From: Pierre-Yves Ritschard Date: Fri, 17 Jan 2014 10:05:49 +0000 (+0100) Subject: Add threshold checks for riemann output. X-Git-Tag: collectd-5.5.0~291^2~5 X-Git-Url: https://git.octo.it/?p=collectd.git;a=commitdiff_plain;h=a3dd2679c84398244a10a39e54613043f4feb940 Add threshold checks for riemann output. --- diff --git a/src/Makefile.am b/src/Makefile.am index a9d85823..30c779b9 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -133,9 +133,9 @@ collectd_tg_LDADD += libcollectdclient/libcollectdclient.la collectd_tg_DEPENDENCIES = libcollectdclient/libcollectdclient.la -pkglib_LTLIBRARIES = +pkglib_LTLIBRARIES = -BUILT_SOURCES = +BUILT_SOURCES = CLEANFILES = if BUILD_PLUGIN_AGGREGATION @@ -273,7 +273,7 @@ pkglib_LTLIBRARIES += cpu.la cpu_la_SOURCES = cpu.c cpu_la_CFLAGS = $(AM_CFLAGS) cpu_la_LDFLAGS = -module -avoid-version -cpu_la_LIBADD = +cpu_la_LIBADD = if BUILD_WITH_LIBKSTAT cpu_la_LIBADD += -lkstat endif @@ -371,7 +371,7 @@ pkglib_LTLIBRARIES += disk.la disk_la_SOURCES = disk.c disk_la_CFLAGS = $(AM_CFLAGS) disk_la_LDFLAGS = -module -avoid-version -disk_la_LIBADD = +disk_la_LIBADD = if BUILD_WITH_LIBKSTAT disk_la_LIBADD += -lkstat endif @@ -382,7 +382,7 @@ if BUILD_WITH_LIBIOKIT disk_la_LDFLAGS += -framework IOKit endif if BUILD_WITH_LIBSTATGRAB -disk_la_CFLAGS += $(BUILD_WITH_LIBSTATGRAB_CFLAGS) +disk_la_CFLAGS += $(BUILD_WITH_LIBSTATGRAB_CFLAGS) disk_la_LIBADD += $(BUILD_WITH_LIBSTATGRAB_LDFLAGS) endif if BUILD_WITH_PERFSTAT @@ -862,7 +862,7 @@ if BUILD_PLUGIN_OLSRD pkglib_LTLIBRARIES += olsrd.la olsrd_la_SOURCES = olsrd.c olsrd_la_LDFLAGS = -module -avoid-version -olsrd_la_LIBADD = +olsrd_la_LIBADD = if BUILD_WITH_LIBSOCKET olsrd_la_LIBADD += -lsocket endif @@ -1409,7 +1409,7 @@ endif if BUILD_PLUGIN_WRITE_RIEMANN pkglib_LTLIBRARIES += write_riemann.la -write_riemann_la_SOURCES = write_riemann.c +write_riemann_la_SOURCES = write_riemann.c write_riemann_threshold.c nodist_write_riemann_la_SOURCES = riemann.pb-c.c riemann.pb-c.h write_riemann_la_LDFLAGS = -module -avoid-version write_riemann_la_LIBADD = -lprotobuf-c diff --git a/src/write_riemann.c b/src/write_riemann.c index c890818d..c99aefdc 100644 --- a/src/write_riemann.c +++ b/src/write_riemann.c @@ -39,6 +39,9 @@ #define RIEMANN_PORT "5555" #define RIEMANN_TTL_FACTOR 2.0 +int write_riemann_threshold_check(const data_set_t *, const value_list_t *, int *); +int write_riemann_threshold_config(oconfig_item_t *); + struct riemann_host { char *name; #define F_CONNECT 0x01 @@ -450,7 +453,8 @@ static Msg *riemann_notification_to_protobuf (struct riemann_host *host, /* {{{ static Event *riemann_value_to_protobuf (struct riemann_host const *host, /* {{{ */ data_set_t const *ds, value_list_t const *vl, size_t index, - gauge_t const *rates) + gauge_t const *rates, + int status) { Event *event; char name_buffer[5 * DATA_MAX_NAME_LEN]; @@ -471,6 +475,21 @@ static Event *riemann_value_to_protobuf (struct riemann_host const *host, /* {{{ event->time = CDTIME_T_TO_TIME_T (vl->time); event->has_time = 1; + switch (status) { + case STATE_OKAY: + event->state = strdup("ok"); + break; + case STATE_ERROR: + event->state = strdup("critical"); + break; + case STATE_WARNING: + event->state = strdup("warning"); + break; + case STATE_MISSING: + event->state = strdup("unknown"); + break; + } + ttl = CDTIME_T_TO_DOUBLE (vl->interval) * host->ttl_factor; event->ttl = (float) ttl; event->has_ttl = 1; @@ -554,8 +573,9 @@ static Event *riemann_value_to_protobuf (struct riemann_host const *host, /* {{{ } /* }}} Event *riemann_value_to_protobuf */ static Msg *riemann_value_list_to_protobuf (struct riemann_host const *host, /* {{{ */ - data_set_t const *ds, - value_list_t const *vl) + data_set_t const *ds, + value_list_t const *vl, + int *statuses) { Msg *msg; size_t i; @@ -595,7 +615,7 @@ static Msg *riemann_value_list_to_protobuf (struct riemann_host const *host, /* for (i = 0; i < msg->n_events; i++) { msg->events[i] = riemann_value_to_protobuf (host, ds, vl, - (int) i, rates); + (int) i, rates, statuses[i]); if (msg->events[i] == NULL) { riemann_msg_protobuf_free (msg); @@ -632,10 +652,12 @@ static int riemann_write(const data_set_t *ds, /* {{{ */ user_data_t *ud) { int status; + int statuses[vl->values_len]; struct riemann_host *host = ud->data; Msg *msg; - msg = riemann_value_list_to_protobuf (host, ds, vl); + write_riemann_threshold_check(ds, vl, statuses); + msg = riemann_value_list_to_protobuf (host, ds, vl, statuses); if (msg == NULL) return (-1); @@ -712,6 +734,10 @@ static int riemann_config_node(oconfig_item_t *ci) /* {{{ */ status = cf_util_get_string (child, &host->node); if (status != 0) break; + } else if (strcasecmp ("Threshold", child->key) == 0) { + status = write_riemann_threshold_config(child); + if (status != 0) + break; } else if (strcasecmp ("Port", child->key) == 0) { status = cf_util_get_service (child, &host->service); if (status != 0) { diff --git a/src/write_riemann_threshold.c b/src/write_riemann_threshold.c index 86367747..0ef728a2 100644 --- a/src/write_riemann_threshold.c +++ b/src/write_riemann_threshold.c @@ -3,6 +3,7 @@ * Copyright (C) 2007-2010 Florian Forster * Copyright (C) 2008-2009 Sebastian Harl * Copyright (C) 2009 Andrés J. Díaz + * Copyright (C) 2014 Pierre-Yves Ritschard * * This program is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License as published by the @@ -18,6 +19,7 @@ * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA * * Author: + * Pierre-Yves Ritschard * Florian octo Forster * Sebastian Harl * Andrés J. Díaz @@ -171,7 +173,7 @@ static int ut_threshold_add (const threshold_t *th) return (status); } /* }}} int ut_threshold_add */ -/* +/* * threshold_t *threshold_search * * Searches for a threshold configuration using all the possible variations of @@ -521,195 +523,6 @@ static int ut_config_host (const threshold_t *th_orig, oconfig_item_t *ci) /* }}} */ /* - * int ut_report_state - * - * Checks if the `state' differs from the old state and creates a notification - * if appropriate. - * Does not fail. - */ -static int ut_report_state (const data_set_t *ds, - const value_list_t *vl, - const threshold_t *th, - const gauge_t *values, - int ds_index, - int state) -{ /* {{{ */ - int state_old; - notification_t n; - - char *buf; - size_t bufsize; - - int status; - - /* Check if hits matched */ - if ( (th->hits != 0) ) - { - int hits = uc_get_hits(ds,vl); - /* STATE_OKAY resets hits unless PERSIST_OK flag is set. Hits resets if - * threshold is hit. */ - if ( ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) ) || (hits > th->hits) ) - { - DEBUG("ut_report_state: reset uc_get_hits = 0"); - uc_set_hits(ds,vl,0); /* reset hit counter and notify */ - } else { - DEBUG("ut_report_state: th->hits = %d, uc_get_hits = %d",th->hits,uc_get_hits(ds,vl)); - (void) uc_inc_hits(ds,vl,1); /* increase hit counter */ - return (0); - } - } /* end check hits */ - - state_old = uc_get_state (ds, vl); - - /* If the state didn't change, report if `persistent' is specified. If the - * state is `okay', then only report if `persist_ok` flag is set. */ - if (state == state_old) - { - if ((th->flags & UT_FLAG_PERSIST) == 0) - return (0); - else if ( (state == STATE_OKAY) && ((th->flags & UT_FLAG_PERSIST_OK) == 0) ) - return (0); - } - - if (state != state_old) - uc_set_state (ds, vl, state); - - NOTIFICATION_INIT_VL (&n, vl); - - buf = n.message; - bufsize = sizeof (n.message); - - if (state == STATE_OKAY) - n.severity = NOTIF_OKAY; - else if (state == STATE_WARNING) - n.severity = NOTIF_WARNING; - else - n.severity = NOTIF_FAILURE; - - n.time = vl->time; - - status = ssnprintf (buf, bufsize, "Host %s, plugin %s", - vl->host, vl->plugin); - buf += status; - bufsize -= status; - - if (vl->plugin_instance[0] != '\0') - { - status = ssnprintf (buf, bufsize, " (instance %s)", - vl->plugin_instance); - buf += status; - bufsize -= status; - } - - status = ssnprintf (buf, bufsize, " type %s", vl->type); - buf += status; - bufsize -= status; - - if (vl->type_instance[0] != '\0') - { - status = ssnprintf (buf, bufsize, " (instance %s)", - vl->type_instance); - buf += status; - bufsize -= status; - } - - plugin_notification_meta_add_string (&n, "DataSource", - ds->ds[ds_index].name); - plugin_notification_meta_add_double (&n, "CurrentValue", values[ds_index]); - plugin_notification_meta_add_double (&n, "WarningMin", th->warning_min); - plugin_notification_meta_add_double (&n, "WarningMax", th->warning_max); - plugin_notification_meta_add_double (&n, "FailureMin", th->failure_min); - plugin_notification_meta_add_double (&n, "FailureMax", th->failure_max); - - /* Send an okay notification */ - if (state == STATE_OKAY) - { - if (state_old == STATE_MISSING) - status = ssnprintf (buf, bufsize, - ": Value is no longer missing."); - else - status = ssnprintf (buf, bufsize, - ": All data sources are within range again."); - buf += status; - bufsize -= status; - } - else - { - double min; - double max; - - min = (state == STATE_ERROR) ? th->failure_min : th->warning_min; - max = (state == STATE_ERROR) ? th->failure_max : th->warning_max; - - if (th->flags & UT_FLAG_INVERT) - { - if (!isnan (min) && !isnan (max)) - { - status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently " - "%f. That is within the %s region of %f%s and %f%s.", - ds->ds[ds_index].name, values[ds_index], - (state == STATE_ERROR) ? "failure" : "warning", - min, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : "", - max, ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : ""); - } - else - { - status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently " - "%f. That is %s the %s threshold of %f%s.", - ds->ds[ds_index].name, values[ds_index], - isnan (min) ? "below" : "above", - (state == STATE_ERROR) ? "failure" : "warning", - isnan (min) ? max : min, - ((th->flags & UT_FLAG_PERCENTAGE) != 0) ? "%" : ""); - } - } - else if (th->flags & UT_FLAG_PERCENTAGE) - { - gauge_t value; - gauge_t sum; - int i; - - sum = 0.0; - for (i = 0; i < vl->values_len; i++) - { - if (isnan (values[i])) - continue; - - sum += values[i]; - } - - if (sum == 0.0) - value = NAN; - else - value = 100.0 * values[ds_index] / sum; - - status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently " - "%g (%.2f%%). That is %s the %s threshold of %.2f%%.", - ds->ds[ds_index].name, values[ds_index], value, - (value < min) ? "below" : "above", - (state == STATE_ERROR) ? "failure" : "warning", - (value < min) ? min : max); - } - else /* is not inverted */ - { - status = ssnprintf (buf, bufsize, ": Data source \"%s\" is currently " - "%f. That is %s the %s threshold of %f.", - ds->ds[ds_index].name, values[ds_index], - (values[ds_index] < min) ? "below" : "above", - (state == STATE_ERROR) ? "failure" : "warning", - (values[ds_index] < min) ? min : max); - } - buf += status; - bufsize -= status; - } - - plugin_dispatch_notification (&n); - - plugin_notification_meta_free (n.meta); - return (0); -} /* }}} int ut_report_state */ - -/* * int ut_check_one_data_source * * Checks one data source against the given threshold configuration. If the @@ -796,11 +609,11 @@ static int ut_check_one_threshold (const data_set_t *ds, const value_list_t *vl, const threshold_t *th, const gauge_t *values, - int *ret_ds_index) + int *statuses) { /* {{{ */ int ret = -1; - int ds_index = -1; int i; + int status; gauge_t values_copy[ds->ds_num]; memcpy (values_copy, values, sizeof (values_copy)); @@ -841,19 +654,14 @@ static int ut_check_one_threshold (const data_set_t *ds, for (i = 0; i < ds->ds_num; i++) { - int status; - status = ut_check_one_data_source (ds, vl, th, values_copy, i); - if (ret < status) - { - ret = status; - ds_index = i; + if (status != -1) { + ret = 0; + if (statuses[i] < status) + statuses[i] = status; } } /* for (ds->ds_num) */ - if (ret_ds_index != NULL) - *ret_ds_index = ds_index; - return (ret); } /* }}} int ut_check_one_threshold */ @@ -862,23 +670,22 @@ static int ut_check_one_threshold (const data_set_t *ds, * * Gets a list of matching thresholds and searches for the worst status by one * of the thresholds. Then reports that status using the ut_report_state - * function above. + * function above. * Returns zero on success and if no threshold has been configured. Returns * less than zero on failure. */ -static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl, - __attribute__((unused)) user_data_t *ud) +int write_riemann_threshold_check (const data_set_t *ds, const value_list_t *vl, + int *statuses) { /* {{{ */ threshold_t *th; gauge_t *values; int status; - int worst_state = -1; - threshold_t *worst_th = NULL; - int worst_ds_index = -1; + memset(statuses, 0, vl->values_len * sizeof(*statuses)); + if (threshold_tree == NULL) - return (0); + return 0; /* Is this lock really necessary? So far, thresholds are only inserted at * startup. -octo */ @@ -886,19 +693,17 @@ static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl, th = threshold_search (vl); pthread_mutex_unlock (&threshold_lock); if (th == NULL) - return (0); + return (0); DEBUG ("ut_check_threshold: Found matching threshold(s)"); values = uc_get_rate (ds, vl); if (values == NULL) - return (0); + return (0); while (th != NULL) { - int ds_index = -1; - - status = ut_check_one_threshold (ds, vl, th, values, &ds_index); + status = ut_check_one_threshold (ds, vl, th, values, statuses); if (status < 0) { ERROR ("ut_check_threshold: ut_check_one_threshold failed."); @@ -906,64 +711,14 @@ static int ut_check_threshold (const data_set_t *ds, const value_list_t *vl, return (-1); } - if (worst_state < status) - { - worst_state = status; - worst_th = th; - worst_ds_index = ds_index; - } - th = th->next; } /* while (th) */ - status = ut_report_state (ds, vl, worst_th, values, - worst_ds_index, worst_state); - if (status != 0) - { - ERROR ("ut_check_threshold: ut_report_state failed."); - sfree (values); - return (-1); - } - sfree (values); return (0); } /* }}} int ut_check_threshold */ -/* - * int ut_missing - * - * This function is called whenever a value goes "missing". - */ -static int ut_missing (const value_list_t *vl, - __attribute__((unused)) user_data_t *ud) -{ /* {{{ */ - threshold_t *th; - cdtime_t missing_time; - char identifier[6 * DATA_MAX_NAME_LEN]; - notification_t n; - - if (threshold_tree == NULL) - return (0); - - th = threshold_search (vl); - /* dispatch notifications for "interesting" values only */ - if ((th == NULL) || ((th->flags & UT_FLAG_INTERESTING) == 0)) - return (0); - - missing_time = cdtime () - vl->time; - FORMAT_VL (identifier, sizeof (identifier), vl); - - NOTIFICATION_INIT_VL (&n, vl); - ssnprintf (n.message, sizeof (n.message), - "%s has not been updated for %.3f seconds.", - identifier, CDTIME_T_TO_DOUBLE (missing_time)); - - plugin_dispatch_notification (&n); - - return (0); -} /* }}} int ut_missing */ - int write_riemann_threshold_config (oconfig_item_t *ci) { /* {{{ */ int i; @@ -990,7 +745,7 @@ int write_riemann_threshold_config (oconfig_item_t *ci) th.hits = 0; th.hysteresis = 0; th.flags = UT_FLAG_INTERESTING; /* interesting by default */ - + for (i = 0; i < ci->children_num; i++) { oconfig_item_t *option = ci->children + i; @@ -1012,13 +767,6 @@ int write_riemann_threshold_config (oconfig_item_t *ci) break; } - if (c_avl_size (threshold_tree) > 0) { - plugin_register_missing ("threshold", ut_missing, - /* user data = */ NULL); - plugin_register_write ("threshold", ut_check_threshold, - /* user data = */ NULL); - } - return (status); } /* }}} int um_config */