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
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.
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.
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
21 * Florian octo Forster <octo at collectd.org>
22 * Sebastian Harl <sh at tokkee.org>
23 * Andrés J. Díaz <ajdiaz at connectical.com>
29 #include "utils_avltree.h"
30 #include "utils_cache.h"
31 #include "utils_threshold.h"
34 * Threshold management
35 * ====================
36 * The following functions add, delete, search, etc. configured thresholds to
37 * the underlying AVL trees.
41 * int ut_threshold_add
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.
47 static int ut_threshold_add (const threshold_t *th)
49 char name[6 * DATA_MAX_NAME_LEN];
55 if (format_name (name, sizeof (name), th->host,
56 th->plugin, th->plugin_instance,
57 th->type, th->type_instance) != 0)
59 ERROR ("ut_threshold_add: format_name failed.");
63 name_copy = strdup (name);
64 if (name_copy == NULL)
66 ERROR ("ut_threshold_add: strdup failed.");
70 th_copy = malloc (sizeof (*th_copy));
74 ERROR ("ut_threshold_add: malloc failed.");
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 */
91 status = c_avl_insert (threshold_tree, name_copy, th_copy);
93 else /* th_ptr points to the last threshold in the list */
95 th_ptr->next = th_copy;
96 /* name_copy isn't needed */
100 pthread_mutex_unlock (&threshold_lock);
104 ERROR ("ut_threshold_add: c_avl_insert (%s) failed.", name);
110 } /* }}} int ut_threshold_add */
115 * The following approximately two hundred functions are used to handle the
116 * configuration and fill the threshold list.
118 static int ut_config_type_datasource (threshold_t *th, oconfig_item_t *ci)
120 if ((ci->values_num != 1)
121 || (ci->values[0].type != OCONFIG_TYPE_STRING))
123 WARNING ("threshold values: The `DataSource' option needs exactly one "
128 sstrncpy (th->data_source, ci->values[0].value.string,
129 sizeof (th->data_source));
132 } /* int ut_config_type_datasource */
134 static int ut_config_type_instance (threshold_t *th, oconfig_item_t *ci)
136 if ((ci->values_num != 1)
137 || (ci->values[0].type != OCONFIG_TYPE_STRING))
139 WARNING ("threshold values: The `Instance' option needs exactly one "
144 sstrncpy (th->type_instance, ci->values[0].value.string,
145 sizeof (th->type_instance));
148 } /* int ut_config_type_instance */
150 static int ut_config_type_max (threshold_t *th, oconfig_item_t *ci)
152 if ((ci->values_num != 1)
153 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
155 WARNING ("threshold values: The `%s' option needs exactly one "
156 "number argument.", ci->key);
160 if (strcasecmp (ci->key, "WarningMax") == 0)
161 th->warning_max = ci->values[0].value.number;
163 th->failure_max = ci->values[0].value.number;
166 } /* int ut_config_type_max */
168 static int ut_config_type_min (threshold_t *th, oconfig_item_t *ci)
170 if ((ci->values_num != 1)
171 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
173 WARNING ("threshold values: The `%s' option needs exactly one "
174 "number argument.", ci->key);
178 if (strcasecmp (ci->key, "WarningMin") == 0)
179 th->warning_min = ci->values[0].value.number;
181 th->failure_min = ci->values[0].value.number;
184 } /* int ut_config_type_min */
186 static int ut_config_type_hits (threshold_t *th, oconfig_item_t *ci)
188 if ((ci->values_num != 1)
189 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
191 WARNING ("threshold values: The `%s' option needs exactly one "
192 "number argument.", ci->key);
196 th->hits = ci->values[0].value.number;
199 } /* int ut_config_type_hits */
201 static int ut_config_type_hysteresis (threshold_t *th, oconfig_item_t *ci)
203 if ((ci->values_num != 1)
204 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
206 WARNING ("threshold values: The `%s' option needs exactly one "
207 "number argument.", ci->key);
211 th->hysteresis = ci->values[0].value.number;
214 } /* int ut_config_type_hysteresis */
216 static int ut_config_type (const threshold_t *th_orig, oconfig_item_t *ci)
222 if ((ci->values_num != 1)
223 || (ci->values[0].type != OCONFIG_TYPE_STRING))
225 WARNING ("threshold values: The `Type' block needs exactly one string "
230 if (ci->children_num < 1)
232 WARNING ("threshold values: The `Type' block needs at least one option.");
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;
245 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
247 for (i = 0; i < ci->children_num; i++)
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);
277 WARNING ("threshold values: Option `%s' not allowed inside a `Type' "
278 "block.", option->key);
288 status = ut_threshold_add (&th);
292 } /* int ut_config_type */
294 static int ut_config_plugin_instance (threshold_t *th, oconfig_item_t *ci)
296 if ((ci->values_num != 1)
297 || (ci->values[0].type != OCONFIG_TYPE_STRING))
299 WARNING ("threshold values: The `Instance' option needs exactly one "
304 sstrncpy (th->plugin_instance, ci->values[0].value.string,
305 sizeof (th->plugin_instance));
308 } /* int ut_config_plugin_instance */
310 static int ut_config_plugin (const threshold_t *th_orig, oconfig_item_t *ci)
316 if ((ci->values_num != 1)
317 || (ci->values[0].type != OCONFIG_TYPE_STRING))
319 WARNING ("threshold values: The `Plugin' block needs exactly one string "
324 if (ci->children_num < 1)
326 WARNING ("threshold values: The `Plugin' block needs at least one nested "
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++)
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);
344 WARNING ("threshold values: Option `%s' not allowed inside a `Plugin' "
345 "block.", option->key);
354 } /* int ut_config_plugin */
356 static int ut_config_host (const threshold_t *th_orig, oconfig_item_t *ci)
362 if ((ci->values_num != 1)
363 || (ci->values[0].type != OCONFIG_TYPE_STRING))
365 WARNING ("threshold values: The `Host' block needs exactly one string "
370 if (ci->children_num < 1)
372 WARNING ("threshold values: The `Host' block needs at least one nested "
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++)
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);
390 WARNING ("threshold values: Option `%s' not allowed inside a `Host' "
391 "block.", option->key);
400 } /* int ut_config_host */
402 * End of the functions used to configure threshold values.
407 * int ut_report_state
409 * Checks if the `state' differs from the old state and creates a notification
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,
428 /* Check if hits matched */
429 if ( (th->hits != 0) )
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) )
436 DEBUG("ut_report_state: reset uc_get_hits = 0");
437 uc_set_hits(ds,vl,0); /* reset hit counter and notify */
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 */
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)
451 if ((th->flags & UT_FLAG_PERSIST) == 0)
453 else if ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) )
457 if (state != state_old)
458 uc_set_state (ds, vl, state);
460 NOTIFICATION_INIT_VL (&n, vl);
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;
470 n.severity = NOTIF_FAILURE;
474 status = ssnprintf (buf, bufsize, "Host %s, plugin %s",
475 vl->host, vl->plugin);
479 if (vl->plugin_instance[0] != '\0')
481 status = ssnprintf (buf, bufsize, " (instance %s)",
482 vl->plugin_instance);
487 status = ssnprintf (buf, bufsize, " type %s", vl->type);
491 if (vl->type_instance[0] != '\0')
493 status = ssnprintf (buf, bufsize, " (instance %s)",
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)
510 if (state_old == STATE_MISSING)
511 ssnprintf (buf, bufsize, ": Value is no longer missing.");
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]);
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)
528 if (!isnan (min) && !isnan (max))
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) ? "%" : "");
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) ? "%" : "");
548 else if (th->flags & UT_FLAG_PERCENTAGE)
555 for (i = 0; i < vl->values_len; i++)
557 if (isnan (values[i]))
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);
575 else /* is not inverted */
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);
586 plugin_dispatch_notification (&n);
588 plugin_notification_meta_free (n.meta);
590 } /* }}} int ut_report_state */
593 * int ut_check_one_data_source
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
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,
611 int prev_state = STATE_OKAY;
613 /* check if this threshold applies to this data source */
616 ds_name = ds->ds[ds_index].name;
617 if ((th->data_source[0] != 0)
618 && (strcmp (ds_name, th->data_source) != 0))
622 if ((th->flags & UT_FLAG_INVERT) != 0)
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)
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.
640 * There is no hysteresis for the OKAY state.
642 gauge_t hysteresis_for_warning = 0, hysteresis_for_failure = 0;
646 hysteresis_for_failure = th->hysteresis;
649 hysteresis_for_warning = th->hysteresis;
652 /* do nothing -- the hysteresis only applies to the non-normal states */
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])))
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])))
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])))
670 if ((!isnan (th->warning_min) && (th->warning_min > values[ds_index]))
671 || (!isnan (th->warning_max) && (th->warning_max < values[ds_index])))
676 return (STATE_ERROR);
679 return (STATE_WARNING);
682 } /* }}} int ut_check_one_data_source */
685 * int ut_check_one_threshold
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.
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,
701 gauge_t values_copy[ds->ds_num];
703 memcpy (values_copy, values, sizeof (values_copy));
705 if ((th->flags & UT_FLAG_PERCENTAGE) != 0)
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);
718 /* Prepare `sum' and `num'. */
719 for (i = 0; i < ds->ds_num; i++)
720 if (!isnan (values[i]))
726 if ((num == 0) /* All data sources are undefined. */
727 || (sum == 0.0)) /* Sum is zero, cannot calculate percentage. */
729 for (i = 0; i < ds->ds_num; i++)
730 values_copy[i] = NAN;
732 else /* We can actually calculate the percentage. */
734 for (i = 0; i < ds->ds_num; i++)
735 values_copy[i] = 100.0 * values[i] / sum;
737 } /* if (UT_FLAG_PERCENTAGE) */
739 for (i = 0; i < ds->ds_num; i++)
743 status = ut_check_one_data_source (ds, vl, th, values_copy, i);
749 } /* for (ds->ds_num) */
751 if (ret_ds_index != NULL)
752 *ret_ds_index = ds_index;
755 } /* }}} int ut_check_one_threshold */
758 * int ut_check_threshold
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
763 * Returns zero on success and if no threshold has been configured. Returns
764 * less than zero on failure.
766 static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl,
767 __attribute__((unused)) user_data_t *ud)
773 int worst_state = -1;
774 threshold_t *worst_th = NULL;
775 int worst_ds_index = -1;
777 if (threshold_tree == NULL)
780 /* Is this lock really necessary? So far, thresholds are only inserted at
782 pthread_mutex_lock (&threshold_lock);
783 th = threshold_search (vl);
784 pthread_mutex_unlock (&threshold_lock);
788 DEBUG ("ut_check_threshold: Found matching threshold(s)");
790 values = uc_get_rate (ds, vl);
798 status = ut_check_one_threshold (ds, vl, th, values, &ds_index);
801 ERROR ("ut_check_threshold: ut_check_one_threshold failed.");
806 if (worst_state < status)
808 worst_state = status;
810 worst_ds_index = ds_index;
816 status = ut_report_state (ds, vl, worst_th, values,
817 worst_ds_index, worst_state);
820 ERROR ("ut_check_threshold: ut_report_state failed.");
828 } /* }}} int ut_check_threshold */
833 * This function is called whenever a value goes "missing".
835 static int ut_missing (const value_list_t *vl,
836 __attribute__((unused)) user_data_t *ud)
839 cdtime_t missing_time;
840 char identifier[6 * DATA_MAX_NAME_LEN];
844 if (threshold_tree == NULL)
847 th = threshold_search (vl);
848 /* dispatch notifications for "interesting" values only */
849 if ((th == NULL) || ((th->flags & UT_FLAG_INTERESTING) == 0))
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));
862 plugin_dispatch_notification (&n);
865 } /* }}} int ut_missing */
867 static int ut_config (oconfig_item_t *ci)
871 int old_size = c_avl_size (threshold_tree);
875 if (threshold_tree == NULL)
877 threshold_tree = c_avl_create ((void *) strcmp);
878 if (threshold_tree == NULL)
880 ERROR ("ut_config: c_avl_create failed.");
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;
893 th.flags = UT_FLAG_INTERESTING; /* interesting by default */
895 for (i = 0; i < ci->children_num; i++)
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);
907 WARNING ("threshold values: Option `%s' not allowed here.", option->key);
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))
918 plugin_register_missing ("threshold", ut_missing,
919 /* user data = */ NULL);
920 plugin_register_write ("threshold", ut_check_threshold,
921 /* user data = */ NULL);
925 } /* }}} int um_config */
927 void module_register (void)
929 plugin_register_complex_config ("threshold", ut_config);
932 /* vim: set sw=2 ts=8 sts=2 tw=78 et fdm=marker : */