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>
30 #include "utils_avltree.h"
31 #include "utils_cache.h"
32 #include "utils_threshold.h"
35 * Threshold management
36 * ====================
37 * The following functions add, delete, search, etc. configured thresholds to
38 * the underlying AVL trees.
42 * int ut_threshold_add
44 * Adds a threshold configuration to the list of thresholds. The threshold_t
45 * structure is copied and may be destroyed after this call. Returns zero on
46 * success, non-zero otherwise.
48 static int ut_threshold_add (const threshold_t *th)
50 char name[6 * DATA_MAX_NAME_LEN];
56 if (format_name (name, sizeof (name), th->host,
57 th->plugin, th->plugin_instance,
58 th->type, th->type_instance) != 0)
60 ERROR ("ut_threshold_add: format_name failed.");
64 name_copy = strdup (name);
65 if (name_copy == NULL)
67 ERROR ("ut_threshold_add: strdup failed.");
71 th_copy = malloc (sizeof (*th_copy));
75 ERROR ("ut_threshold_add: malloc failed.");
78 memcpy (th_copy, th, sizeof (threshold_t));
80 DEBUG ("ut_threshold_add: Adding entry `%s'", name);
82 pthread_mutex_lock (&threshold_lock);
84 th_ptr = threshold_get (th->host, th->plugin, th->plugin_instance,
85 th->type, th->type_instance);
87 while ((th_ptr != NULL) && (th_ptr->next != NULL))
88 th_ptr = th_ptr->next;
90 if (th_ptr == NULL) /* no such threshold yet */
92 status = c_avl_insert (threshold_tree, name_copy, th_copy);
94 else /* th_ptr points to the last threshold in the list */
96 th_ptr->next = th_copy;
97 /* name_copy isn't needed */
101 pthread_mutex_unlock (&threshold_lock);
105 ERROR ("ut_threshold_add: c_avl_insert (%s) failed.", name);
111 } /* }}} int ut_threshold_add */
116 * The following approximately two hundred functions are used to handle the
117 * configuration and fill the threshold list.
119 static int ut_config_type_datasource (threshold_t *th, oconfig_item_t *ci)
121 if ((ci->values_num != 1)
122 || (ci->values[0].type != OCONFIG_TYPE_STRING))
124 WARNING ("threshold values: The `DataSource' option needs exactly one "
129 sstrncpy (th->data_source, ci->values[0].value.string,
130 sizeof (th->data_source));
133 } /* int ut_config_type_datasource */
135 static int ut_config_type_instance (threshold_t *th, oconfig_item_t *ci)
137 if ((ci->values_num != 1)
138 || (ci->values[0].type != OCONFIG_TYPE_STRING))
140 WARNING ("threshold values: The `Instance' option needs exactly one "
145 sstrncpy (th->type_instance, ci->values[0].value.string,
146 sizeof (th->type_instance));
149 } /* int ut_config_type_instance */
151 static int ut_config_type_max (threshold_t *th, oconfig_item_t *ci)
153 if ((ci->values_num != 1)
154 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
156 WARNING ("threshold values: The `%s' option needs exactly one "
157 "number argument.", ci->key);
161 if (strcasecmp (ci->key, "WarningMax") == 0)
162 th->warning_max = ci->values[0].value.number;
164 th->failure_max = ci->values[0].value.number;
167 } /* int ut_config_type_max */
169 static int ut_config_type_min (threshold_t *th, oconfig_item_t *ci)
171 if ((ci->values_num != 1)
172 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
174 WARNING ("threshold values: The `%s' option needs exactly one "
175 "number argument.", ci->key);
179 if (strcasecmp (ci->key, "WarningMin") == 0)
180 th->warning_min = ci->values[0].value.number;
182 th->failure_min = ci->values[0].value.number;
185 } /* int ut_config_type_min */
187 static int ut_config_type_hits (threshold_t *th, oconfig_item_t *ci)
189 if ((ci->values_num != 1)
190 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
192 WARNING ("threshold values: The `%s' option needs exactly one "
193 "number argument.", ci->key);
197 th->hits = ci->values[0].value.number;
200 } /* int ut_config_type_hits */
202 static int ut_config_type_hysteresis (threshold_t *th, oconfig_item_t *ci)
204 if ((ci->values_num != 1)
205 || (ci->values[0].type != OCONFIG_TYPE_NUMBER))
207 WARNING ("threshold values: The `%s' option needs exactly one "
208 "number argument.", ci->key);
212 th->hysteresis = ci->values[0].value.number;
215 } /* int ut_config_type_hysteresis */
217 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 (int 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)
315 if ((ci->values_num != 1)
316 || (ci->values[0].type != OCONFIG_TYPE_STRING))
318 WARNING ("threshold values: The `Plugin' block needs exactly one string "
323 if (ci->children_num < 1)
325 WARNING ("threshold values: The `Plugin' block needs at least one nested "
330 memcpy (&th, th_orig, sizeof (th));
331 sstrncpy (th.plugin, ci->values[0].value.string, sizeof (th.plugin));
333 for (int i = 0; i < ci->children_num; i++)
335 oconfig_item_t *option = ci->children + i;
337 if (strcasecmp ("Type", option->key) == 0)
338 status = ut_config_type (&th, option);
339 else if (strcasecmp ("Instance", option->key) == 0)
340 status = ut_config_plugin_instance (&th, option);
343 WARNING ("threshold values: Option `%s' not allowed inside a `Plugin' "
344 "block.", option->key);
353 } /* int ut_config_plugin */
355 static int ut_config_host (const threshold_t *th_orig, oconfig_item_t *ci)
360 if ((ci->values_num != 1)
361 || (ci->values[0].type != OCONFIG_TYPE_STRING))
363 WARNING ("threshold values: The `Host' block needs exactly one string "
368 if (ci->children_num < 1)
370 WARNING ("threshold values: The `Host' block needs at least one nested "
375 memcpy (&th, th_orig, sizeof (th));
376 sstrncpy (th.host, ci->values[0].value.string, sizeof (th.host));
378 for (int i = 0; i < ci->children_num; i++)
380 oconfig_item_t *option = ci->children + i;
382 if (strcasecmp ("Type", option->key) == 0)
383 status = ut_config_type (&th, option);
384 else if (strcasecmp ("Plugin", option->key) == 0)
385 status = ut_config_plugin (&th, option);
388 WARNING ("threshold values: Option `%s' not allowed inside a `Host' "
389 "block.", option->key);
398 } /* int ut_config_host */
400 * End of the functions used to configure threshold values.
405 * int ut_report_state
407 * Checks if the `state' differs from the old state and creates a notification
411 static int ut_report_state (const data_set_t *ds,
412 const value_list_t *vl,
413 const threshold_t *th,
414 const gauge_t *values,
426 /* Check if hits matched */
427 if ( (th->hits != 0) )
429 int hits = uc_get_hits(ds,vl);
430 /* STATE_OKAY resets hits unless PERSIST_OK flag is set. Hits resets if
431 * threshold is hit. */
432 if ( ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) ) || (hits > th->hits) )
434 DEBUG("ut_report_state: reset uc_get_hits = 0");
435 uc_set_hits(ds,vl,0); /* reset hit counter and notify */
437 DEBUG("ut_report_state: th->hits = %d, uc_get_hits = %d",th->hits,uc_get_hits(ds,vl));
438 (void) uc_inc_hits(ds,vl,1); /* increase hit counter */
441 } /* end check hits */
443 state_old = uc_get_state (ds, vl);
445 /* If the state didn't change, report if `persistent' is specified. If the
446 * state is `okay', then only report if `persist_ok` flag is set. */
447 if (state == state_old)
449 if ((th->flags & UT_FLAG_PERSIST) == 0)
451 else if ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) )
455 if (state != state_old)
456 uc_set_state (ds, vl, state);
458 NOTIFICATION_INIT_VL (&n, vl);
461 bufsize = sizeof (n.message);
463 if (state == STATE_OKAY)
464 n.severity = NOTIF_OKAY;
465 else if (state == STATE_WARNING)
466 n.severity = NOTIF_WARNING;
468 n.severity = NOTIF_FAILURE;
472 status = ssnprintf (buf, bufsize, "Host %s, plugin %s",
473 vl->host, vl->plugin);
477 if (vl->plugin_instance[0] != '\0')
479 status = ssnprintf (buf, bufsize, " (instance %s)",
480 vl->plugin_instance);
485 status = ssnprintf (buf, bufsize, " type %s", vl->type);
489 if (vl->type_instance[0] != '\0')
491 status = ssnprintf (buf, bufsize, " (instance %s)",
497 plugin_notification_meta_add_string (&n, "DataSource",
498 ds->ds[ds_index].name);
499 plugin_notification_meta_add_double (&n, "CurrentValue", values[ds_index]);
500 plugin_notification_meta_add_double (&n, "WarningMin", th->warning_min);
501 plugin_notification_meta_add_double (&n, "WarningMax", th->warning_max);
502 plugin_notification_meta_add_double (&n, "FailureMin", th->failure_min);
503 plugin_notification_meta_add_double (&n, "FailureMax", th->failure_max);
505 /* Send an okay notification */
506 if (state == STATE_OKAY)
508 if (state_old == STATE_MISSING)
509 ssnprintf (buf, bufsize, ": Value is no longer missing.");
511 ssnprintf (buf, bufsize,
512 ": All data sources are within range again. "
513 "Current value of \"%s\" is %f.",
514 ds->ds[ds_index].name, values[ds_index]);
521 min = (state == STATE_ERROR) ? th->failure_min : th->warning_min;
522 max = (state == STATE_ERROR) ? th->failure_max : th->warning_max;
524 if (th->flags & UT_FLAG_INVERT)
526 if (!isnan (min) && !isnan (max))
528 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
529 "%f. That is within the %s region of %f%s and %f%s.",
530 ds->ds[ds_index].name, values[ds_index],
531 (state == STATE_ERROR) ? "failure" : "warning",
532 min, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "",
533 max, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
537 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
538 "%f. That is %s the %s threshold of %f%s.",
539 ds->ds[ds_index].name, values[ds_index],
540 isnan (min) ? "below" : "above",
541 (state == STATE_ERROR) ? "failure" : "warning",
542 isnan (min) ? max : min,
543 ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "");
546 else if (th->flags & UT_FLAG_PERCENTAGE)
552 for (size_t i = 0; i < vl->values_len; i++)
554 if (isnan (values[i]))
563 value = 100.0 * values[ds_index] / sum;
565 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
566 "%g (%.2f%%). That is %s the %s threshold of %.2f%%.",
567 ds->ds[ds_index].name, values[ds_index], value,
568 (value < min) ? "below" : "above",
569 (state == STATE_ERROR) ? "failure" : "warning",
570 (value < min) ? min : max);
572 else /* is not inverted */
574 ssnprintf (buf, bufsize, ": Data source \"%s\" is currently "
575 "%f. That is %s the %s threshold of %f.",
576 ds->ds[ds_index].name, values[ds_index],
577 (values[ds_index] < min) ? "below" : "above",
578 (state == STATE_ERROR) ? "failure" : "warning",
579 (values[ds_index] < min) ? min : max);
583 plugin_dispatch_notification (&n);
585 plugin_notification_meta_free (n.meta);
587 } /* }}} int ut_report_state */
590 * int ut_check_one_data_source
592 * Checks one data source against the given threshold configuration. If the
593 * `DataSource' option is set in the threshold, and the name does NOT match,
594 * `okay' is returned. If the threshold does match, its failure and warning
595 * min and max values are checked and `failure' or `warning' is returned if
599 static int ut_check_one_data_source (const data_set_t *ds,
600 const value_list_t __attribute__((unused)) *vl,
601 const threshold_t *th,
602 const gauge_t *values,
608 int prev_state = STATE_OKAY;
610 /* check if this threshold applies to this data source */
613 ds_name = ds->ds[ds_index].name;
614 if ((th->data_source[0] != 0)
615 && (strcmp (ds_name, th->data_source) != 0))
619 if ((th->flags & UT_FLAG_INVERT) != 0)
625 /* XXX: This is an experimental code, not optimized, not fast, not reliable,
626 * and probably, do not work as you expect. Enjoy! :D */
627 if (th->hysteresis > 0)
629 prev_state = uc_get_state(ds,vl);
630 /* The purpose of hysteresis is elliminating flapping state when the value
631 * oscilates around the thresholds. In other words, what is important is
632 * the previous state; if the new value would trigger a transition, make
633 * sure that we artificially widen the range which is considered to apply
634 * for the previous state, and only trigger the notification if the value
635 * is outside of this expanded range.
637 * There is no hysteresis for the OKAY state.
639 gauge_t hysteresis_for_warning = 0, hysteresis_for_failure = 0;
643 hysteresis_for_failure = th->hysteresis;
646 hysteresis_for_warning = th->hysteresis;
649 /* do nothing -- the hysteresis only applies to the non-normal states */
653 if ((!isnan (th->failure_min) && (th->failure_min + hysteresis_for_failure > values[ds_index]))
654 || (!isnan (th->failure_max) && (th->failure_max - hysteresis_for_failure < values[ds_index])))
657 if ((!isnan (th->warning_min) && (th->warning_min + hysteresis_for_warning > values[ds_index]))
658 || (!isnan (th->warning_max) && (th->warning_max - hysteresis_for_warning < values[ds_index])))
662 else { /* no hysteresis */
663 if ((!isnan (th->failure_min) && (th->failure_min > values[ds_index]))
664 || (!isnan (th->failure_max) && (th->failure_max < values[ds_index])))
667 if ((!isnan (th->warning_min) && (th->warning_min > values[ds_index]))
668 || (!isnan (th->warning_max) && (th->warning_max < values[ds_index])))
673 return (STATE_ERROR);
676 return (STATE_WARNING);
679 } /* }}} int ut_check_one_data_source */
682 * int ut_check_one_threshold
684 * Checks all data sources of a value list against the given threshold, using
685 * the ut_check_one_data_source function above. Returns the worst status,
686 * which is `okay' if nothing has failed.
687 * Returns less than zero if the data set doesn't have any data sources.
689 static int ut_check_one_threshold (const data_set_t *ds,
690 const value_list_t *vl,
691 const threshold_t *th,
692 const gauge_t *values,
697 gauge_t values_copy[ds->ds_num];
699 memcpy (values_copy, values, sizeof (values_copy));
701 if ((th->flags & UT_FLAG_PERCENTAGE) != 0)
708 WARNING ("ut_check_one_threshold: The %s type has only one data "
709 "source, but you have configured to check this as a percentage. "
710 "That doesn't make much sense, because the percentage will always "
711 "be 100%%!", ds->type);
714 /* Prepare `sum' and `num'. */
715 for (size_t i = 0; i < ds->ds_num; i++)
716 if (!isnan (values[i]))
722 if ((num == 0) /* All data sources are undefined. */
723 || (sum == 0.0)) /* Sum is zero, cannot calculate percentage. */
725 for (size_t i = 0; i < ds->ds_num; i++)
726 values_copy[i] = NAN;
728 else /* We can actually calculate the percentage. */
730 for (size_t i = 0; i < ds->ds_num; i++)
731 values_copy[i] = 100.0 * values[i] / sum;
733 } /* if (UT_FLAG_PERCENTAGE) */
735 for (size_t i = 0; i < ds->ds_num; i++)
739 status = ut_check_one_data_source (ds, vl, th, values_copy, i);
745 } /* for (ds->ds_num) */
747 if (ret_ds_index != NULL)
748 *ret_ds_index = ds_index;
751 } /* }}} int ut_check_one_threshold */
754 * int ut_check_threshold
756 * Gets a list of matching thresholds and searches for the worst status by one
757 * of the thresholds. Then reports that status using the ut_report_state
759 * Returns zero on success and if no threshold has been configured. Returns
760 * less than zero on failure.
762 static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl,
763 __attribute__((unused)) user_data_t *ud)
769 int worst_state = -1;
770 threshold_t *worst_th = NULL;
771 int worst_ds_index = -1;
773 if (threshold_tree == NULL)
776 /* Is this lock really necessary? So far, thresholds are only inserted at
778 pthread_mutex_lock (&threshold_lock);
779 th = threshold_search (vl);
780 pthread_mutex_unlock (&threshold_lock);
784 DEBUG ("ut_check_threshold: Found matching threshold(s)");
786 values = uc_get_rate (ds, vl);
794 status = ut_check_one_threshold (ds, vl, th, values, &ds_index);
797 ERROR ("ut_check_threshold: ut_check_one_threshold failed.");
802 if (worst_state < status)
804 worst_state = status;
806 worst_ds_index = ds_index;
812 status = ut_report_state (ds, vl, worst_th, values,
813 worst_ds_index, worst_state);
816 ERROR ("ut_check_threshold: ut_report_state failed.");
824 } /* }}} int ut_check_threshold */
829 * This function is called whenever a value goes "missing".
831 static int ut_missing (const value_list_t *vl,
832 __attribute__((unused)) user_data_t *ud)
835 cdtime_t missing_time;
836 char identifier[6 * DATA_MAX_NAME_LEN];
840 if (threshold_tree == NULL)
843 th = threshold_search (vl);
844 /* dispatch notifications for "interesting" values only */
845 if ((th == NULL) || ((th->flags & UT_FLAG_INTERESTING) == 0))
849 missing_time = now - vl->time;
850 FORMAT_VL (identifier, sizeof (identifier), vl);
852 NOTIFICATION_INIT_VL (&n, vl);
853 ssnprintf (n.message, sizeof (n.message),
854 "%s has not been updated for %.3f seconds.",
855 identifier, CDTIME_T_TO_DOUBLE (missing_time));
858 plugin_dispatch_notification (&n);
861 } /* }}} int ut_missing */
863 static int ut_config (oconfig_item_t *ci)
866 int old_size = c_avl_size (threshold_tree);
868 if (threshold_tree == NULL)
870 threshold_tree = c_avl_create ((int (*) (const void *, const void *)) strcmp);
871 if (threshold_tree == NULL)
873 ERROR ("ut_config: c_avl_create failed.");
883 .flags = UT_FLAG_INTERESTING /* interesting by default */
886 for (int i = 0; i < ci->children_num; i++)
888 oconfig_item_t *option = ci->children + i;
890 if (strcasecmp ("Type", option->key) == 0)
891 status = ut_config_type (&th, option);
892 else if (strcasecmp ("Plugin", option->key) == 0)
893 status = ut_config_plugin (&th, option);
894 else if (strcasecmp ("Host", option->key) == 0)
895 status = ut_config_host (&th, option);
898 WARNING ("threshold values: Option `%s' not allowed here.", option->key);
906 /* register callbacks if this is the first time we see a valid config */
907 if ((old_size == 0) && (c_avl_size (threshold_tree) > 0))
909 plugin_register_missing ("threshold", ut_missing,
910 /* user data = */ NULL);
911 plugin_register_write ("threshold", ut_check_threshold,
912 /* user data = */ NULL);
916 } /* }}} int um_config */
918 void module_register (void)
920 plugin_register_complex_config ("threshold", ut_config);
923 /* vim: set sw=2 ts=8 sts=2 tw=78 et fdm=marker : */