, sizeof (th));
444 sstrncpy (th.plugin, ci->values[0].value.string, sizeof (th.plugin));
446 for (i = 0; i < ci->children_num; i++)
447 {
448 oconfig_item_t *option = ci->children + i;
449 status = 0;
451 if (strcasecmp ("Type", option->key) == 0)
452 status = ut_config_type (&th, option);
453 else if (strcasecmp ("Instance", option->key) == 0)
454 status = ut_config_plugin_instance (&th, option);
455 else
456 {
457 WARNING ("threshold values: Option `%s' not allowed inside a `Plugin' "
458 "block.", option->key);
459 status = -1;
460 }
462 if (status != 0)
463 break;
464 }
466 return (status);
467 } /* int ut_config_plugin */
469 static int ut_config_host (const threshold_t *th_orig, oconfig_item_t *ci)
470 {
471 int i;
472 threshold_t th;
473 int status = 0;
475 if ((ci->values_num != 1)
476 || (ci->values[0].type != OCONFIG_TYPE_STRING))
477 {
478 WARNING ("threshold values: The `Host' block needs exactly one string "
479 "argument.");
480 return (-1);
481 }
483 if (ci->children_num < 1)
484 {
485 WARNING ("threshold values: The `Host' block needs at least one nested "
486 "block.");
487 return (-1);
488 }
490 memcpy (&th, th_orig, sizeof (th));
491 sstrncpy (th.host, ci->values[0].value.string, sizeof (th.host));
493 for (i = 0; i < ci->children_num; i++)
494 {
495 oconfig_item_t *option = ci->children + i;
496 status = 0;
498 if (strcasecmp ("Type", option->key) == 0)
499 status = ut_config_type (&th, option);
500 else if (strcasecmp ("Plugin", option->key) == 0)
501 status = ut_config_plugin (&th, option);
502 else
503 {
504 WARNING ("threshold values: Option `%s' not allowed inside a `Host' "
505 "block.", option->key);
506 status = -1;
507 }
509 if (status != 0)
510 break;
511 }
513 return (status);
514 } /* int ut_config_host */
516 int ut_config (oconfig_item_t *ci)
517 {
518 int i;
519 int status = 0;
521 threshold_t th;
523 if (ci->values_num != 0)
524 {
525 ERROR ("threshold values: The `Threshold' block may not have any "
526 "arguments.");
527 return (-1);
528 }
530 if (threshold_tree == NULL)
531 {
532 threshold_tree = c_avl_create ((void *) strcmp);
533 if (threshold_tree == NULL)
534 {
535 ERROR ("ut_config: c_avl_create failed.");
536 return (-1);
537 }
538 }
540 memset (&th, '\0', sizeof (th));
541 th.warning_min = NAN;
542 th.warning_max = NAN;
543 th.failure_min = NAN;
544 th.failure_max = NAN;
546 th.hits = 0;
547 th.hysteresis = 0;
548 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
550 for (i = 0; i < ci->children_num; i++)
551 {
552 oconfig_item_t *option = ci->children + i;
553 status = 0;
555 if (strcasecmp ("Type", option->key) == 0)
556 status = ut_config_type (&th, option);
557 else if (strcasecmp ("Plugin", option->key) == 0)
558 status = ut_config_plugin (&th, option);
559 else if (strcasecmp ("Host", option->key) == 0)
560 status = ut_config_host (&th, option);
561 else
562 {
563 WARNING ("threshold values: Option `%s' not allowed here.", option->key);
564 status = -1;
565 }
567 if (status != 0)
568 break;
569 }
571 return (status);
572 } /* int um_config */
573 /*
574 * End of the functions used to configure threshold values.
575 */
576 /* }}} */
578 /*
579 * int ut_report_state
580 *
581 * Checks if the `state' differs from the old state and creates a notification
582 * if appropriate.
583 * Does not fail.
584 */
585 static int ut_report_state (const data_set_t *ds,
586 const value_list_t *vl,
587 const threshold_t *th,
588 const gauge_t *values,
589 int ds_index,
590 int state)
591 { /* {{{ */
592 int state_old;
593 notification_t n;
595 char *buf;
596 size_t bufsize;
598 int status;
600 /* Check if hits matched */
601 if ( (th->hits != 0) )
602 {
603 int hits = uc_get_hits(ds,vl);
604 /* The STATE_OKAY always reset hits, or if hits reaise the limit */
605 if ( (state == STATE_OKAY) || (hits > th->hits) )
606 {
607 DEBUG("ut_report_state: reset uc_get_hits = 0");
608 uc_set_hits(ds,vl,0); /* reset hit counter and notify */
609 } else {
610 DEBUG("ut_report_state: th->hits = %d, uc_get_hits = %d",th->hits,uc_get_hits(ds,vl));
611 (void) uc_inc_hits(ds,vl,1); /* increase hit counter */
612 return (0);
613 }
614 } /* end check hits */
616 state_old = uc_get_state (ds, vl);
618 /* If the state didn't change, only report if `persistent' is specified and
619 * the state is not `okay'. */
620 if (state == state_old)
621 {
622 if ((th->flags & UT_FLAG_PERSIST) == 0)
623 return (0);
624 else if (state == STATE_OKAY)
625 return (0);
626 }
628 if (state != state_old)
629 uc_set_state (ds, vl, state);
631 NOTIFICATION_INIT_VL (&n, vl);
633 buf = n.message;
634 bufsize = sizeof (n.message);
636 if (state == STATE_OKAY)
637 n.severity = NOTIF_OKAY;
638 else if (state == STATE_WARNING)
639 n.severity = NOTIF_WARNING;
640 else
641 n.severity = NOTIF_FAILURE;
643 n.time = vl->time;
645 status = ssnprintf (buf, bufsize, "Host %s, plugin %s",
646 vl->host, vl->plugin);
647 buf += status;
648 bufsize -= status;
650 if (vl->plugin_instance[0] != '\0')
651 {
652 status = ssnprintf (buf, bufsize, " (instance %s)",
653 vl->plugin_instance);
654 buf += status;
655 bufsize -= status;
656 }
658 status = ssnprintf (buf, bufsize, " type %s", vl->type);
659 buf += status;
660 bufsize -= status;
662 if (vl->type_instance[0] != '\0')
663 {
664 status = ssnprintf (buf, bufsize, " (instance %s)",
665 vl->type_instance);
666 buf += status;
667 bufsize -= status;
668 }
670 plugin_notification_meta_add_string (&n, "DataSource",
671 ds->ds[ds_index].name);
672 plugin_notification_meta_add_double (&n, "CurrentValue", values[ds_index]);
673 plugin_notification_meta_add_double (&n, "WarningMin", th->warning_min);
674 plugin_notification_meta_add_double (&n, "WarningMax", th->warning_max);
675 plugin_notification_meta_add_double (&n, "FailureMin", th->failure_min);
676 plugin_notification_meta_add_double (&n, "FailureMax", th->failure_max);
678 /* Send an okay notification */
679 if (state == STATE_OKAY)
680 {
681 if (state_old == STATE_MISSING)
682 status = ssnprintf (buf, bufsize,
683 ": Value is no longer missing.");
684 else
685 status = ssnprintf (buf, bufsize,
686 ": All data sources are within range again.");
687 buf += status;
688 bufsize -= status;
689 }
690 else
691 {
692 double min;
693 double max;
695 min = (state == STATE_ERROR) ? th->failure_min : th->warning_min;
696 max = (state == STATE_ERROR) ? th->failure_max : th->warning_max;
698 if (th->flags & UT_FLAG_INVERT)
699 {
700 if (!isnan (min) && !isnan (max))
701 {
702 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
703 "%f. That is within the %s region of %f%s and %f%s.",
704 ds->ds[ds_index].name, values[ds_index],
705 (state == STATE_ERROR) ? "failure" : "warning",
706 min, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "",
707 max, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
708 }
709 else
710 {
711 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
712 "%f. That is %s the %s threshold of %f%s.",
713 ds->ds[ds_index].name, values[ds_index],
714 isnan (min) ? "below" : "above",
715 (state == STATE_ERROR) ? "failure" : "warning",
716 isnan (min) ? max : min,
717 ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
718 }
719 }
720 else if (th->flags & UT_FLAG_PERCENTAGE)
721 {
722 gauge_t value;
723 gauge_t sum;
724 int i;
726 sum = 0.0;
727 for (i = 0; i < vl->values_len; i++)
728 {
729 if (isnan (values[i]))
730 continue;
732 sum += values[i];
733 }
735 if (sum == 0.0)
736 value = NAN;
737 else
738 value = 100.0 * values[ds_index] / sum;
740 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
741 "%g (%.2f%%). That is %s the %s threshold of %.2f%%.",
742 ds->ds[ds_index].name, values[ds_index], value,
743 (value < min) ? "below" : "above",
744 (state == STATE_ERROR) ? "failure" : "warning",
745 (value < min) ? min : max);
746 }
747 else /* is not inverted */
748 {
749 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
750 "%f. That is %s the %s threshold of %f.",
751 ds->ds[ds_index].name, values[ds_index],
752 (values[ds_index] < min) ? "below" : "above",
753 (state == STATE_ERROR) ? "failure" : "warning",
754 (values[ds_index] < min) ? min : max);
755 }
756 buf += status;
757 bufsize -= status;
758 }
760 plugin_dispatch_notification (&n);
762 plugin_notification_meta_free (n.meta);
763 return (0);
764 } /* }}} int ut_report_state */
766 /*
767 * int ut_check_one_data_source
768 *
769 * Checks one data source against the given threshold configuration. If the
770 * `DataSource' option is set in the threshold, and the name does NOT match,
771 * `okay' is returned. If the threshold does match, its failure and warning
772 * min and max values are checked and `failure' or `warning' is returned if
773 * appropriate.
774 * Does not fail.
775 */
776 static int ut_check_one_data_source (const data_set_t *ds,
777 const value_list_t __attribute__((unused)) *vl,
778 const threshold_t *th,
779 const gauge_t *values,
780 int ds_index)
781 { /* {{{ */
782 const char *ds_name;
783 int is_warning = 0;
784 int is_failure = 0;
785 int prev_state = STATE_OKAY;
787 /* check if this threshold applies to this data source */
788 if (ds != NULL)
789 {
790 ds_name = ds->ds[ds_index].name;
791 if ((th->data_source[0] != 0)
792 && (strcmp (ds_name, th->data_source) != 0))
793 return (STATE_OKAY);
794 }
796 if ((th->flags & UT_FLAG_INVERT) != 0)
797 {
798 is_warning--;
799 is_failure--;
800 }
802 /* XXX: This is an experimental code, not optimized, not fast, not reliable,
803 * and probably, do not work as you expect. Enjoy! :D */
804 if ( (th->hysteresis > 0) && ((prev_state = uc_get_state(ds,vl)) != STATE_OKAY) )
805 {
806 switch(prev_state)
807 {
808 case STATE_ERROR:
809 if ( (!isnan (th->failure_min) && ((th->failure_min + th->hysteresis) < values[ds_index])) ||
810 (!isnan (th->failure_max) && ((th->failure_max - th->hysteresis) > values[ds_index])) )
811 return (STATE_OKAY);
812 else
813 is_failure++;
814 case STATE_WARNING:
815 if ( (!isnan (th->warning_min) && ((th->warning_min + th->hysteresis) < values[ds_index])) ||
816 (!isnan (th->warning_max) && ((th->warning_max - th->hysteresis) > values[ds_index])) )
817 return (STATE_OKAY);
818 else
819 is_warning++;
820 }
821 }
822 else { /* no hysteresis */
823 if ((!isnan (th->failure_min) && (th->failure_min > values[ds_index]))
824 || (!isnan (th->failure_max) && (th->failure_max < values[ds_index])))
825 is_failure++;
827 if ((!isnan (th->warning_min) && (th->warning_min > values[ds_index]))
828 || (!isnan (th->warning_max) && (th->warning_max < values[ds_index])))
829 is_warning++;
830 }
832 if (is_failure != 0)
833 return (STATE_ERROR);
835 if (is_warning != 0)
836 return (STATE_WARNING);
838 return (STATE_OKAY);
839 } /* }}} int ut_check_one_data_source */
841 /*
842 * int ut_check_one_threshold
843 *
844 * Checks all data sources of a value list against the given threshold, using
845 * the ut_check_one_data_source function above. Returns the worst status,
846 * which is `okay' if nothing has failed.
847 * Returns less than zero if the data set doesn't have any data sources.
848 */
849 static int ut_check_one_threshold (const data_set_t *ds,
850 const value_list_t *vl,
851 const threshold_t *th,
852 const gauge_t *values,
853 int *ret_ds_index)
854 { /* {{{ */
855 int ret = -1;
856 int ds_index = -1;
857 int i;
858 gauge_t values_copy[ds->ds_num];
860 memcpy (values_copy, values, sizeof (values_copy));
862 if ((th->flags & UT_FLAG_PERCENTAGE) != 0)
863 {
864 int num = 0;
865 gauge_t sum=0.0;
867 if (ds->ds_num == 1)
868 {
869 WARNING ("ut_check_one_threshold: The %s type has only one data "
870 "source, but you have configured to check this as a percentage. "
871 "That doesn't make much sense, because the percentage will always "
872 "be 100%%!", ds->type);
873 }
875 /* Prepare `sum' and `num'. */
876 for (i = 0; i < ds->ds_num; i++)
877 if (!isnan (values[i]))
878 {
879 num++;
880 sum += values[i];
881 }
883 if ((num == 0) /* All data sources are undefined. */
884 || (sum == 0.0)) /* Sum is zero, cannot calculate percentage. */
885 {
886 for (i = 0; i < ds->ds_num; i++)
887 values_copy[i] = NAN;
888 }
889 else /* We can actually calculate the percentage. */
890 {
891 for (i = 0; i < ds->ds_num; i++)
892 values_copy[i] = 100.0 * values[i] / sum;
893 }
894 } /* if (UT_FLAG_PERCENTAGE) */
896 for (i = 0; i < ds->ds_num; i++)
897 {
898 int status;
900 status = ut_check_one_data_source (ds, vl, th, values_copy, i);
901 if (ret < status)
902 {
903 ret = status;
904 ds_index = i;
905 }
906 } /* for (ds->ds_num) */
908 if (ret_ds_index != NULL)
909 *ret_ds_index = ds_index;
911 return (ret);
912 } /* }}} int ut_check_one_threshold */
914 /*
915 * int ut_check_threshold
916 *
917 * Gets a list of matching thresholds and searches for the worst status by one
918 * of the thresholds. Then reports that status using the ut_report_state
919 * function above.
920 * Returns zero on success and if no threshold has been configured. Returns
921 * less than zero on failure.
922 */
923 static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl,
924 __attribute__((unused)) user_data_t *ud)
925 { /* {{{ */
926 threshold_t *th;
927 gauge_t *values;
928 int status;
930 int worst_state = -1;
931 threshold_t *worst_th = NULL;
932 int worst_ds_index = -1;
934 if (threshold_tree == NULL)
935 return (0);
937 /* Is this lock really necessary? So far, thresholds are only inserted at
938 * startup. -octo */
939 pthread_mutex_lock (&threshold_lock);
940 th = threshold_search (vl);
941 pthread_mutex_unlock (&threshold_lock);
942 if (th == NULL)
943 return (0);
945 DEBUG ("ut_check_threshold: Found matching threshold(s)");
947 values = uc_get_rate (ds, vl);
948 if (values == NULL)
949 return (0);
951 while (th != NULL)
952 {
953 int ds_index = -1;
955 status = ut_check_one_threshold (ds, vl, th, values, &ds_index);
956 if (status < 0)
957 {
958 ERROR ("ut_check_threshold: ut_check_one_threshold failed.");
959 sfree (values);
960 return (-1);
961 }
963 if (worst_state < status)
964 {
965 worst_state = status;
966 worst_th = th;
967 worst_ds_index = ds_index;
968 }
970 th = th->next;
971 } /* while (th) */
973 status = ut_report_state (ds, vl, worst_th, values,
974 worst_ds_index, worst_state);
975 if (status != 0)
976 {
977 ERROR ("ut_check_threshold: ut_report_state failed.");
978 sfree (values);
979 return (-1);
980 }
982 sfree (values);
984 return (0);
985 } /* }}} int ut_check_threshold */
987 /*
988 * int ut_missing
989 *
990 * This function is called whenever a value goes "missing".
991 */
992 static int ut_missing (const value_list_t *vl,
993 __attribute__((unused)) user_data_t *ud)
994 { /* {{{ */
995 threshold_t *th;
996 cdtime_t missing_time;
997 char identifier[6 * DATA_MAX_NAME_LEN];
998 notification_t n;
1000 th = threshold_search (vl);
1001 if (th == NULL)
1002 return (0);
1004 missing_time = cdtime () - vl->time;
1005 FORMAT_VL (identifier, sizeof (identifier), vl);
1007 NOTIFICATION_INIT_VL (&n, vl);
1008 ssnprintf (n.message, sizeof (n.message),
1009 "%s has not been updated for %.3f seconds.",
1010 identifier, CDTIME_T_TO_DOUBLE (missing_time));
1012 plugin_dispatch_notification (&n);
1014 return (0);
1015 } /* }}} int ut_missing */
1017 void module_register (void)
1018 {
1019 plugin_register_complex_config ("threshold", ut_config);
1020 plugin_register_missing ("threshold", ut_missing,
1021 /* user data = */ NULL);
1022 plugin_register_write ("threshold", ut_check_threshold,
1023 /* user data = */ NULL);
1024 }
1026 /* vim: set sw=2 ts=8 sts=2 tw=78 et fdm=marker : */