a4449741d0c566ee4df8dac7cb196c33e98dd9e5
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 #define UT_FLAG_PERSIST_OK 0x10
43 typedef struct threshold_s
44 {
45 char host[DATA_MAX_NAME_LEN];
46 char plugin[DATA_MAX_NAME_LEN];
47 char plugin_instance[DATA_MAX_NAME_LEN];
48 char type[DATA_MAX_NAME_LEN];
49 char type_instance[DATA_MAX_NAME_LEN];
50 char data_source[DATA_MAX_NAME_LEN];
51 gauge_t warning_min;
52 gauge_t warning_max;
53 gauge_t failure_min;
54 gauge_t failure_max;
55 gauge_t hysteresis;
56 unsigned int flags;
57 int hits;
58 struct threshold_s *next;
59 } threshold_t;
60 /* }}} */
62 /*
63 * Private (static) variables
64 * {{{ */
65 static c_avl_tree_t *threshold_tree = NULL;
66 static pthread_mutex_t threshold_lock = PTHREAD_MUTEX_INITIALIZER;
67 /* }}} */
69 /*
70 * Threshold management
71 * ====================
72 * The following functions add, delete, search, etc. configured thresholds to
73 * the underlying AVL trees.
74 */
75 /*
76 * threshold_t *threshold_get
77 *
78 * Retrieve one specific threshold configuration. For looking up a threshold
79 * matching a value_list_t, see "threshold_search" below. Returns NULL if the
80 * specified threshold doesn't exist.
81 */
82 static threshold_t *threshold_get (const char *hostname,
83 const char *plugin, const char *plugin_instance,
84 const char *type, const char *type_instance)
85 { /* {{{ */
86 char name[6 * DATA_MAX_NAME_LEN];
87 threshold_t *th = NULL;
89 format_name (name, sizeof (name),
90 (hostname == NULL) ? "" : hostname,
91 (plugin == NULL) ? "" : plugin, plugin_instance,
92 (type == NULL) ? "" : type, type_instance);
93 name[sizeof (name) - 1] = '\0';
95 if (c_avl_get (threshold_tree, name, (void *) &th) == 0)
96 return (th);
97 else
98 return (NULL);
99 } /* }}} threshold_t *threshold_get */
101 /*
102 * int ut_threshold_add
103 *
104 * Adds a threshold configuration to the list of thresholds. The threshold_t
105 * structure is copied and may be destroyed after this call. Returns zero on
106 * success, non-zero otherwise.
107 */
108 static int ut_threshold_add (const threshold_t *th)
109 { /* {{{ */
110 char name[6 * DATA_MAX_NAME_LEN];
111 char *name_copy;
112 threshold_t *th_copy;
113 threshold_t *th_ptr;
114 int status = 0;
116 if (format_name (name, sizeof (name), th->host,
117 th->plugin, th->plugin_instance,
118 th->type, th->type_instance) != 0)
119 {
120 ERROR ("ut_threshold_add: format_name failed.");
121 return (-1);
122 }
124 name_copy = strdup (name);
125 if (name_copy == NULL)
126 {
127 ERROR ("ut_threshold_add: strdup failed.");
128 return (-1);
129 }
131 th_copy = (threshold_t *) malloc (sizeof (threshold_t));
132 if (th_copy == NULL)
133 {
134 sfree (name_copy);
135 ERROR ("ut_threshold_add: malloc failed.");
136 return (-1);
137 }
138 memcpy (th_copy, th, sizeof (threshold_t));
139 th_ptr = NULL;
141 DEBUG ("ut_threshold_add: Adding entry `%s'", name);
143 pthread_mutex_lock (&threshold_lock);
145 th_ptr = threshold_get (th->host, th->plugin, th->plugin_instance,
146 th->type, th->type_instance);
148 while ((th_ptr != NULL) && (th_ptr->next != NULL))
149 th_ptr = th_ptr->next;
151 if (th_ptr == NULL) /* no such threshold yet */
152 {
153 status = c_avl_insert (threshold_tree, name_copy, th_copy);
154 }
155 else /* th_ptr points to the last threshold in the list */
156 {
157 th_ptr->next = th_copy;
158 /* name_copy isn't needed */
159 sfree (name_copy);
160 }
162 pthread_mutex_unlock (&threshold_lock);
164 if (status != 0)
165 {
166 ERROR ("ut_threshold_add: c_avl_insert (%s) failed.", name);
167 sfree (name_copy);
168 sfree (th_copy);
169 }
171 return (status);
172 } /* }}} int ut_threshold_add */
174 /*
175 * threshold_t *threshold_search
176 *
177 * Searches for a threshold configuration using all the possible variations of
178 * "Host", "Plugin" and "Type" blocks. Returns NULL if no threshold could be
179 * found.
180 * XXX: This is likely the least efficient function in collectd.
181 */
182 static threshold_t *threshold_search (const value_list_t *vl)
183 { /* {{{ */
184 threshold_t *th;
186 if ((th = threshold_get (vl->host, vl->plugin, vl->plugin_instance,
187 vl->type, vl->type_instance)) != NULL)
188 return (th);
189 else if ((th = threshold_get (vl->host, vl->plugin, vl->plugin_instance,
190 vl->type, NULL)) != NULL)
191 return (th);
192 else if ((th = threshold_get (vl->host, vl->plugin, NULL,
193 vl->type, vl->type_instance)) != NULL)
194 return (th);
195 else if ((th = threshold_get (vl->host, vl->plugin, NULL,
196 vl->type, NULL)) != NULL)
197 return (th);
198 else if ((th = threshold_get (vl->host, "", NULL,
199 vl->type, vl->type_instance)) != NULL)
200 return (th);
201 else if ((th = threshold_get (vl->host, "", NULL,
202 vl->type, NULL)) != NULL)
203 return (th);
204 else if ((th = threshold_get ("", vl->plugin, vl->plugin_instance,
205 vl->type, vl->type_instance)) != NULL)
206 return (th);
207 else if ((th = threshold_get ("", vl->plugin, vl->plugin_instance,
208 vl->type, NULL)) != NULL)
209 return (th);
210 else if ((th = threshold_get ("", vl->plugin, NULL,
211 vl->type, vl->type_instance)) != NULL)
212 return (th);
213 else if ((th = threshold_get ("", vl->plugin, NULL,
214 vl->type, NULL)) != NULL)
215 return (th);
216 else if ((th = threshold_get ("", "", NULL,
217 vl->type, vl->type_instance)) != NULL)
218 return (th);
219 else if ((th = threshold_get ("", "", NULL,
220 vl->type, NULL)) != NULL)
221 return (th);
223 return (NULL);
224 } /* }}} threshold_t *threshold_search */
226 /*
227 * Configuration
228 * =============
229 * The following approximately two hundred functions are used to handle the
230 * configuration and fill the threshold list.
231 * {{{ */
232 static int ut_config_type_datasource (threshold_t *th, oconfig_item_t *ci)
233 {
234 if ((ci->values_num != 1)
235 || (ci->values[0].type != OCONFIG_TYPE_STRING))
236 {
237 WARNING ("threshold values: The `DataSource' option needs exactly one "
238 "string argument.");
239 return (-1);
240 }
242 sstrncpy (th->data_source, ci->values[0].value.string,
243 sizeof (th->data_source));
245 return (0);
246 } /* int ut_config_type_datasource */
248 static int ut_config_type_instance (threshold_t *th, oconfig_item_t *ci)
249 {
250 if ((ci->values_num != 1)
251 || (ci->values[0].type != OCONFIG_TYPE_STRING))
252 {
253 WARNING ("threshold values: The `Instance' option needs exactly one "
254 "string argument.");
255 return (-1);
256 }
258 sstrncpy (th->type_instance, ci->values[0].value.string,
259 sizeof (th->type_instance));
261 return (0);
262 } /* int ut_config_type_instance */
264 static int ut_config_type_max (threshold_t *th, oconfig_item_t *ci)
265 {
266 if ((ci->values_num != 1)
267 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
268 {
269 WARNING ("threshold values: The `%s' option needs exactly one "
270 "number argument.", ci->key);
271 return (-1);
272 }
274 if (strcasecmp (ci->key, "WarningMax") == 0)
275 th->warning_max = ci->values[0].value.number;
276 else
277 th->failure_max = ci->values[0].value.number;
279 return (0);
280 } /* int ut_config_type_max */
282 static int ut_config_type_min (threshold_t *th, oconfig_item_t *ci)
283 {
284 if ((ci->values_num != 1)
285 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
286 {
287 WARNING ("threshold values: The `%s' option needs exactly one "
288 "number argument.", ci->key);
289 return (-1);
290 }
292 if (strcasecmp (ci->key, "WarningMin") == 0)
293 th->warning_min = ci->values[0].value.number;
294 else
295 th->failure_min = ci->values[0].value.number;
297 return (0);
298 } /* int ut_config_type_min */
300 static int ut_config_type_hits (threshold_t *th, oconfig_item_t *ci)
301 {
302 if ((ci->values_num != 1)
303 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
304 {
305 WARNING ("threshold values: The `%s' option needs exactly one "
306 "number argument.", ci->key);
307 return (-1);
308 }
310 th->hits = ci->values[0].value.number;
312 return (0);
313 } /* int ut_config_type_hits */
315 static int ut_config_type_hysteresis (threshold_t *th, oconfig_item_t *ci)
316 {
317 if ((ci->values_num != 1)
318 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
319 {
320 WARNING ("threshold values: The `%s' option needs exactly one "
321 "number argument.", ci->key);
322 return (-1);
323 }
325 th->hysteresis = ci->values[0].value.number;
327 return (0);
328 } /* int ut_config_type_hysteresis */
330 static int ut_config_type (const threshold_t *th_orig, oconfig_item_t *ci)
331 {
332 int i;
333 threshold_t th;
334 int status = 0;
336 if ((ci->values_num != 1)
337 || (ci->values[0].type != OCONFIG_TYPE_STRING))
338 {
339 WARNING ("threshold values: The `Type' block needs exactly one string "
340 "argument.");
341 return (-1);
342 }
344 if (ci->children_num < 1)
345 {
346 WARNING ("threshold values: The `Type' block needs at least one option.");
347 return (-1);
348 }
350 memcpy (&th, th_orig, sizeof (th));
351 sstrncpy (th.type, ci->values[0].value.string, sizeof (th.type));
353 th.warning_min = NAN;
354 th.warning_max = NAN;
355 th.failure_min = NAN;
356 th.failure_max = NAN;
357 th.hits = 0;
358 th.hysteresis = 0;
359 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
361 for (i = 0; i < ci->children_num; i++)
362 {
363 oconfig_item_t *option = ci->children + i;
364 status = 0;
366 if (strcasecmp ("Instance", option->key) == 0)
367 status = ut_config_type_instance (&th, option);
368 else if (strcasecmp ("DataSource", option->key) == 0)
369 status = ut_config_type_datasource (&th, option);
370 else if ((strcasecmp ("WarningMax", option->key) == 0)
371 || (strcasecmp ("FailureMax", option->key) == 0))
372 status = ut_config_type_max (&th, option);
373 else if ((strcasecmp ("WarningMin", option->key) == 0)
374 || (strcasecmp ("FailureMin", option->key) == 0))
375 status = ut_config_type_min (&th, option);
376 else if (strcasecmp ("Interesting", option->key) == 0)
377 status = cf_util_get_flag (option, &th.flags, UT_FLAG_INTERESTING);
378 else if (strcasecmp ("Invert", option->key) == 0)
379 status = cf_util_get_flag (option, &th.flags, UT_FLAG_INVERT);
380 else if (strcasecmp ("Persist", option->key) == 0)
381 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERSIST);
382 else if (strcasecmp ("PersistOK", option->key) == 0)
383 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERSIST_OK);
384 else if (strcasecmp ("Percentage", option->key) == 0)
385 status = cf_util_get_flag (option, &th.flags, UT_FLAG_PERCENTAGE);
386 else if (strcasecmp ("Hits", option->key) == 0)
387 status = ut_config_type_hits (&th, option);
388 else if (strcasecmp ("Hysteresis", option->key) == 0)
389 status = ut_config_type_hysteresis (&th, option);
390 else
391 {
392 WARNING ("threshold values: Option `%s' not allowed inside a `Type' "
393 "block.", option->key);
394 status = -1;
395 }
397 if (status != 0)
398 break;
399 }
401 if (status == 0)
402 {
403 status = ut_threshold_add (&th);
404 }
406 return (status);
407 } /* int ut_config_type */
409 static int ut_config_plugin_instance (threshold_t *th, oconfig_item_t *ci)
410 {
411 if ((ci->values_num != 1)
412 || (ci->values[0].type != OCONFIG_TYPE_STRING))
413 {
414 WARNING ("threshold values: The `Instance' option needs exactly one "
415 "string argument.");
416 return (-1);
417 }
419 sstrncpy (th->plugin_instance, ci->values[0].value.string,
420 sizeof (th->plugin_instance));
422 return (0);
423 } /* int ut_config_plugin_instance */
425 static int ut_config_plugin (const threshold_t *th_orig, oconfig_item_t *ci)
426 {
427 int i;
428 threshold_t th;
429 int status = 0;
431 if ((ci->values_num != 1)
432 || (ci->values[0].type != OCONFIG_TYPE_STRING))
433 {
434 WARNING ("threshold values: The `Plugin' block needs exactly one string "
435 "argument.");
436 return (-1);
437 }
439 if (ci->children_num < 1)
440 {
441 WARNING ("threshold values: The `Plugin' block needs at least one nested "
442 "block.");
443 return (-1);
444 }
446 memcpy (&th, th_orig, sizeof (th));
447 sstrncpy (th.plugin, ci->values[0].value.string, sizeof (th.plugin));
449 for (i = 0; i < ci->children_num; i++)
450 {
451 oconfig_item_t *option = ci->children + i;
452 status = 0;
454 if (strcasecmp ("Type", option->key) == 0)
455 status = ut_config_type (&th, option);
456 else if (strcasecmp ("Instance", option->key) == 0)
457 status = ut_config_plugin_instance (&th, option);
458 else
459 {
460 WARNING ("threshold values: Option `%s' not allowed inside a `Plugin' "
461 "block.", option->key);
462 status = -1;
463 }
465 if (status != 0)
466 break;
467 }
469 return (status);
470 } /* int ut_config_plugin */
472 static int ut_config_host (const threshold_t *th_orig, oconfig_item_t *ci)
473 {
474 int i;
475 threshold_t th;
476 int status = 0;
478 if ((ci->values_num != 1)
479 || (ci->values[0].type != OCONFIG_TYPE_STRING))
480 {
481 WARNING ("threshold values: The `Host' block needs exactly one string "
482 "argument.");
483 return (-1);
484 }
486 if (ci->children_num < 1)
487 {
488 WARNING ("threshold values: The `Host' block needs at least one nested "
489 "block.");
490 return (-1);
491 }
493 memcpy (&th, th_orig, sizeof (th));
494 sstrncpy (th.host, ci->values[0].value.string, sizeof (th.host));
496 for (i = 0; i < ci->children_num; i++)
497 {
498 oconfig_item_t *option = ci->children + i;
499 status = 0;
501 if (strcasecmp ("Type", option->key) == 0)
502 status = ut_config_type (&th, option);
503 else if (strcasecmp ("Plugin", option->key) == 0)
504 status = ut_config_plugin (&th, option);
505 else
506 {
507 WARNING ("threshold values: Option `%s' not allowed inside a `Host' "
508 "block.", option->key);
509 status = -1;
510 }
512 if (status != 0)
513 break;
514 }
516 return (status);
517 } /* int ut_config_host */
518 /*
519 * End of the functions used to configure threshold values.
520 */
521 /* }}} */
523 /*
524 * int ut_report_state
525 *
526 * Checks if the `state' differs from the old state and creates a notification
527 * if appropriate.
528 * Does not fail.
529 */
530 static int ut_report_state (const data_set_t *ds,
531 const value_list_t *vl,
532 const threshold_t *th,
533 const gauge_t *values,
534 int ds_index,
535 int state)
536 { /* {{{ */
537 int state_old;
538 notification_t n;
540 char *buf;
541 size_t bufsize;
543 int status;
545 /* Check if hits matched */
546 if ( (th->hits != 0) )
547 {
548 int hits = uc_get_hits(ds,vl);
549 /* STATE_OKAY resets hits unless PERSIST_OK flag is set. Hits resets if
550 * threshold is hit. */
551 if ( ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) ) || (hits > th->hits) )
552 {
553 DEBUG("ut_report_state: reset uc_get_hits = 0");
554 uc_set_hits(ds,vl,0); /* reset hit counter and notify */
555 } else {
556 DEBUG("ut_report_state: th->hits = %d, uc_get_hits = %d",th->hits,uc_get_hits(ds,vl));
557 (void) uc_inc_hits(ds,vl,1); /* increase hit counter */
558 return (0);
559 }
560 } /* end check hits */
562 state_old = uc_get_state (ds, vl);
564 /* If the state didn't change, report if `persistent' is specified. If the
565 * state is `okay', then only report if `persist_ok` flag is set. */
566 if (state == state_old)
567 {
568 if ((th->flags & UT_FLAG_PERSIST) == 0)
569 return (0);
570 else if ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) )
571 return (0);
572 }
574 if (state != state_old)
575 uc_set_state (ds, vl, state);
577 NOTIFICATION_INIT_VL (&n, vl);
579 buf = n.message;
580 bufsize = sizeof (n.message);
582 if (state == STATE_OKAY)
583 n.severity = NOTIF_OKAY;
584 else if (state == STATE_WARNING)
585 n.severity = NOTIF_WARNING;
586 else
587 n.severity = NOTIF_FAILURE;
589 n.time = vl->time;
591 status = ssnprintf (buf, bufsize, "Host %s, plugin %s",
592 vl->host, vl->plugin);
593 buf += status;
594 bufsize -= status;
596 if (vl->plugin_instance[0] != '\0')
597 {
598 status = ssnprintf (buf, bufsize, " (instance %s)",
599 vl->plugin_instance);
600 buf += status;
601 bufsize -= status;
602 }
604 status = ssnprintf (buf, bufsize, " type %s", vl->type);
605 buf += status;
606 bufsize -= status;
608 if (vl->type_instance[0] != '\0')
609 {
610 status = ssnprintf (buf, bufsize, " (instance %s)",
611 vl->type_instance);
612 buf += status;
613 bufsize -= status;
614 }
616 plugin_notification_meta_add_string (&n, "DataSource",
617 ds->ds[ds_index].name);
618 plugin_notification_meta_add_double (&n, "CurrentValue", values[ds_index]);
619 plugin_notification_meta_add_double (&n, "WarningMin", th->warning_min);
620 plugin_notification_meta_add_double (&n, "WarningMax", th->warning_max);
621 plugin_notification_meta_add_double (&n, "FailureMin", th->failure_min);
622 plugin_notification_meta_add_double (&n, "FailureMax", th->failure_max);
624 /* Send an okay notification */
625 if (state == STATE_OKAY)
626 {
627 if (state_old == STATE_MISSING)
628 status = ssnprintf (buf, bufsize,
629 ": Value is no longer missing.");
630 else
631 status = ssnprintf (buf, bufsize,
632 ": All data sources are within range again. "
633 "Current value of \"%s\" is %f.",
634 ds->ds[ds_index].name, values[ds_index]);
635 buf += status;
636 bufsize -= status;
637 }
638 else
639 {
640 double min;
641 double max;
643 min = (state == STATE_ERROR) ? th->failure_min : th->warning_min;
644 max = (state == STATE_ERROR) ? th->failure_max : th->warning_max;
646 if (th->flags & UT_FLAG_INVERT)
647 {
648 if (!isnan (min) && !isnan (max))
649 {
650 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
651 "%f. That is within the %s region of %f%s and %f%s.",
652 ds->ds[ds_index].name, values[ds_index],
653 (state == STATE_ERROR) ? "failure" : "warning",
654 min, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "",
655 max, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
656 }
657 else
658 {
659 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
660 "%f. That is %s the %s threshold of %f%s.",
661 ds->ds[ds_index].name, values[ds_index],
662 isnan (min) ? "below" : "above",
663 (state == STATE_ERROR) ? "failure" : "warning",
664 isnan (min) ? max : min,
665 ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
666 }
667 }
668 else if (th->flags & UT_FLAG_PERCENTAGE)
669 {
670 gauge_t value;
671 gauge_t sum;
672 int i;
674 sum = 0.0;
675 for (i = 0; i < vl->values_len; i++)
676 {
677 if (isnan (values[i]))
678 continue;
680 sum += values[i];
681 }
683 if (sum == 0.0)
684 value = NAN;
685 else
686 value = 100.0 * values[ds_index] / sum;
688 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
689 "%g (%.2f%%). That is %s the %s threshold of %.2f%%.",
690 ds->ds[ds_index].name, values[ds_index], value,
691 (value < min) ? "below" : "above",
692 (state == STATE_ERROR) ? "failure" : "warning",
693 (value < min) ? min : max);
694 }
695 else /* is not inverted */
696 {
697 status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
698 "%f. That is %s the %s threshold of %f.",
699 ds->ds[ds_index].name, values[ds_index],
700 (values[ds_index] < min) ? "below" : "above",
701 (state == STATE_ERROR) ? "failure" : "warning",
702 (values[ds_index] < min) ? min : max);
703 }
704 buf += status;
705 bufsize -= status;
706 }
708 plugin_dispatch_notification (&n);
710 plugin_notification_meta_free (n.meta);
711 return (0);
712 } /* }}} int ut_report_state */
714 /*
715 * int ut_check_one_data_source
716 *
717 * Checks one data source against the given threshold configuration. If the
718 * `DataSource' option is set in the threshold, and the name does NOT match,
719 * `okay' is returned. If the threshold does match, its failure and warning
720 * min and max values are checked and `failure' or `warning' is returned if
721 * appropriate.
722 * Does not fail.
723 */
724 static int ut_check_one_data_source (const data_set_t *ds,
725 const value_list_t __attribute__((unused)) *vl,
726 const threshold_t *th,
727 const gauge_t *values,
728 int ds_index)
729 { /* {{{ */
730 const char *ds_name;
731 int is_warning = 0;
732 int is_failure = 0;
733 int prev_state = STATE_OKAY;
735 /* check if this threshold applies to this data source */
736 if (ds != NULL)
737 {
738 ds_name = ds->ds[ds_index].name;
739 if ((th->data_source[0] != 0)
740 && (strcmp (ds_name, th->data_source) != 0))
741 return (STATE_OKAY);
742 }
744 if ((th->flags & UT_FLAG_INVERT) != 0)
745 {
746 is_warning--;
747 is_failure--;
748 }
750 /* XXX: This is an experimental code, not optimized, not fast, not reliable,
751 * and probably, do not work as you expect. Enjoy! :D */
752 if ( (th->hysteresis > 0) && ((prev_state = uc_get_state(ds,vl)) != STATE_OKAY) )
753 {
754 switch(prev_state)
755 {
756 case STATE_ERROR:
757 if ( (!isnan (th->failure_min) && ((th->failure_min + th->hysteresis) < values[ds_index])) ||
758 (!isnan (th->failure_max) && ((th->failure_max - th->hysteresis) > values[ds_index])) )
759 return (STATE_OKAY);
760 else
761 is_failure++;
762 case STATE_WARNING:
763 if ( (!isnan (th->warning_min) && ((th->warning_min + th->hysteresis) < values[ds_index])) ||
764 (!isnan (th->warning_max) && ((th->warning_max - th->hysteresis) > values[ds_index])) )
765 return (STATE_OKAY);
766 else
767 is_warning++;
768 }
769 }
770 else { /* no hysteresis */
771 if ((!isnan (th->failure_min) && (th->failure_min > values[ds_index]))
772 || (!isnan (th->failure_max) && (th->failure_max < values[ds_index])))
773 is_failure++;
775 if ((!isnan (th->warning_min) && (th->warning_min > values[ds_index]))
776 || (!isnan (th->warning_max) && (th->warning_max < values[ds_index])))
777 is_warning++;
778 }
780 if (is_failure != 0)
781 return (STATE_ERROR);
783 if (is_warning != 0)
784 return (STATE_WARNING);
786 return (STATE_OKAY);
787 } /* }}} int ut_check_one_data_source */
789 /*
790 * int ut_check_one_threshold
791 *
792 * Checks all data sources of a value list against the given threshold, using
793 * the ut_check_one_data_source function above. Returns the worst status,
794 * which is `okay' if nothing has failed.
795 * Returns less than zero if the data set doesn't have any data sources.
796 */
797 static int ut_check_one_threshold (const data_set_t *ds,
798 const value_list_t *vl,
799 const threshold_t *th,
800 const gauge_t *values,
801 int *ret_ds_index)
802 { /* {{{ */
803 int ret = -1;
804 int ds_index = -1;
805 int i;
806 gauge_t values_copy[ds->ds_num];
808 memcpy (values_copy, values, sizeof (values_copy));
810 if ((th->flags & UT_FLAG_PERCENTAGE) != 0)
811 {
812 int num = 0;
813 gauge_t sum=0.0;
815 if (ds->ds_num == 1)
816 {
817 WARNING ("ut_check_one_threshold: The %s type has only one data "
818 "source, but you have configured to check this as a percentage. "
819 "That doesn't make much sense, because the percentage will always "
820 "be 100%%!", ds->type);
821 }
823 /* Prepare `sum' and `num'. */
824 for (i = 0; i < ds->ds_num; i++)
825 if (!isnan (values[i]))
826 {
827 num++;
828 sum += values[i];
829 }
831 if ((num == 0) /* All data sources are undefined. */
832 || (sum == 0.0)) /* Sum is zero, cannot calculate percentage. */
833 {
834 for (i = 0; i < ds->ds_num; i++)
835 values_copy[i] = NAN;
836 }
837 else /* We can actually calculate the percentage. */
838 {
839 for (i = 0; i < ds->ds_num; i++)
840 values_copy[i] = 100.0 * values[i] / sum;
841 }
842 } /* if (UT_FLAG_PERCENTAGE) */
844 for (i = 0; i < ds->ds_num; i++)
845 {
846 int status;
848 status = ut_check_one_data_source (ds, vl, th, values_copy, i);
849 if (ret < status)
850 {
851 ret = status;
852 ds_index = i;
853 }
854 } /* for (ds->ds_num) */
856 if (ret_ds_index != NULL)
857 *ret_ds_index = ds_index;
859 return (ret);
860 } /* }}} int ut_check_one_threshold */
862 /*
863 * int ut_check_threshold
864 *
865 * Gets a list of matching thresholds and searches for the worst status by one
866 * of the thresholds. Then reports that status using the ut_report_state
867 * function above.
868 * Returns zero on success and if no threshold has been configured. Returns
869 * less than zero on failure.
870 */
871 static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl,
872 __attribute__((unused)) user_data_t *ud)
873 { /* {{{ */
874 threshold_t *th;
875 gauge_t *values;
876 int status;
878 int worst_state = -1;
879 threshold_t *worst_th = NULL;
880 int worst_ds_index = -1;
882 if (threshold_tree == NULL)
883 return (0);
885 /* Is this lock really necessary? So far, thresholds are only inserted at
886 * startup. -octo */
887 pthread_mutex_lock (&threshold_lock);
888 th = threshold_search (vl);
889 pthread_mutex_unlock (&threshold_lock);
890 if (th == NULL)
891 return (0);
893 DEBUG ("ut_check_threshold: Found matching threshold(s)");
895 values = uc_get_rate (ds, vl);
896 if (values == NULL)
897 return (0);
899 while (th != NULL)
900 {
901 int ds_index = -1;
903 status = ut_check_one_threshold (ds, vl, th, values, &ds_index);
904 if (status < 0)
905 {
906 ERROR ("ut_check_threshold: ut_check_one_threshold failed.");
907 sfree (values);
908 return (-1);
909 }
911 if (worst_state < status)
912 {
913 worst_state = status;
914 worst_th = th;
915 worst_ds_index = ds_index;
916 }
918 th = th->next;
919 } /* while (th) */
921 status = ut_report_state (ds, vl, worst_th, values,
922 worst_ds_index, worst_state);
923 if (status != 0)
924 {
925 ERROR ("ut_check_threshold: ut_report_state failed.");
926 sfree (values);
927 return (-1);
928 }
930 sfree (values);
932 return (0);
933 } /* }}} int ut_check_threshold */
935 /*
936 * int ut_missing
937 *
938 * This function is called whenever a value goes "missing".
939 */
940 static int ut_missing (const value_list_t *vl,
941 __attribute__((unused)) user_data_t *ud)
942 { /* {{{ */
943 threshold_t *th;
944 cdtime_t missing_time;
945 char identifier[6 * DATA_MAX_NAME_LEN];
946 notification_t n;
948 if (threshold_tree == NULL)
949 return (0);
951 th = threshold_search (vl);
952 /* dispatch notifications for "interesting" values only */
953 if ((th == NULL) || ((th->flags & UT_FLAG_INTERESTING) == 0))
954 return (0);
956 missing_time = cdtime () - vl->time;
957 FORMAT_VL (identifier, sizeof (identifier), vl);
959 NOTIFICATION_INIT_VL (&n, vl);
960 ssnprintf (n.message, sizeof (n.message),
961 "%s has not been updated for %.3f seconds.",
962 identifier, CDTIME_T_TO_DOUBLE (missing_time));
964 plugin_dispatch_notification (&n);
966 return (0);
967 } /* }}} int ut_missing */
969 int ut_config (oconfig_item_t *ci)
970 { /* {{{ */
971 int i;
972 int status = 0;
974 threshold_t th;
976 if (threshold_tree == NULL)
977 {
978 threshold_tree = c_avl_create ((void *) strcmp);
979 if (threshold_tree == NULL)
980 {
981 ERROR ("ut_config: c_avl_create failed.");
982 return (-1);
983 }
984 }
986 memset (&th, '\0', sizeof (th));
987 th.warning_min = NAN;
988 th.warning_max = NAN;
989 th.failure_min = NAN;
990 th.failure_max = NAN;
992 th.hits = 0;
993 th.hysteresis = 0;
994 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
996 for (i = 0; i < ci->children_num; i++)
997 {
998 oconfig_item_t *option = ci->children + i;
999 status = 0;
1001 if (strcasecmp ("Type", option->key) == 0)
1002 status = ut_config_type (&th, option);
1003 else if (strcasecmp ("Plugin", option->key) == 0)
1004 status = ut_config_plugin (&th, option);
1005 else if (strcasecmp ("Host", option->key) == 0)
1006 status = ut_config_host (&th, option);
1007 else
1008 {
1009 WARNING ("threshold values: Option `%s' not allowed here.", option->key);
1010 status = -1;
1011 }
1013 if (status != 0)
1014 break;
1015 }
1017 if (c_avl_size (threshold_tree) > 0) {
1018 plugin_register_missing ("threshold", ut_missing,
1019 /* user data = */ NULL);
1020 plugin_register_write ("threshold", ut_check_threshold,
1021 /* user data = */ NULL);
1022 }
1024 return (status);
1025 } /* }}} int um_config */
1027 void module_register (void)
1028 {
1029 plugin_register_complex_config ("threshold", ut_config);
1030 }
1032 /* vim: set sw=2 ts=8 sts=2 tw=78 et fdm=marker : */