diff options
Diffstat (limited to 'database/rrdset.c')
| -rw-r--r-- | database/rrdset.c | 2361 |
1 files changed, 2361 insertions, 0 deletions
diff --git a/database/rrdset.c b/database/rrdset.c new file mode 100644 index 00000000..f4bb48aa --- /dev/null +++ b/database/rrdset.c @@ -0,0 +1,2361 @@ +// SPDX-License-Identifier: GPL-3.0-or-later + +#define NETDATA_RRD_INTERNALS +#include "rrd.h" +#include <sched.h> +#include "storage_engine.h" + + +void rrdset_metadata_updated(RRDSET *st) { + __atomic_add_fetch(&st->version, 1, __ATOMIC_RELAXED); + rrdcontext_updated_rrdset(st); +} + +// ---------------------------------------------------------------------------- +// RRDSET rrdpush send chart_slots + +static void rrdset_rrdpush_send_chart_slot_assign(RRDSET *st) { + RRDHOST *host = st->rrdhost; + spinlock_lock(&host->rrdpush.send.pluginsd_chart_slots.available.spinlock); + + if(host->rrdpush.send.pluginsd_chart_slots.available.used > 0) + st->rrdpush.sender.chart_slot = + host->rrdpush.send.pluginsd_chart_slots.available.array[--host->rrdpush.send.pluginsd_chart_slots.available.used]; + else + st->rrdpush.sender.chart_slot = ++host->rrdpush.send.pluginsd_chart_slots.last_used; + + spinlock_unlock(&host->rrdpush.send.pluginsd_chart_slots.available.spinlock); +} + +static void rrdset_rrdpush_send_chart_slot_release(RRDSET *st) { + if(!st->rrdpush.sender.chart_slot || st->rrdhost->rrdpush.send.pluginsd_chart_slots.available.ignore) + return; + + RRDHOST *host = st->rrdhost; + spinlock_lock(&host->rrdpush.send.pluginsd_chart_slots.available.spinlock); + + if(host->rrdpush.send.pluginsd_chart_slots.available.used >= host->rrdpush.send.pluginsd_chart_slots.available.size) { + uint32_t old_size = host->rrdpush.send.pluginsd_chart_slots.available.size; + uint32_t new_size = (old_size > 0) ? (old_size * 2) : 1024; + + host->rrdpush.send.pluginsd_chart_slots.available.array = + reallocz(host->rrdpush.send.pluginsd_chart_slots.available.array, new_size * sizeof(uint32_t)); + + host->rrdpush.send.pluginsd_chart_slots.available.size = new_size; + } + + host->rrdpush.send.pluginsd_chart_slots.available.array[host->rrdpush.send.pluginsd_chart_slots.available.used++] = + st->rrdpush.sender.chart_slot; + + st->rrdpush.sender.chart_slot = 0; + spinlock_unlock(&host->rrdpush.send.pluginsd_chart_slots.available.spinlock); +} + +void rrdhost_pluginsd_send_chart_slots_free(RRDHOST *host) { + spinlock_lock(&host->rrdpush.send.pluginsd_chart_slots.available.spinlock); + host->rrdpush.send.pluginsd_chart_slots.available.ignore = true; + freez(host->rrdpush.send.pluginsd_chart_slots.available.array); + host->rrdpush.send.pluginsd_chart_slots.available.array = NULL; + host->rrdpush.send.pluginsd_chart_slots.available.used = 0; + host->rrdpush.send.pluginsd_chart_slots.available.size = 0; + spinlock_unlock(&host->rrdpush.send.pluginsd_chart_slots.available.spinlock); + + // zero all the slots on all charts, so that they will not attempt to access the array + RRDSET *st; + rrdset_foreach_read(st, host) { + st->rrdpush.sender.chart_slot = 0; + } + rrdset_foreach_done(st); +} + +void rrdset_pluginsd_receive_unslot(RRDSET *st) { + for(size_t i = 0; i < st->pluginsd.size ;i++) { + rrddim_acquired_release(st->pluginsd.prd_array[i].rda); // can be NULL + st->pluginsd.prd_array[i].rda = NULL; + st->pluginsd.prd_array[i].rd = NULL; + st->pluginsd.prd_array[i].id = NULL; + } + + RRDHOST *host = st->rrdhost; + + if(st->pluginsd.last_slot >= 0 && + (uint32_t)st->pluginsd.last_slot < host->rrdpush.receive.pluginsd_chart_slots.size && + host->rrdpush.receive.pluginsd_chart_slots.array[st->pluginsd.last_slot] == st) { + host->rrdpush.receive.pluginsd_chart_slots.array[st->pluginsd.last_slot] = NULL; + } + + st->pluginsd.last_slot = -1; + st->pluginsd.dims_with_slots = false; +} + +void rrdset_pluginsd_receive_unslot_and_cleanup(RRDSET *st) { + if(!st) + return; + + spinlock_lock(&st->pluginsd.spinlock); + + rrdset_pluginsd_receive_unslot(st); + + freez(st->pluginsd.prd_array); + st->pluginsd.prd_array = NULL; + st->pluginsd.size = 0; + st->pluginsd.pos = 0; + st->pluginsd.set = false; + st->pluginsd.last_slot = -1; + st->pluginsd.dims_with_slots = false; + st->pluginsd.collector_tid = 0; + + spinlock_unlock(&st->pluginsd.spinlock); +} + +static void rrdset_pluginsd_receive_slots_initialize(RRDSET *st) { + spinlock_init(&st->pluginsd.spinlock); + st->pluginsd.last_slot = -1; +} + +void rrdhost_pluginsd_receive_chart_slots_free(RRDHOST *host) { + spinlock_lock(&host->rrdpush.receive.pluginsd_chart_slots.spinlock); + + if(host->rrdpush.receive.pluginsd_chart_slots.array) { + for (size_t s = 0; s < host->rrdpush.receive.pluginsd_chart_slots.size; s++) + rrdset_pluginsd_receive_unslot_and_cleanup(host->rrdpush.receive.pluginsd_chart_slots.array[s]); + + freez(host->rrdpush.receive.pluginsd_chart_slots.array); + host->rrdpush.receive.pluginsd_chart_slots.array = NULL; + host->rrdpush.receive.pluginsd_chart_slots.size = 0; + } + + spinlock_unlock(&host->rrdpush.receive.pluginsd_chart_slots.spinlock); +} + +// ---------------------------------------------------------------------------- +// RRDSET name index + +static void rrdset_name_insert_callback(const DICTIONARY_ITEM *item __maybe_unused, void *rrdset, void *rrdhost __maybe_unused) { + RRDSET *st = rrdset; + rrdset_flag_set(st, RRDSET_FLAG_INDEXED_NAME); +} +static void rrdset_name_delete_callback(const DICTIONARY_ITEM *item __maybe_unused, void *rrdset, void *rrdhost __maybe_unused) { + RRDSET *st = rrdset; + rrdset_flag_clear(st, RRDSET_FLAG_INDEXED_NAME); +} + +static inline void rrdset_index_add_name(RRDHOST *host, RRDSET *st) { + if(!st->name) return; + dictionary_set(host->rrdset_root_index_name, rrdset_name(st), st, sizeof(RRDSET)); +} + +static inline void rrdset_index_del_name(RRDHOST *host, RRDSET *st) { + if(rrdset_flag_check(st, RRDSET_FLAG_INDEXED_NAME)) + dictionary_del(host->rrdset_root_index_name, rrdset_name(st)); +} + +static inline RRDSET *rrdset_index_find_name(RRDHOST *host, const char *name) { + if (unlikely(!host->rrdset_root_index_name)) + return NULL; + return dictionary_get(host->rrdset_root_index_name, name); +} + +// ---------------------------------------------------------------------------- +// RRDSET index + +static inline void rrdset_update_permanent_labels(RRDSET *st) { + if(!st->rrdlabels) return; + + rrdlabels_add(st->rrdlabels, "_collect_plugin", rrdset_plugin_name(st), RRDLABEL_SRC_AUTO | RRDLABEL_FLAG_DONT_DELETE); + rrdlabels_add(st->rrdlabels, "_collect_module", rrdset_module_name(st), RRDLABEL_SRC_AUTO | RRDLABEL_FLAG_DONT_DELETE); +} + +static STRING *rrdset_fix_name(RRDHOST *host, const char *chart_full_id, const char *type, const char *current_name, const char *name) { + if(!name || !*name) return NULL; + + char full_name[RRD_ID_LENGTH_MAX + 1]; + char sanitized_name[CONFIG_MAX_VALUE + 1]; + char new_name[CONFIG_MAX_VALUE + 1]; + + snprintfz(full_name, RRD_ID_LENGTH_MAX, "%s.%s", type, name); + rrdset_strncpyz_name(sanitized_name, full_name, CONFIG_MAX_VALUE); + strncpyz(new_name, sanitized_name, CONFIG_MAX_VALUE); + + if(rrdset_index_find_name(host, new_name)) { + netdata_log_debug(D_RRD_CALLS, "RRDSET: chart name '%s' on host '%s' already exists.", new_name, rrdhost_hostname(host)); + if(!strcmp(chart_full_id, full_name) && (!current_name || !*current_name)) { + unsigned i = 1; + + do { + snprintfz(new_name, CONFIG_MAX_VALUE, "%s_%u", sanitized_name, i); + i++; + } while (rrdset_index_find_name(host, new_name)); + +// netdata_log_info("RRDSET: using name '%s' for chart '%s' on host '%s'.", new_name, full_name, rrdhost_hostname(host)); + } + else + return NULL; + } + + return string_strdupz(new_name); +} + +struct rrdset_constructor { + RRDHOST *host; + const char *type; + const char *id; + const char *name; + const char *family; + const char *context; + const char *title; + const char *units; + const char *plugin; + const char *module; + long priority; + int update_every; + RRDSET_TYPE chart_type; + RRD_MEMORY_MODE memory_mode; + long history_entries; + + enum { + RRDSET_REACT_NONE = 0, + RRDSET_REACT_NEW = (1 << 0), + RRDSET_REACT_UPDATED = (1 << 1), + RRDSET_REACT_PLUGIN_UPDATED = (1 << 2), + RRDSET_REACT_MODULE_UPDATED = (1 << 3), + RRDSET_REACT_CHART_ACTIVATED = (1 << 4), + } react_action; +}; + +// the constructor - the dictionary is write locked while this runs +static void rrdset_insert_callback(const DICTIONARY_ITEM *item __maybe_unused, void *rrdset, void *constructor_data) { + struct rrdset_constructor *ctr = constructor_data; + RRDHOST *host = ctr->host; + RRDSET *st = rrdset; + + const char *chart_full_id = dictionary_acquired_item_name(item); + + st->id = string_strdupz(chart_full_id); + + st->name = rrdset_fix_name(host, chart_full_id, ctr->type, NULL, ctr->name); + if(!st->name) + st->name = rrdset_fix_name(host, chart_full_id, ctr->type, NULL, ctr->id); + rrdset_index_add_name(host, st); + + st->parts.id = string_strdupz(ctr->id); + st->parts.type = string_strdupz(ctr->type); + st->parts.name = string_strdupz(ctr->name); + + st->family = (ctr->family && *ctr->family) ? rrd_string_strdupz(ctr->family) : rrd_string_strdupz(ctr->type); + st->context = (ctr->context && *ctr->context) ? rrd_string_strdupz(ctr->context) : rrd_string_strdupz(chart_full_id); + + st->units = rrd_string_strdupz(ctr->units); + st->title = rrd_string_strdupz(ctr->title); + st->plugin_name = rrd_string_strdupz(ctr->plugin); + st->module_name = rrd_string_strdupz(ctr->module); + st->priority = ctr->priority; + + st->db.entries = (ctr->memory_mode != RRD_MEMORY_MODE_DBENGINE) ? align_entries_to_pagesize(ctr->memory_mode, ctr->history_entries) : 5; + st->update_every = ctr->update_every; + st->rrd_memory_mode = ctr->memory_mode; + + st->chart_type = ctr->chart_type; + st->rrdhost = host; + + rrdset_rrdpush_send_chart_slot_assign(st); + + spinlock_init(&st->data_collection_lock); + + st->flags = RRDSET_FLAG_SYNC_CLOCK + | RRDSET_FLAG_INDEXED_ID + | RRDSET_FLAG_RECEIVER_REPLICATION_FINISHED + | RRDSET_FLAG_SENDER_REPLICATION_FINISHED + ; + + rw_spinlock_init(&st->alerts.spinlock); + + if(st->rrd_memory_mode == RRD_MEMORY_MODE_SAVE || st->rrd_memory_mode == RRD_MEMORY_MODE_MAP) { + if(!rrdset_memory_load_or_create_map_save(st, st->rrd_memory_mode)) { + netdata_log_info("Failed to use db mode %s for chart '%s', falling back to ram mode.", (st->rrd_memory_mode == RRD_MEMORY_MODE_MAP)?"map":"save", rrdset_name(st)); + st->rrd_memory_mode = RRD_MEMORY_MODE_RAM; + } + } + + // initialize the db tiers + { + for(size_t tier = 0; tier < storage_tiers ; tier++) { + STORAGE_ENGINE *eng = st->rrdhost->db[tier].eng; + if(!eng) continue; + + st->storage_metrics_groups[tier] = storage_engine_metrics_group_get(eng->backend, host->db[tier].instance, &st->chart_uuid); + } + } + + rrddim_index_init(st); + + // chart variables - we need this for data collection to work (collector given chart variables) - not only health + rrdsetvar_index_init(st); + + if (host->health.health_enabled) { + st->rrdfamily = rrdfamily_add_and_acquire(host, rrdset_family(st)); + st->rrdvars = rrdvariables_create(); + rrddimvar_index_init(st); + } + + st->rrdlabels = rrdlabels_create(); + rrdset_update_permanent_labels(st); + + st->green = NAN; + st->red = NAN; + + rrdset_pluginsd_receive_slots_initialize(st); + + ctr->react_action = RRDSET_REACT_NEW; + + ml_chart_new(st); +} + +void rrdset_finalize_collection(RRDSET *st, bool dimensions_too) { + RRDHOST *host = st->rrdhost; + + rrdset_flag_set(st, RRDSET_FLAG_COLLECTION_FINISHED); + + if(dimensions_too) { + RRDDIM *rd; + rrddim_foreach_read(rd, st) + rrddim_finalize_collection_and_check_retention(rd); + rrddim_foreach_done(rd); + } + + for(size_t tier = 0; tier < storage_tiers ; tier++) { + STORAGE_ENGINE *eng = st->rrdhost->db[tier].eng; + if(!eng) continue; + + if(st->storage_metrics_groups[tier]) { + storage_engine_metrics_group_release(eng->backend, host->db[tier].instance, st->storage_metrics_groups[tier]); + st->storage_metrics_groups[tier] = NULL; + } + } + + rrdset_pluginsd_receive_unslot_and_cleanup(st); +} + +// the destructor - the dictionary is write locked while this runs +static void rrdset_delete_callback(const DICTIONARY_ITEM *item __maybe_unused, void *rrdset, void *rrdhost) { + RRDHOST *host = rrdhost; + RRDSET *st = rrdset; + + rrdset_flag_clear(st, RRDSET_FLAG_INDEXED_ID); + + rrdset_finalize_collection(st, false); + + rrdset_rrdpush_send_chart_slot_release(st); + + // remove it from the name index + rrdset_index_del_name(host, st); + + // release the collector info + dictionary_destroy(st->functions_view); + + rrdcalc_unlink_all_rrdset_alerts(st); + + // ------------------------------------------------------------------------ + // the order of destruction is important here + + // 1. delete RRDDIMVAR index - this will speed up the destruction of RRDDIMs + // because each dimension loops to find its own variables in this index. + // There are no references to the items on this index from the dimensions. + // To find their own, they have to walk-through the dictionary. + rrddimvar_index_destroy(st); // destroy the rrddimvar index + + // 2. delete RRDSETVAR index + rrdsetvar_index_destroy(st); // destroy the rrdsetvar index + + // 3. delete RRDVAR index after the above, to avoid triggering its garbage collector (they have references on this) + rrdvariables_destroy(st->rrdvars); // free all variables and destroy the rrdvar dictionary + + // 4. delete RRDFAMILY - this has to be last, because RRDDIMVAR and RRDSETVAR need the reference counter + rrdfamily_release(host, st->rrdfamily); // release the acquired rrdfamily -- has to be after all variables + + // 5. delete RRDDIMs, now their variables are not existing, so this is fast + rrddim_index_destroy(st); // free all the dimensions and destroy the dimensions index + + // 6. this has to be after the dimensions are freed, but before labels are freed (contexts need the labels) + rrdcontext_removed_rrdset(st); // let contexts know + + // 7. destroy the chart labels + rrdlabels_destroy(st->rrdlabels); // destroy the labels, after letting the contexts know + + // 8. destroy the ml handle + ml_chart_delete(st); + + rrdset_memory_file_free(st); // remove files of db mode save and map + + // ------------------------------------------------------------------------ + // free it + + string_freez(st->id); + string_freez(st->name); + string_freez(st->parts.id); + string_freez(st->parts.type); + string_freez(st->parts.name); + string_freez(st->family); + string_freez(st->title); + string_freez(st->units); + string_freez(st->context); + string_freez(st->plugin_name); + string_freez(st->module_name); + + freez(st->exporting_flags); + freez(st->db.cache_dir); +} + +// the item to be inserted, is already in the dictionary +// this callback deals with the situation, migrating the existing object to the new values +// the dictionary is write locked while this runs +static bool rrdset_conflict_callback(const DICTIONARY_ITEM *item __maybe_unused, void *rrdset, void *new_rrdset, void *constructor_data) { + (void)new_rrdset; // it is NULL + + struct rrdset_constructor *ctr = constructor_data; + RRDSET *st = rrdset; + + rrdset_isnot_obsolete___safe_from_collector_thread(st); + + ctr->react_action = RRDSET_REACT_NONE; + + if (rrdset_reset_name(st, (ctr->name && *ctr->name) ? ctr->name : ctr->id) == 2) + ctr->react_action |= RRDSET_REACT_UPDATED; + + if (unlikely(st->priority != ctr->priority)) { + st->priority = ctr->priority; + ctr->react_action |= RRDSET_REACT_UPDATED; + } + + if (unlikely(st->update_every != ctr->update_every)) { + rrdset_set_update_every_s(st, ctr->update_every); + ctr->react_action |= RRDSET_REACT_UPDATED; + } + + if(ctr->plugin && *ctr->plugin) { + STRING *old_plugin = st->plugin_name; + st->plugin_name = rrd_string_strdupz(ctr->plugin); + if (old_plugin != st->plugin_name) + ctr->react_action |= RRDSET_REACT_PLUGIN_UPDATED; + string_freez(old_plugin); + } + + if(ctr->module && *ctr->module) { + STRING *old_module = st->module_name; + st->module_name = rrd_string_strdupz(ctr->module); + if (old_module != st->module_name) + ctr->react_action |= RRDSET_REACT_MODULE_UPDATED; + string_freez(old_module); + } + + if(ctr->title && *ctr->title) { + STRING *old_title = st->title; + st->title = rrd_string_strdupz(ctr->title); + if(old_title != st->title) + ctr->react_action |= RRDSET_REACT_UPDATED; + string_freez(old_title); + } + + if(ctr->units && *ctr->units) { + STRING *old_units = st->units; + st->units = rrd_string_strdupz(ctr->units); + if(old_units != st->units) + ctr->react_action |= RRDSET_REACT_UPDATED; + string_freez(old_units); + } + + if(ctr->family && *ctr->family) { + STRING *old_family = st->family; + st->family = rrd_string_strdupz(ctr->family); + if(old_family != st->family) + ctr->react_action |= RRDSET_REACT_UPDATED; + string_freez(old_family); + + // TODO - we should rename RRDFAMILY variables + } + + if(ctr->context && *ctr->context) { + STRING *old_context = st->context; + st->context = rrd_string_strdupz(ctr->context); + if(old_context != st->context) + ctr->react_action |= RRDSET_REACT_UPDATED; + string_freez(old_context); + } + + if(st->chart_type != ctr->chart_type) { + st->chart_type = ctr->chart_type; + ctr->react_action |= RRDSET_REACT_UPDATED; + } + + rrdset_update_permanent_labels(st); + + rrdset_flag_set(st, RRDSET_FLAG_SYNC_CLOCK); + + return ctr->react_action != RRDSET_REACT_NONE; +} + +// this is called after all insertions/conflicts, with the dictionary unlocked, with a reference to RRDSET +// so, any actions requiring locks on other objects, should be placed here +static void rrdset_react_callback(const DICTIONARY_ITEM *item __maybe_unused, void *rrdset, void *constructor_data) { + struct rrdset_constructor *ctr = constructor_data; + RRDSET *st = rrdset; + RRDHOST *host = st->rrdhost; + + st->last_accessed_time_s = now_realtime_sec(); + + if(host->health.health_enabled && (ctr->react_action & (RRDSET_REACT_NEW | RRDSET_REACT_CHART_ACTIVATED))) { + rrdset_flag_set(st, RRDSET_FLAG_PENDING_HEALTH_INITIALIZATION); + rrdhost_flag_set(st->rrdhost, RRDHOST_FLAG_PENDING_HEALTH_INITIALIZATION); + } + + if(ctr->react_action & (RRDSET_REACT_NEW | RRDSET_REACT_PLUGIN_UPDATED | RRDSET_REACT_MODULE_UPDATED)) { + if (ctr->react_action & RRDSET_REACT_NEW) { + if(unlikely(rrdcontext_find_chart_uuid(st, &st->chart_uuid))) + uuid_generate(st->chart_uuid); + } + rrdset_flag_set(st, RRDSET_FLAG_METADATA_UPDATE); + rrdhost_flag_set(st->rrdhost, RRDHOST_FLAG_METADATA_UPDATE); + } + + rrdset_metadata_updated(st); +} + +void rrdset_index_init(RRDHOST *host) { + if(!host->rrdset_root_index) { + host->rrdset_root_index = dictionary_create_advanced(DICT_OPTION_DONT_OVERWRITE_VALUE | DICT_OPTION_FIXED_SIZE, + &dictionary_stats_category_rrdset_rrddim, sizeof(RRDSET)); + + dictionary_register_insert_callback(host->rrdset_root_index, rrdset_insert_callback, NULL); + dictionary_register_conflict_callback(host->rrdset_root_index, rrdset_conflict_callback, NULL); + dictionary_register_react_callback(host->rrdset_root_index, rrdset_react_callback, NULL); + dictionary_register_delete_callback(host->rrdset_root_index, rrdset_delete_callback, host); + } + + if(!host->rrdset_root_index_name) { + host->rrdset_root_index_name = dictionary_create_advanced( + DICT_OPTION_NAME_LINK_DONT_CLONE | DICT_OPTION_VALUE_LINK_DONT_CLONE | DICT_OPTION_DONT_OVERWRITE_VALUE, + &dictionary_stats_category_rrdset_rrddim, 0); + + dictionary_register_insert_callback(host->rrdset_root_index_name, rrdset_name_insert_callback, host); + dictionary_register_delete_callback(host->rrdset_root_index_name, rrdset_name_delete_callback, host); + } +} + +void rrdset_index_destroy(RRDHOST *host) { + // destroy the name index first + dictionary_destroy(host->rrdset_root_index_name); + host->rrdset_root_index_name = NULL; + + // destroy the id index last + dictionary_destroy(host->rrdset_root_index); + host->rrdset_root_index = NULL; +} + +static inline RRDSET *rrdset_index_add(RRDHOST *host, const char *id, struct rrdset_constructor *st_ctr) { + return dictionary_set_advanced(host->rrdset_root_index, id, -1, NULL, sizeof(RRDSET), st_ctr); +} + +static inline void rrdset_index_del(RRDHOST *host, RRDSET *st) { + if(rrdset_flag_check(st, RRDSET_FLAG_INDEXED_ID)) + dictionary_del(host->rrdset_root_index, rrdset_id(st)); +} + +static RRDSET *rrdset_index_find(RRDHOST *host, const char *id) { + // TODO - the name index should have an acquired dictionary item, not just a pointer to RRDSET + if (unlikely(!host->rrdset_root_index)) + return NULL; + return dictionary_get(host->rrdset_root_index, id); +} + +// ---------------------------------------------------------------------------- +// RRDSET - find charts + +inline RRDSET *rrdset_find(RRDHOST *host, const char *id) { + netdata_log_debug(D_RRD_CALLS, "rrdset_find() for chart '%s' in host '%s'", id, rrdhost_hostname(host)); + RRDSET *st = rrdset_index_find(host, id); + + if(st) + st->last_accessed_time_s = now_realtime_sec(); + + return(st); +} + +inline RRDSET *rrdset_find_bytype(RRDHOST *host, const char *type, const char *id) { + netdata_log_debug(D_RRD_CALLS, "rrdset_find_bytype() for chart '%s.%s' in host '%s'", type, id, rrdhost_hostname(host)); + + char buf[RRD_ID_LENGTH_MAX + 1]; + strncpyz(buf, type, RRD_ID_LENGTH_MAX - 1); + strcat(buf, "."); + int len = (int) strlen(buf); + strncpyz(&buf[len], id, (size_t) (RRD_ID_LENGTH_MAX - len)); + + return(rrdset_find(host, buf)); +} + +inline RRDSET *rrdset_find_byname(RRDHOST *host, const char *name) { + netdata_log_debug(D_RRD_CALLS, "rrdset_find_byname() for chart '%s' in host '%s'", name, rrdhost_hostname(host)); + RRDSET *st = rrdset_index_find_name(host, name); + return(st); +} + +RRDSET_ACQUIRED *rrdset_find_and_acquire(RRDHOST *host, const char *id) { + netdata_log_debug(D_RRD_CALLS, "rrdset_find_and_acquire() for host %s, chart %s", rrdhost_hostname(host), id); + + return (RRDSET_ACQUIRED *)dictionary_get_and_acquire_item(host->rrdset_root_index, id); +} + +RRDSET *rrdset_acquired_to_rrdset(RRDSET_ACQUIRED *rsa) { + if(unlikely(!rsa)) + return NULL; + + return (RRDSET *) dictionary_acquired_item_value((const DICTIONARY_ITEM *)rsa); +} + +void rrdset_acquired_release(RRDSET_ACQUIRED *rsa) { + if(unlikely(!rsa)) + return; + + RRDSET *rs = rrdset_acquired_to_rrdset(rsa); + dictionary_acquired_item_release(rs->rrdhost->rrdset_root_index, (const DICTIONARY_ITEM *)rsa); +} + +// ---------------------------------------------------------------------------- +// RRDSET - rename charts + +char *rrdset_strncpyz_name(char *to, const char *from, size_t length) { + char c, *p = to; + + while (length-- && (c = *from++)) { + if(c != '.' && c != '-' && !isalnum(c)) + c = '_'; + + *p++ = c; + } + + *p = '\0'; + + return to; +} + +int rrdset_reset_name(RRDSET *st, const char *name) { + if(unlikely(!strcmp(rrdset_name(st), name))) + return 1; + + RRDHOST *host = st->rrdhost; + + netdata_log_debug(D_RRD_CALLS, "rrdset_reset_name() old: '%s', new: '%s'", rrdset_name(st), name); + + STRING *name_string = rrdset_fix_name(host, rrdset_id(st), rrdset_parts_type(st), string2str(st->name), name); + if(!name_string) return 0; + + if(st->name) { + rrdset_index_del_name(host, st); + string_freez(st->name); + st->name = name_string; + rrdsetvar_rename_all(st); + } + else + st->name = name_string; + + RRDDIM *rd; + rrddim_foreach_read(rd, st) + rrddimvar_rename_all(rd); + rrddim_foreach_done(rd); + + rrdset_index_add_name(host, st); + + rrdset_flag_clear(st, RRDSET_FLAG_EXPORTING_SEND); + rrdset_flag_clear(st, RRDSET_FLAG_EXPORTING_IGNORE); + rrdset_flag_clear(st, RRDSET_FLAG_UPSTREAM_SEND); + rrdset_flag_clear(st, RRDSET_FLAG_UPSTREAM_IGNORE); + rrdset_metadata_updated(st); + + rrdcontext_updated_rrdset_name(st); + return 2; +} + +// get the timestamp of the last entry in the round-robin database +time_t rrdset_last_entry_s(RRDSET *st) { + RRDDIM *rd; + time_t last_entry_s = 0; + + rrddim_foreach_read(rd, st) { + time_t t = rrddim_last_entry_s(rd); + if(t > last_entry_s) last_entry_s = t; + } + rrddim_foreach_done(rd); + + return last_entry_s; +} + +time_t rrdset_last_entry_s_of_tier(RRDSET *st, size_t tier) { + RRDDIM *rd; + time_t last_entry_s = 0; + + rrddim_foreach_read(rd, st) { + time_t t = rrddim_last_entry_s_of_tier(rd, tier); + if(t > last_entry_s) last_entry_s = t; + } + rrddim_foreach_done(rd); + + return last_entry_s; +} + +// get the timestamp of first entry in the round-robin database +time_t rrdset_first_entry_s(RRDSET *st) { + RRDDIM *rd; + time_t first_entry_s = LONG_MAX; + + rrddim_foreach_read(rd, st) { + time_t t = rrddim_first_entry_s(rd); + if(t < first_entry_s) + first_entry_s = t; + } + rrddim_foreach_done(rd); + + if (unlikely(LONG_MAX == first_entry_s)) return 0; + return first_entry_s; +} + +time_t rrdset_first_entry_s_of_tier(RRDSET *st, size_t tier) { + if(unlikely(tier > storage_tiers)) + return 0; + + RRDDIM *rd; + time_t first_entry_s = LONG_MAX; + + rrddim_foreach_read(rd, st) { + time_t t = rrddim_first_entry_s_of_tier(rd, tier); + if(t && t < first_entry_s) + first_entry_s = t; + } + rrddim_foreach_done(rd); + + if (unlikely(LONG_MAX == first_entry_s)) return 0; + return first_entry_s; +} + +void rrdset_get_retention_of_tier_for_collected_chart(RRDSET *st, time_t *first_time_s, time_t *last_time_s, time_t now_s, size_t tier) { + if(!now_s) + now_s = now_realtime_sec(); + + time_t db_first_entry_s = rrdset_first_entry_s_of_tier(st, tier); + time_t db_last_entry_s = st->last_updated.tv_sec; // we assume this is a collected RRDSET + + if(unlikely(!db_last_entry_s)) { + db_last_entry_s = rrdset_last_entry_s_of_tier(st, tier); + + if (unlikely(!db_last_entry_s)) { + // we assume this is a collected RRDSET + db_first_entry_s = 0; + db_last_entry_s = 0; + } + } + + if(unlikely(db_last_entry_s > now_s)) { + internal_error(db_last_entry_s > now_s + 1, + "RRDSET: 'host:%s/chart:%s' latest db time %ld is in the future, adjusting it to now %ld", + rrdhost_hostname(st->rrdhost), rrdset_id(st), + db_last_entry_s, now_s); + db_last_entry_s = now_s; + } + + if(unlikely(db_first_entry_s && db_last_entry_s && db_first_entry_s >= db_last_entry_s)) { + internal_error(db_first_entry_s > db_last_entry_s, + "RRDSET: 'host:%s/chart:%s' oldest db time %ld is bigger than latest db time %ld, adjusting it to (latest time %ld - update every %ld)", + rrdhost_hostname(st->rrdhost), rrdset_id(st), + db_first_entry_s, db_last_entry_s, + db_last_entry_s, (time_t)st->update_every); + db_first_entry_s = db_last_entry_s - st->update_every; + } + + if(unlikely(!db_first_entry_s && db_last_entry_s)) + // this can be the case on the first data collection of a chart + db_first_entry_s = db_last_entry_s - st->update_every; + + *first_time_s = db_first_entry_s; + *last_time_s = db_last_entry_s; +} + +inline void rrdset_is_obsolete___safe_from_collector_thread(RRDSET *st) { + rrdset_pluginsd_receive_unslot(st); + + if(unlikely(!(rrdset_flag_check(st, RRDSET_FLAG_OBSOLETE)))) { +// netdata_log_info("Setting obsolete flag on chart 'host:%s/chart:%s'", +// rrdhost_hostname(st->rrdhost), rrdset_id(st)); + + rrdset_flag_set(st, RRDSET_FLAG_OBSOLETE); + rrdhost_flag_set(st->rrdhost, RRDHOST_FLAG_PENDING_OBSOLETE_CHARTS); + + st->last_accessed_time_s = now_realtime_sec(); + + rrdset_metadata_updated(st); + + // the chart will not get more updates (data collection) + // so, we have to push its definition now + rrdset_push_chart_definition_now(st); + rrdcontext_updated_rrdset_flags(st); + } +} + +inline void rrdset_isnot_obsolete___safe_from_collector_thread(RRDSET *st) { + if(unlikely((rrdset_flag_check(st, RRDSET_FLAG_OBSOLETE)))) { + +// netdata_log_info("Clearing obsolete flag on chart 'host:%s/chart:%s'", +// rrdhost_hostname(st->rrdhost), rrdset_id(st)); + + rrdset_flag_clear(st, RRDSET_FLAG_OBSOLETE); + st->last_accessed_time_s = now_realtime_sec(); + + rrdset_metadata_updated(st); + + // the chart will be pushed upstream automatically + // due to data collection + rrdcontext_updated_rrdset_flags(st); + } +} + +inline void rrdset_update_heterogeneous_flag(RRDSET *st) { + RRDHOST *host = st->rrdhost; + (void)host; + + RRDDIM *rd; + + rrdset_flag_clear(st, RRDSET_FLAG_HOMOGENEOUS_CHECK); + + bool init = false, is_heterogeneous = false; + RRD_ALGORITHM algorithm; + int32_t multiplier; + int32_t divisor; + + rrddim_foreach_read(rd, st) { + if(!init) { + algorithm = rd->algorithm; + multiplier = rd->multiplier; + divisor = ABS(rd->divisor); + init = true; + continue; + } + + if(algorithm != rd->algorithm || multiplier != ABS(rd->multiplier) || divisor != ABS(rd->divisor)) { + if(!rrdset_flag_check(st, RRDSET_FLAG_HETEROGENEOUS)) { + #ifdef NETDATA_INTERNAL_CHECKS + netdata_log_info("Dimension '%s' added on chart '%s' of host '%s' is not homogeneous to other dimensions already present " + "(algorithm is '%s' vs '%s', multiplier is %d vs %d, " + "divisor is %d vs %d).", + rrddim_name(rd), + rrdset_name(st), + rrdhost_hostname(host), + rrd_algorithm_name(rd->algorithm), rrd_algorithm_name(algorithm), + rd->multiplier, multiplier, + rd->divisor, divisor + ); + #endif + rrdset_flag_set(st, RRDSET_FLAG_HETEROGENEOUS); + } + + is_heterogeneous = true; + break; + } + } + rrddim_foreach_done(rd); + + if(!is_heterogeneous) { + rrdset_flag_clear(st, RRDSET_FLAG_HETEROGENEOUS); + rrdcontext_updated_rrdset_flags(st); + } +} + +// ---------------------------------------------------------------------------- +// RRDSET - reset a chart + +void rrdset_reset(RRDSET *st) { + netdata_log_debug(D_RRD_CALLS, "rrdset_reset() %s", rrdset_name(st)); + + st->last_collected_time.tv_sec = 0; + st->last_collected_time.tv_usec = 0; + st->last_updated.tv_sec = 0; + st->last_updated.tv_usec = 0; + st->db.current_entry = 0; + st->counter = 0; + st->counter_done = 0; + + RRDDIM *rd; + rrddim_foreach_read(rd, st) { + rd->collector.last_collected_time.tv_sec = 0; + rd->collector.last_collected_time.tv_usec = 0; + rd->collector.counter = 0; + + if(!rrddim_flag_check(rd, RRDDIM_FLAG_ARCHIVED)) { + for(size_t tier = 0; tier < storage_tiers ;tier++) + storage_engine_store_flush(rd->tiers[tier].db_collection_handle); + } + } + rrddim_foreach_done(rd); +} + +// ---------------------------------------------------------------------------- +// RRDSET - helpers for rrdset_create() + +inline long align_entries_to_pagesize(RRD_MEMORY_MODE mode, long entries) { + if(mode == RRD_MEMORY_MODE_DBENGINE) return 0; + if(mode == RRD_MEMORY_MODE_NONE) return 5; + + if(entries < 5) entries = 5; + if(entries > RRD_HISTORY_ENTRIES_MAX) entries = RRD_HISTORY_ENTRIES_MAX; + + if(mode == RRD_MEMORY_MODE_MAP || mode == RRD_MEMORY_MODE_SAVE || mode == RRD_MEMORY_MODE_RAM) { + long header_size = 0; + + if(mode == RRD_MEMORY_MODE_MAP || mode == RRD_MEMORY_MODE_SAVE) + header_size = (long)rrddim_memory_file_header_size(); + + long page = (long)sysconf(_SC_PAGESIZE); + long size = (long)(header_size + entries * sizeof(storage_number)); + if (unlikely(size % page)) { + size -= (size % page); + size += page; + + long n = (long)((size - header_size) / sizeof(storage_number)); + return n; + } + } + + return entries; +} + +static inline void last_collected_time_align(RRDSET *st) { + st->last_collected_time.tv_sec -= st->last_collected_time.tv_sec % st->update_every; + + if(unlikely(rrdset_flag_check(st, RRDSET_FLAG_STORE_FIRST))) + st->last_collected_time.tv_usec = 0; + else + st->last_collected_time.tv_usec = 500000; +} + +static inline void last_updated_time_align(RRDSET *st) { + st->last_updated.tv_sec -= st->last_updated.tv_sec % st->update_every; + st->last_updated.tv_usec = 0; +} + +// ---------------------------------------------------------------------------- +// RRDSET - free a chart + +void rrdset_free(RRDSET *st) { + if(unlikely(!st)) return; + rrdset_index_del(st->rrdhost, st); +} + +void rrdset_save(RRDSET *st) { + rrdset_memory_file_save(st); + + RRDDIM *rd; + rrddim_foreach_read(rd, st) + rrddim_memory_file_save(rd); + rrddim_foreach_done(rd); +} + +void rrdset_delete_files(RRDSET *st) { + RRDDIM *rd; + + netdata_log_info("Deleting chart '%s' ('%s') from disk...", rrdset_id(st), rrdset_name(st)); + + if(st->rrd_memory_mode == RRD_MEMORY_MODE_SAVE || st->rrd_memory_mode == RRD_MEMORY_MODE_MAP) { + const char *cache_filename = rrdset_cache_filename(st); + if(cache_filename) { + netdata_log_info("Deleting chart header file '%s'.", cache_filename); + if (unlikely(unlink(cache_filename) == -1)) + netdata_log_error("Cannot delete chart header file '%s'", cache_filename); + } + else + netdata_log_error("Cannot find the cache filename of chart '%s'", rrdset_id(st)); + } + + rrddim_foreach_read(rd, st) { + const char *cache_filename = rrddim_cache_filename(rd); + if(!cache_filename) continue; + + netdata_log_info("Deleting dimension file '%s'.", cache_filename); + if(unlikely(unlink(cache_filename) == -1)) + netdata_log_error("Cannot delete dimension file '%s'", cache_filename); + } + rrddim_foreach_done(rd); + + if(st->db.cache_dir) + recursively_delete_dir(st->db.cache_dir, "left-over chart"); +} + +void rrdset_delete_obsolete_dimensions(RRDSET *st) { + RRDDIM *rd; + + netdata_log_info("Deleting dimensions of chart '%s' ('%s') from disk...", rrdset_id(st), rrdset_name(st)); + + rrddim_foreach_read(rd, st) { + if(rrddim_flag_check(rd, RRDDIM_FLAG_OBSOLETE)) { + const char *cache_filename = rrddim_cache_filename(rd); + if(!cache_filename) continue; + netdata_log_info("Deleting dimension file '%s'.", cache_filename); + if(unlikely(unlink(cache_filename) == -1)) + netdata_log_error("Cannot delete dimension file '%s'", cache_filename); + } + } + rrddim_foreach_done(rd); +} + +// ---------------------------------------------------------------------------- +// RRDSET - create a chart + +RRDSET *rrdset_create_custom( + RRDHOST *host + , const char *type + , const char *id + , const char *name + , const char *family + , const char *context + , const char *title + , const char *units + , const char *plugin + , const char *module + , long priority + , int update_every + , RRDSET_TYPE chart_type + , RRD_MEMORY_MODE memory_mode + , long history_entries +) { + if (host != localhost) + host->child_last_chart_command = now_realtime_sec(); + + if(!type || !type[0]) + fatal("Cannot create rrd stats without a type: id '%s', name '%s', family '%s', context '%s', title '%s', units '%s', plugin '%s', module '%s'." + , (id && *id)?id:"<unset>" + , (name && *name)?name:"<unset>" + , (family && *family)?family:"<unset>" + , (context && *context)?context:"<unset>" + , (title && *title)?title:"<unset>" + , (units && *units)?units:"<unset>" + , (plugin && *plugin)?plugin:"<unset>" + , (module && *module)?module:"<unset>" + ); + + if(!id || !id[0]) + fatal("Cannot create rrd stats without an id: type '%s', name '%s', family '%s', context '%s', title '%s', units '%s', plugin '%s', module '%s'." + , type + , (name && *name)?name:"<unset>" + , (family && *family)?family:"<unset>" + , (context && *context)?context:"<unset>" + , (title && *title)?title:"<unset>" + , (units && *units)?units:"<unset>" + , (plugin && *plugin)?plugin:"<unset>" + , (module && *module)?module:"<unset>" + ); + + // ------------------------------------------------------------------------ + // check if it already exists + + char full_id[RRD_ID_LENGTH_MAX + 1]; + snprintfz(full_id, RRD_ID_LENGTH_MAX, "%s.%s", type, id); + + // ------------------------------------------------------------------------ + // allocate it + + netdata_log_debug(D_RRD_CALLS, "Creating RRD_STATS for '%s.%s'.", type, id); + + struct rrdset_constructor tmp = { + .host = host, + .type = type, + .id = id, + .name = name, + .family = family, + .context = context, + .title = title, + .units = units, + .plugin = plugin, + .module = module, + .priority = priority, + .update_every = update_every, + .chart_type = chart_type, + .memory_mode = memory_mode, + .history_entries = history_entries, + }; + + RRDSET *st = rrdset_index_add(host, full_id, &tmp); + return(st); +} + +// ---------------------------------------------------------------------------- +// RRDSET - data collection iteration control + +void rrdset_timed_next(RRDSET *st, struct timeval now, usec_t duration_since_last_update) { + #ifdef NETDATA_INTERNAL_CHECKS + char *discard_reason = NULL; + usec_t discarded = duration_since_last_update; + #endif + + if(unlikely(rrdset_flag_check(st, RRDSET_FLAG_SYNC_CLOCK))) { + // the chart needs to be re-synced to current time + rrdset_flag_clear(st, RRDSET_FLAG_SYNC_CLOCK); + + // discard the duration supplied + duration_since_last_update = 0; + + #ifdef NETDATA_INTERNAL_CHECKS + if(!discard_reason) discard_reason = "SYNC CLOCK FLAG"; + #endif + } + + if(unlikely(!st->last_collected_time.tv_sec)) { + // the first entry + duration_since_last_update = st->update_every * USEC_PER_SEC; + #ifdef NETDATA_INTERNAL_CHECKS + if(!discard_reason) discard_reason = "FIRST DATA COLLECTION"; + #endif + } + else if(unlikely(!duration_since_last_update)) { + // no dt given by the plugin + duration_since_last_update = dt_usec(&now, &st->last_collected_time); + #ifdef NETDATA_INTERNAL_CHECKS + if(!discard_reason) discard_reason = "NO USEC GIVEN BY COLLECTOR"; + #endif + } + else { + // microseconds has the time since the last collection + susec_t since_last_usec = dt_usec_signed(&now, &st->last_collected_time); + + if(unlikely(since_last_usec < 0)) { + // oops! the database is in the future + #ifdef NETDATA_INTERNAL_CHECKS + netdata_log_info("RRD database for chart '%s' on host '%s' is %0.5" NETDATA_DOUBLE_MODIFIER + " secs in the future (counter #%u, update #%u). Adjusting it to current time." + , rrdset_id(st) + , rrdhost_hostname(st->rrdhost) + , (NETDATA_DOUBLE)-since_last_usec / USEC_PER_SEC + , st->counter + , st->counter_done + ); + #endif + + duration_since_last_update = 0; + + #ifdef NETDATA_INTERNAL_CHECKS + if(!discard_reason) discard_reason = "COLLECTION TIME IN FUTURE"; + #endif + } + else if(unlikely((usec_t)since_last_usec > (usec_t)(st->update_every * 5 * USEC_PER_SEC))) { + // oops! the database is too far behind + #ifdef NETDATA_INTERNAL_CHECKS + netdata_log_info("RRD database for chart '%s' on host '%s' is %0.5" NETDATA_DOUBLE_MODIFIER + " secs in the past (counter #%u, update #%u). Adjusting it to current time.", + rrdset_id(st), rrdhost_hostname(st->rrdhost), (NETDATA_DOUBLE)since_last_usec / USEC_PER_SEC, + st->counter, st->counter_done); + #endif + + duration_since_last_update = (usec_t)since_last_usec; + + #ifdef NETDATA_INTERNAL_CHECKS + if(!discard_reason) discard_reason = "COLLECTION TIME TOO FAR IN THE PAST"; + #endif + } + +#ifdef NETDATA_INTERNAL_CHECKS + if(since_last_usec > 0 && (susec_t) duration_since_last_update < since_last_usec) { + static __thread susec_t min_delta = USEC_PER_SEC * 3600, permanent_min_delta = 0; + static __thread time_t last_time_s = 0; + + // the first time initialize it so that it will make the check later + if(last_time_s == 0) last_time_s = now.tv_sec + 60; + + susec_t delta = since_last_usec - (susec_t) duration_since_last_update; + if(delta < min_delta) min_delta = delta; + + if(now.tv_sec >= last_time_s + 60) { + last_time_s = now.tv_sec; + + if(min_delta > permanent_min_delta) { + netdata_log_info("MINIMUM MICROSECONDS DELTA of thread %d increased from %"PRIi64" to %"PRIi64" (+%"PRIi64")", gettid(), permanent_min_delta, min_delta, min_delta - permanent_min_delta); + permanent_min_delta = min_delta; + } + + min_delta = USEC_PER_SEC * 3600; + } + } +#endif + } + + netdata_log_debug(D_RRD_CALLS, "rrdset_timed_next() for chart %s with duration since last update %"PRIu64" usec", rrdset_name(st), duration_since_last_update); + rrdset_debug(st, "NEXT: %"PRIu64" microseconds", duration_since_last_update); + + internal_error(discarded && discarded != duration_since_last_update, + "host '%s', chart '%s': discarded data collection time of %"PRIu64" usec, " + "replaced with %"PRIu64" usec, reason: '%s'" + , rrdhost_hostname(st->rrdhost) + , rrdset_id(st) + , discarded + , duration_since_last_update + , discard_reason?discard_reason:"UNDEFINED" + ); + + st->usec_since_last_update = duration_since_last_update; +} + +inline void rrdset_next_usec_unfiltered(RRDSET *st, usec_t duration_since_last_update) { + if(unlikely(!st->last_collected_time.tv_sec || !duration_since_last_update || (rrdset_flag_check(st, RRDSET_FLAG_SYNC_CLOCK)))) { + // call the full next_usec() function + rrdset_next_usec(st, duration_since_last_update); + return; + } + + st->usec_since_last_update = duration_since_last_update; +} + +inline void rrdset_next_usec(RRDSET *st, usec_t duration_since_last_update) { + struct timeval now; + + now_realtime_timeval(&now); + rrdset_timed_next(st, now, duration_since_last_update); +} + +// ---------------------------------------------------------------------------- +// RRDSET - process the collected values for all dimensions of a chart + +static inline usec_t rrdset_init_last_collected_time(RRDSET *st, struct timeval now) { + st->last_collected_time = now; + last_collected_time_align(st); + + usec_t last_collect_ut = st->last_collected_time.tv_sec * USEC_PER_SEC + st->last_collected_time.tv_usec; + + rrdset_debug(st, "initialized last collected time to %0.3" NETDATA_DOUBLE_MODIFIER, (NETDATA_DOUBLE)last_collect_ut / USEC_PER_SEC); + + return last_collect_ut; +} + +static inline usec_t rrdset_update_last_collected_time(RRDSET *st) { + usec_t last_collect_ut = st->last_collected_time.tv_sec * USEC_PER_SEC + st->last_collected_time.tv_usec; + usec_t ut = last_collect_ut + st->usec_since_last_update; + st->last_collected_time.tv_sec = (time_t) (ut / USEC_PER_SEC); + st->last_collected_time.tv_usec = (suseconds_t) (ut % USEC_PER_SEC); + + rrdset_debug(st, "updated last collected time to %0.3" NETDATA_DOUBLE_MODIFIER, (NETDATA_DOUBLE)last_collect_ut / USEC_PER_SEC); + + return last_collect_ut; +} + +static inline void rrdset_init_last_updated_time(RRDSET *st) { + // copy the last collected time to last updated time + st->last_updated.tv_sec = st->last_collected_time.tv_sec; + st->last_updated.tv_usec = st->last_collected_time.tv_usec; + + if(rrdset_flag_check(st, RRDSET_FLAG_STORE_FIRST)) + st->last_updated.tv_sec -= st->update_every; + + last_updated_time_align(st); +} + +static __thread size_t rrdset_done_statistics_points_stored_per_tier[RRD_STORAGE_TIERS]; + +static inline time_t tier_next_point_time_s(RRDDIM *rd, struct rrddim_tier *t, time_t now_s) { + time_t loop = (time_t)rd->rrdset->update_every * (time_t)t->tier_grouping; + return now_s + loop - ((now_s + loop) % loop); +} + +void store_metric_at_tier(RRDDIM *rd, size_t tier, struct rrddim_tier *t, STORAGE_POINT sp, usec_t now_ut __maybe_unused) { + if (unlikely(!t->next_point_end_time_s)) + t->next_point_end_time_s = tier_next_point_time_s(rd, t, sp.end_time_s); + + if(unlikely(sp.start_time_s >= t->next_point_end_time_s)) { + // flush the virtual point, it is done + + if (likely(!storage_point_is_unset(t->virtual_point))) { + + storage_engine_store_metric( + t->db_collection_handle, + t->next_point_end_time_s * USEC_PER_SEC, + t->virtual_point.sum, + t->virtual_point.min, + t->virtual_point.max, + t->virtual_point.count, + t->virtual_point.anomaly_count, + t->virtual_point.flags); + } + else { + storage_engine_store_metric( + t->db_collection_handle, + t->next_point_end_time_s * USEC_PER_SEC, + NAN, + NAN, + NAN, + 0, + 0, SN_FLAG_NONE); + } + + rrdset_done_statistics_points_stored_per_tier[tier]++; + t->virtual_point.count = 0; // make the point unset + t->next_point_end_time_s = tier_next_point_time_s(rd, t, sp.end_time_s); + } + + // merge the dates into our virtual point + if (unlikely(sp.start_time_s < t->virtual_point.start_time_s)) + t->virtual_point.start_time_s = sp.start_time_s; + + if (likely(sp.end_time_s > t->virtual_point.end_time_s)) + t->virtual_point.end_time_s = sp.end_time_s; + + // merge the values into our virtual point + if (likely(!storage_point_is_gap(sp))) { + // we aggregate only non NULLs into higher tiers + + if (likely(!storage_point_is_unset(t->virtual_point))) { + // merge the collected point to our virtual one + t->virtual_point.sum += sp.sum; + t->virtual_point.min = MIN(t->virtual_point.min, sp.min); + t->virtual_point.max = MAX(t->virtual_point.max, sp.max); + t->virtual_point.count += sp.count; + t->virtual_point.anomaly_count += sp.anomaly_count; + t->virtual_point.flags |= sp.flags; + } + else { + // reset our virtual point to this one + t->virtual_point = sp; + } + } +} +#ifdef NETDATA_LOG_COLLECTION_ERRORS +void rrddim_store_metric_with_trace(RRDDIM *rd, usec_t point_end_time_ut, NETDATA_DOUBLE n, SN_FLAGS flags, const char *function) { +#else // !NETDATA_LOG_COLLECTION_ERRORS +void rrddim_store_metric(RRDDIM *rd, usec_t point_end_time_ut, NETDATA_DOUBLE n, SN_FLAGS flags) { +#endif // !NETDATA_LOG_COLLECTION_ERRORS + + static __thread struct log_stack_entry lgs[] = { + [0] = ND_LOG_FIELD_STR(NDF_NIDL_DIMENSION, NULL), + [1] = ND_LOG_FIELD_END(), + }; + lgs[0].str = rd->id; + log_stack_push(lgs); + +#ifdef NETDATA_LOG_COLLECTION_ERRORS + rd->rrddim_store_metric_count++; + + if(likely(rd->rrddim_store_metric_count > 1)) { + usec_t expected = rd->rrddim_store_metric_last_ut + rd->update_every * USEC_PER_SEC; + + if(point_end_time_ut != rd->rrddim_store_metric_last_ut) { + internal_error(true, + "%s COLLECTION: 'host:%s/chart:%s/dim:%s' granularity %d, collection %zu, expected to store at tier 0 a value at %llu, but it gave %llu [%s%llu usec] (called from %s(), previously by %s())", + (point_end_time_ut < rd->rrddim_store_metric_last_ut) ? "**PAST**" : "GAP", + rrdhost_hostname(rd->rrdset->rrdhost), rrdset_id(rd->rrdset), rrddim_id(rd), + rd->update_every, + rd->rrddim_store_metric_count, + expected, point_end_time_ut, + (point_end_time_ut < rd->rrddim_store_metric_last_ut)?"by -" : "gap ", + expected - point_end_time_ut, + function, + rd->rrddim_store_metric_last_caller?rd->rrddim_store_metric_last_caller:"none"); + } + } + + rd->rrddim_store_metric_last_ut = point_end_time_ut; + rd->rrddim_store_metric_last_caller = function; +#endif // NETDATA_LOG_COLLECTION_ERRORS + + // store the metric on tier 0 + storage_engine_store_metric(rd->tiers[0].db_collection_handle, point_end_time_ut, + n, 0, 0, + 1, 0, flags); + + rrdset_done_statistics_points_stored_per_tier[0]++; + + time_t now_s = (time_t)(point_end_time_ut / USEC_PER_SEC); + + STORAGE_POINT sp = { + .start_time_s = now_s - rd->rrdset->update_every, + .end_time_s = now_s, + .min = n, + .max = n, + .sum = n, + .count = 1, + .anomaly_count = (flags & SN_FLAG_NOT_ANOMALOUS) ? 0 : 1, + .flags = flags + }; + + for(size_t tier = 1; tier < storage_tiers ;tier++) { + if(unlikely(!rd->tiers[tier].db_metric_handle)) continue; + + struct rrddim_tier *t = &rd->tiers[tier]; + + if(!rrddim_option_check(rd, RRDDIM_OPTION_BACKFILLED_HIGH_TIERS)) { + // we have not collected this tier before + // let's fill any gap that may exist + rrdr_fill_tier_gap_from_smaller_tiers(rd, tier, now_s); + rrddim_option_set(rd, RRDDIM_OPTION_BACKFILLED_HIGH_TIERS); + } + + store_metric_at_tier(rd, tier, t, sp, point_end_time_ut); + } + + rrdcontext_collected_rrddim(rd); + log_stack_pop(&lgs); +} + +void store_metric_collection_completed() { + global_statistics_rrdset_done_chart_collection_completed(rrdset_done_statistics_points_stored_per_tier); +} + +// caching of dimensions rrdset_done() and rrdset_done_interpolate() loop through +struct rda_item { + const DICTIONARY_ITEM *item; + RRDDIM *rd; +}; + +static __thread struct rda_item *thread_rda = NULL; +static __thread size_t thread_rda_entries = 0; + +struct rda_item *rrdset_thread_rda_get(size_t *dimensions) { + + if(unlikely(!thread_rda || (*dimensions) > thread_rda_entries)) { + size_t old_mem = thread_rda_entries * sizeof(struct rda_item); + freez(thread_rda); + thread_rda_entries = *dimensions; + size_t new_mem = thread_rda_entries * sizeof(struct rda_item); + thread_rda = mallocz(new_mem); + + __atomic_add_fetch(&netdata_buffers_statistics.rrdset_done_rda_size, new_mem - old_mem, __ATOMIC_RELAXED); + } + + *dimensions = thread_rda_entries; + return thread_rda; +} + +void rrdset_thread_rda_free(void) { + __atomic_sub_fetch(&netdata_buffers_statistics.rrdset_done_rda_size, thread_rda_entries * sizeof(struct rda_item), __ATOMIC_RELAXED); + + freez(thread_rda); + thread_rda = NULL; + thread_rda_entries = 0; +} + +static inline size_t rrdset_done_interpolate( + RRDSET_STREAM_BUFFER *rsb + , RRDSET *st + , struct rda_item *rda_base + , size_t rda_slots + , usec_t update_every_ut + , usec_t last_stored_ut + , usec_t next_store_ut + , usec_t last_collect_ut + , usec_t now_collect_ut + , char store_this_entry + , uint32_t has_reset_value +) { + RRDDIM *rd; + + size_t stored_entries = 0; // the number of entries we have stored in the db, during this call to rrdset_done() + + usec_t first_ut = last_stored_ut, last_ut = 0; + (void)first_ut; + + ssize_t iterations = (ssize_t)((now_collect_ut - last_stored_ut) / (update_every_ut)); + if((now_collect_ut % (update_every_ut)) == 0) iterations++; + + size_t counter = st->counter; + long current_entry = st->db.current_entry; + + SN_FLAGS storage_flags = SN_DEFAULT_FLAGS; + + if (has_reset_value) + storage_flags |= SN_FLAG_RESET; + + for( ; next_store_ut <= now_collect_ut ; last_collect_ut = next_store_ut, next_store_ut += update_every_ut, iterations-- ) { + + internal_error(iterations < 0, + "RRDSET: '%s': iterations calculation wrapped! " + "first_ut = %"PRIu64", last_stored_ut = %"PRIu64", next_store_ut = %"PRIu64", now_collect_ut = %"PRIu64"" + , rrdset_id(st) + , first_ut + , last_stored_ut + , next_store_ut + , now_collect_ut + ); + + rrdset_debug(st, "last_stored_ut = %0.3" NETDATA_DOUBLE_MODIFIER " (last updated time)", (NETDATA_DOUBLE)last_stored_ut/USEC_PER_SEC); + rrdset_debug(st, "next_store_ut = %0.3" NETDATA_DOUBLE_MODIFIER " (next interpolation point)", (NETDATA_DOUBLE)next_store_ut/USEC_PER_SEC); + + last_ut = next_store_ut; + + ml_chart_update_begin(st); + + struct rda_item *rda; + size_t dim_id; + for(dim_id = 0, rda = rda_base ; dim_id < rda_slots ; ++dim_id, ++rda) { + rd = rda->rd; + if(unlikely(!rd)) continue; + + NETDATA_DOUBLE new_value; + + switch(rd->algorithm) { + case RRD_ALGORITHM_INCREMENTAL: + new_value = (NETDATA_DOUBLE) + ( rd->collector.calculated_value + * (NETDATA_DOUBLE)(next_store_ut - last_collect_ut) + / (NETDATA_DOUBLE)(now_collect_ut - last_collect_ut) + ); + + rrdset_debug(st, "%s: CALC2 INC " NETDATA_DOUBLE_FORMAT " = " + NETDATA_DOUBLE_FORMAT + " * (%"PRIu64" - %"PRIu64")" + " / (%"PRIu64" - %"PRIu64"" + , rrddim_name(rd) + , new_value + , rd->collector.calculated_value + , next_store_ut, last_collect_ut + , now_collect_ut, last_collect_ut + ); + + rd->collector.calculated_value -= new_value; + new_value += rd->collector.last_calculated_value; + rd->collector.last_calculated_value = 0; + new_value /= (NETDATA_DOUBLE)st->update_every; + + if(unlikely(next_store_ut - last_stored_ut < update_every_ut)) { + + rrdset_debug(st, "%s: COLLECTION POINT IS SHORT " NETDATA_DOUBLE_FORMAT " - EXTRAPOLATING", + rrddim_name(rd) + , (NETDATA_DOUBLE)(next_store_ut - last_stored_ut) + ); + + new_value = new_value * (NETDATA_DOUBLE)(st->update_every * USEC_PER_SEC) / (NETDATA_DOUBLE)(next_store_ut - last_stored_ut); + } + break; + + case RRD_ALGORITHM_ABSOLUTE: + case RRD_ALGORITHM_PCENT_OVER_ROW_TOTAL: + case RRD_ALGORITHM_PCENT_OVER_DIFF_TOTAL: + default: + if(iterations == 1) { + // this is the last iteration + // do not interpolate + // just show the calculated value + + new_value = rd->collector.calculated_value; + } + else { + // we have missed an update + // interpolate in the middle values + + new_value = (NETDATA_DOUBLE) + ( ( (rd->collector.calculated_value - rd->collector.last_calculated_value) + * (NETDATA_DOUBLE)(next_store_ut - last_collect_ut) + / (NETDATA_DOUBLE)(now_collect_ut - last_collect_ut) + ) + + rd->collector.last_calculated_value + ); + + rrdset_debug(st, "%s: CALC2 DEF " NETDATA_DOUBLE_FORMAT " = (((" + "(" NETDATA_DOUBLE_FORMAT " - " NETDATA_DOUBLE_FORMAT ")" + " * %"PRIu64"" + " / %"PRIu64") + " NETDATA_DOUBLE_FORMAT, rrddim_name(rd) + , new_value + , rd->collector.calculated_value, rd->collector.last_calculated_value + , (next_store_ut - first_ut) + , (now_collect_ut - first_ut), rd->collector.last_calculated_value + ); + } + break; + } + + time_t current_time_s = (time_t) (next_store_ut / USEC_PER_SEC); + + if(unlikely(!store_this_entry)) { + (void) ml_dimension_is_anomalous(rd, current_time_s, 0, false); + + if(rsb->wb && rsb->v2) + rrddim_push_metrics_v2(rsb, rd, next_store_ut, NAN, SN_FLAG_NONE); + + rrddim_store_metric(rd, next_store_ut, NAN, SN_FLAG_NONE); + continue; + } + + if(likely(rrddim_check_updated(rd) && rd->collector.counter > 1 && iterations < gap_when_lost_iterations_above)) { + uint32_t dim_storage_flags = storage_flags; + + if (ml_dimension_is_anomalous(rd, current_time_s, new_value, true)) { + // clear anomaly bit: 0 -> is anomalous, 1 -> not anomalous + dim_storage_flags &= ~((storage_number)SN_FLAG_NOT_ANOMALOUS); + } + + if(rsb->wb && rsb->v2) + rrddim_push_metrics_v2(rsb, rd, next_store_ut, new_value, dim_storage_flags); + + rrddim_store_metric(rd, next_store_ut, new_value, dim_storage_flags); + rd->collector.last_stored_value = new_value; + } + else { + (void) ml_dimension_is_anomalous(rd, current_time_s, 0, false); + + rrdset_debug(st, "%s: STORE[%ld] = NON EXISTING ", rrddim_name(rd), current_entry); + + if(rsb->wb && rsb->v2) + rrddim_push_metrics_v2(rsb, rd, next_store_ut, NAN, SN_FLAG_NONE); + + rrddim_store_metric(rd, next_store_ut, NAN, SN_FLAG_NONE); + rd->collector.last_stored_value = NAN; + } + + stored_entries++; + } + + ml_chart_update_end(st); + + // reset the storage flags for the next point, if any; + storage_flags = SN_DEFAULT_FLAGS; + + st->counter = ++counter; + st->db.current_entry = current_entry = ((current_entry + 1) >= st->db.entries) ? 0 : current_entry + 1; + + st->last_updated.tv_sec = (time_t) (last_ut / USEC_PER_SEC); + st->last_updated.tv_usec = 0; + + last_stored_ut = next_store_ut; + } + +/* + st->counter = counter; + st->current_entry = current_entry; + + if(likely(last_ut)) { + st->last_updated.tv_sec = (time_t) (last_ut / USEC_PER_SEC); + st->last_updated.tv_usec = 0; + } +*/ + + return stored_entries; +} + +void rrdset_done(RRDSET *st) { + struct timeval now; + + now_realtime_timeval(&now); + rrdset_timed_done(st, now, /* pending_rrdset_next = */ st->counter_done != 0); +} + +void rrdset_timed_done(RRDSET *st, struct timeval now, bool pending_rrdset_next) { + if(unlikely(!service_running(SERVICE_COLLECTORS))) return; + + RRDSET_STREAM_BUFFER stream_buffer = { .wb = NULL, }; + if(unlikely(rrdhost_has_rrdpush_sender_enabled(st->rrdhost))) + stream_buffer = rrdset_push_metric_initialize(st, now.tv_sec); + + spinlock_lock(&st->data_collection_lock); + + if (pending_rrdset_next) + rrdset_timed_next(st, now, 0ULL); + + netdata_log_debug(D_RRD_CALLS, "rrdset_done() for chart '%s'", rrdset_name(st)); + + RRDDIM *rd; + + char + store_this_entry = 1, // boolean: 1 = store this entry, 0 = don't store this entry + first_entry = 0; // boolean: 1 = this is the first entry seen for this chart, 0 = all other entries + + usec_t + last_collect_ut = 0, // the timestamp in microseconds, of the last collected value + now_collect_ut = 0, // the timestamp in microseconds, of this collected value (this is NOW) + last_stored_ut = 0, // the timestamp in microseconds, of the last stored entry in the db + next_store_ut = 0, // the timestamp in microseconds, of the next entry to store in the db + update_every_ut = st->update_every * USEC_PER_SEC; // st->update_every in microseconds + + RRDSET_FLAGS rrdset_flags = rrdset_flag_check(st, ~0); + if(unlikely(rrdset_flags & RRDSET_FLAG_COLLECTION_FINISHED)) { + spinlock_unlock(&st->data_collection_lock); + return; + } + + if (unlikely(rrdset_flags & RRDSET_FLAG_OBSOLETE)) { + netdata_log_error("Chart '%s' has the OBSOLETE flag set, but it is collected.", rrdset_id(st)); + rrdset_isnot_obsolete___safe_from_collector_thread(st); + } + + // check if the chart has a long time to be updated + if(unlikely(st->usec_since_last_update > MAX(st->db.entries, 60) * update_every_ut)) { + nd_log_daemon(NDLP_DEBUG, "host '%s', chart '%s': took too long to be updated (counter #%u, update #%u, %0.3" NETDATA_DOUBLE_MODIFIER + " secs). Resetting it.", rrdhost_hostname(st->rrdhost), rrdset_id(st), st->counter, st->counter_done, + (NETDATA_DOUBLE)st->usec_since_last_update / USEC_PER_SEC); + rrdset_reset(st); + st->usec_since_last_update = update_every_ut; + store_this_entry = 0; + first_entry = 1; + } + + rrdset_debug(st, "microseconds since last update: %"PRIu64"", st->usec_since_last_update); + + // set last_collected_time + if(unlikely(!st->last_collected_time.tv_sec)) { + // it is the first entry + // set the last_collected_time to now + last_collect_ut = rrdset_init_last_collected_time(st, now) - update_every_ut; + + // the first entry should not be stored + store_this_entry = 0; + first_entry = 1; + } + else { + // it is not the first entry + // calculate the proper last_collected_time, using usec_since_last_update + last_collect_ut = rrdset_update_last_collected_time(st); + } + + // if this set has not been updated in the past + // we fake the last_update time to be = now - usec_since_last_update + if(unlikely(!st->last_updated.tv_sec)) { + // it has never been updated before + // set a fake last_updated, in the past using usec_since_last_update + rrdset_init_last_updated_time(st); + + // the first entry should not be stored + store_this_entry = 0; + first_entry = 1; + } + + // check if we will re-write the entire data set + if(unlikely(dt_usec(&st->last_collected_time, &st->last_updated) > st->db.entries * update_every_ut && + st->rrd_memory_mode != RRD_MEMORY_MODE_DBENGINE)) { + netdata_log_info( + "'%s': too old data (last updated at %"PRId64".%"PRId64", last collected at %"PRId64".%"PRId64"). " + "Resetting it. Will not store the next entry.", + rrdset_id(st), + (int64_t)st->last_updated.tv_sec, + (int64_t)st->last_updated.tv_usec, + (int64_t)st->last_collected_time.tv_sec, + (int64_t)st->last_collected_time.tv_usec); + rrdset_reset(st); + rrdset_init_last_updated_time(st); + + st->usec_since_last_update = update_every_ut; + + // the first entry should not be stored + store_this_entry = 0; + first_entry = 1; + } + + // these are the 3 variables that will help us in interpolation + // last_stored_ut = the last time we added a value to the storage + // now_collect_ut = the time the current value has been collected + // next_store_ut = the time of the next interpolation point + now_collect_ut = st->last_collected_time.tv_sec * USEC_PER_SEC + st->last_collected_time.tv_usec; + last_stored_ut = st->last_updated.tv_sec * USEC_PER_SEC + st->last_updated.tv_usec; + next_store_ut = (st->last_updated.tv_sec + st->update_every) * USEC_PER_SEC; + + if(unlikely(!st->counter_done)) { + // set a fake last_updated to jump to current time + rrdset_init_last_updated_time(st); + + last_stored_ut = st->last_updated.tv_sec * USEC_PER_SEC + st->last_updated.tv_usec; + next_store_ut = (st->last_updated.tv_sec + st->update_every) * USEC_PER_SEC; + + if(unlikely(rrdset_flags & RRDSET_FLAG_STORE_FIRST)) { + store_this_entry = 1; + last_collect_ut = next_store_ut - update_every_ut; + + rrdset_debug(st, "Fixed first entry."); + } + else { + store_this_entry = 0; + + rrdset_debug(st, "Will not store the next entry."); + } + } + + st->counter_done++; + + if(stream_buffer.wb && !stream_buffer.v2) + rrdset_push_metrics_v1(&stream_buffer, st); + + uint32_t has_reset_value = 0; + + size_t rda_slots = dictionary_entries(st->rrddim_root_index); + struct rda_item *rda_base = rrdset_thread_rda_get(&rda_slots); + + size_t dim_id; + size_t dimensions = 0; + struct rda_item *rda = rda_base; + total_number collected_total = 0; + total_number last_collected_total = 0; + rrddim_foreach_read(rd, st) { + if(rd_dfe.counter >= rda_slots) + break; + + rda = &rda_base[dimensions++]; + + if(rrddim_flag_check(rd, RRDDIM_FLAG_ARCHIVED)) { + rda->item = NULL; + rda->rd = NULL; + continue; + } + + // store the dimension in the array + rda->item = dictionary_acquired_item_dup(st->rrddim_root_index, rd_dfe.item); + rda->rd = dictionary_acquired_item_value(rda->item); + + // calculate totals + if(likely(rrddim_check_updated(rd))) { + // if the new is smaller than the old (an overflow, or reset), set the old equal to the new + // to reset the calculation (it will give zero as the calculation for this second) + if(unlikely(rd->algorithm == RRD_ALGORITHM_PCENT_OVER_DIFF_TOTAL && rd->collector.last_collected_value > rd->collector.collected_value)) { + netdata_log_debug(D_RRD_STATS, "'%s' / '%s': RESET or OVERFLOW. Last collected value = " COLLECTED_NUMBER_FORMAT ", current = " COLLECTED_NUMBER_FORMAT + , rrdset_id(st) + , rrddim_name(rd) + , rd->collector.last_collected_value + , rd->collector.collected_value + ); + + if(!(rrddim_option_check(rd, RRDDIM_OPTION_DONT_DETECT_RESETS_OR_OVERFLOWS))) + has_reset_value = 1; + + rd->collector.last_collected_value = rd->collector.collected_value; + } + + last_collected_total += rd->collector.last_collected_value; + collected_total += rd->collector.collected_value; + + if(unlikely(rrddim_flag_check(rd, RRDDIM_FLAG_OBSOLETE))) { + netdata_log_error("Dimension %s in chart '%s' has the OBSOLETE flag set, but it is collected.", rrddim_name(rd), rrdset_id(st)); + rrddim_isnot_obsolete___safe_from_collector_thread(st, rd); + } + } + } + rrddim_foreach_done(rd); + rda_slots = dimensions; + + rrdset_debug(st, "last_collect_ut = %0.3" NETDATA_DOUBLE_MODIFIER " (last collection time)", (NETDATA_DOUBLE)last_collect_ut/USEC_PER_SEC); + rrdset_debug(st, "now_collect_ut = %0.3" NETDATA_DOUBLE_MODIFIER " (current collection time)", (NETDATA_DOUBLE)now_collect_ut/USEC_PER_SEC); + rrdset_debug(st, "last_stored_ut = %0.3" NETDATA_DOUBLE_MODIFIER " (last updated time)", (NETDATA_DOUBLE)last_stored_ut/USEC_PER_SEC); + rrdset_debug(st, "next_store_ut = %0.3" NETDATA_DOUBLE_MODIFIER " (next interpolation point)", (NETDATA_DOUBLE)next_store_ut/USEC_PER_SEC); + + // process all dimensions to calculate their values + // based on the collected figures only + // at this stage we do not interpolate anything + for(dim_id = 0, rda = rda_base ; dim_id < rda_slots ; ++dim_id, ++rda) { + rd = rda->rd; + if(unlikely(!rd)) continue; + + if(unlikely(!rrddim_check_updated(rd))) { + rd->collector.calculated_value = 0; + continue; + } + + rrdset_debug(st, "%s: START " + " last_collected_value = " COLLECTED_NUMBER_FORMAT + " collected_value = " COLLECTED_NUMBER_FORMAT + " last_calculated_value = " NETDATA_DOUBLE_FORMAT + " calculated_value = " NETDATA_DOUBLE_FORMAT + , rrddim_name(rd) + , rd->collector.last_collected_value + , rd->collector.collected_value + , rd->collector.last_calculated_value + , rd->collector.calculated_value + ); + + switch(rd->algorithm) { + case RRD_ALGORITHM_ABSOLUTE: + rd->collector.calculated_value = (NETDATA_DOUBLE)rd->collector.collected_value + * (NETDATA_DOUBLE)rd->multiplier + / (NETDATA_DOUBLE)rd->divisor; + + rrdset_debug(st, "%s: CALC ABS/ABS-NO-IN " NETDATA_DOUBLE_FORMAT " = " + COLLECTED_NUMBER_FORMAT + " * " NETDATA_DOUBLE_FORMAT + " / " NETDATA_DOUBLE_FORMAT + , rrddim_name(rd) + , rd->collector.calculated_value + , rd->collector.collected_value + , (NETDATA_DOUBLE)rd->multiplier + , (NETDATA_DOUBLE)rd->divisor + ); + break; + + case RRD_ALGORITHM_PCENT_OVER_ROW_TOTAL: + if(unlikely(!collected_total)) + rd->collector.calculated_value = 0; + else + // the percentage of the current value + // over the total of all dimensions + rd->collector.calculated_value = + (NETDATA_DOUBLE)100 + * (NETDATA_DOUBLE)rd->collector.collected_value + / (NETDATA_DOUBLE)collected_total; + + rrdset_debug(st, "%s: CALC PCENT-ROW " NETDATA_DOUBLE_FORMAT " = 100" + " * " COLLECTED_NUMBER_FORMAT + " / " COLLECTED_NUMBER_FORMAT + , rrddim_name(rd) + , rd->collector.calculated_value + , rd->collector.collected_value + , collected_total + ); + break; + + case RRD_ALGORITHM_INCREMENTAL: + if(unlikely(rd->collector.counter <= 1)) { + rd->collector.calculated_value = 0; + continue; + } + + // If the new is smaller than the old (an overflow, or reset), set the old equal to the new + // to reset the calculation (it will give zero as the calculation for this second). + // It is imperative to set the comparison to uint64_t since type collected_number is signed and + // produces wrong results as far as incremental counters are concerned. + if(unlikely((uint64_t)rd->collector.last_collected_value > (uint64_t)rd->collector.collected_value)) { + netdata_log_debug(D_RRD_STATS, "'%s' / '%s': RESET or OVERFLOW. Last collected value = " COLLECTED_NUMBER_FORMAT ", current = " COLLECTED_NUMBER_FORMAT + , rrdset_id(st) + , rrddim_name(rd) + , rd->collector.last_collected_value + , rd->collector.collected_value); + + if(!(rrddim_option_check(rd, RRDDIM_OPTION_DONT_DETECT_RESETS_OR_OVERFLOWS))) + has_reset_value = 1; + + uint64_t last = (uint64_t)rd->collector.last_collected_value; + uint64_t new = (uint64_t)rd->collector.collected_value; + uint64_t max = (uint64_t)rd->collector.collected_value_max; + uint64_t cap = 0; + + // Signed values are handled by exploiting two's complement which will produce positive deltas + if (max > 0x00000000FFFFFFFFULL) + cap = 0xFFFFFFFFFFFFFFFFULL; // handles signed and unsigned 64-bit counters + else + cap = 0x00000000FFFFFFFFULL; // handles signed and unsigned 32-bit counters + + uint64_t delta = cap - last + new; + uint64_t max_acceptable_rate = (cap / 100) * MAX_INCREMENTAL_PERCENT_RATE; + + // If the delta is less than the maximum acceptable rate and the previous value was near the cap + // then this is an overflow. There can be false positives such that a reset is detected as an + // overflow. + // TODO: remember recent history of rates and compare with current rate to reduce this chance. + if (delta < max_acceptable_rate) { + rd->collector.calculated_value += + (NETDATA_DOUBLE) delta + * (NETDATA_DOUBLE) rd->multiplier + / (NETDATA_DOUBLE) rd->divisor; + } else { + // This is a reset. Any overflow with a rate greater than MAX_INCREMENTAL_PERCENT_RATE will also + // be detected as a reset instead. + rd->collector.calculated_value += (NETDATA_DOUBLE)0; + } + } + else { + rd->collector.calculated_value += + (NETDATA_DOUBLE) (rd->collector.collected_value - rd->collector.last_collected_value) + * (NETDATA_DOUBLE) rd->multiplier + / (NETDATA_DOUBLE) rd->divisor; + } + + rrdset_debug(st, "%s: CALC INC PRE " NETDATA_DOUBLE_FORMAT " = (" + COLLECTED_NUMBER_FORMAT " - " COLLECTED_NUMBER_FORMAT + ")" + " * " NETDATA_DOUBLE_FORMAT + " / " NETDATA_DOUBLE_FORMAT + , rrddim_name(rd) + , rd->collector.calculated_value + , rd->collector.collected_value, rd->collector.last_collected_value + , (NETDATA_DOUBLE)rd->multiplier + , (NETDATA_DOUBLE)rd->divisor + ); + break; + + case RRD_ALGORITHM_PCENT_OVER_DIFF_TOTAL: + if(unlikely(rd->collector.counter <= 1)) { + rd->collector.calculated_value = 0; + continue; + } + + // the percentage of the current increment + // over the increment of all dimensions together + if(unlikely(collected_total == last_collected_total)) + rd->collector.calculated_value = 0; + else + rd->collector.calculated_value = + (NETDATA_DOUBLE)100 + * (NETDATA_DOUBLE)(rd->collector.collected_value - rd->collector.last_collected_value) + / (NETDATA_DOUBLE)(collected_total - last_collected_total); + + rrdset_debug(st, "%s: CALC PCENT-DIFF " NETDATA_DOUBLE_FORMAT " = 100" + " * (" COLLECTED_NUMBER_FORMAT " - " COLLECTED_NUMBER_FORMAT ")" + " / (" COLLECTED_NUMBER_FORMAT " - " COLLECTED_NUMBER_FORMAT ")" + , rrddim_name(rd) + , rd->collector.calculated_value + , rd->collector.collected_value, rd->collector.last_collected_value + , collected_total, last_collected_total + ); + break; + + default: + // make the default zero, to make sure + // it gets noticed when we add new types + rd->collector.calculated_value = 0; + + rrdset_debug(st, "%s: CALC " NETDATA_DOUBLE_FORMAT " = 0" + , rrddim_name(rd) + , rd->collector.calculated_value + ); + break; + } + + rrdset_debug(st, "%s: PHASE2 " + " last_collected_value = " COLLECTED_NUMBER_FORMAT + " collected_value = " COLLECTED_NUMBER_FORMAT + " last_calculated_value = " NETDATA_DOUBLE_FORMAT + " calculated_value = " NETDATA_DOUBLE_FORMAT + , rrddim_name(rd) + , rd->collector.last_collected_value + , rd->collector.collected_value + , rd->collector.last_calculated_value + , rd->collector.calculated_value + ); + } + + // at this point we have all the calculated values ready + // it is now time to interpolate values on a second boundary + +// #ifdef NETDATA_INTERNAL_CHECKS +// if(unlikely(now_collect_ut < next_store_ut && st->counter_done > 1)) { +// // this is collected in the same interpolation point +// rrdset_debug(st, "THIS IS IN THE SAME INTERPOLATION POINT"); +// netdata_log_info("INTERNAL CHECK: host '%s', chart '%s' collection %zu is in the same interpolation point: short by %llu microseconds", st->rrdhost->hostname, rrdset_name(st), st->counter_done, next_store_ut - now_collect_ut); +// } +// #endif + + rrdset_done_interpolate( + &stream_buffer + , st + , rda_base + , rda_slots + , update_every_ut + , last_stored_ut + , next_store_ut + , last_collect_ut + , now_collect_ut + , store_this_entry + , has_reset_value + ); + + for(dim_id = 0, rda = rda_base ; dim_id < rda_slots ; ++dim_id, ++rda) { + rd = rda->rd; + if(unlikely(!rd)) continue; + + if(unlikely(!rrddim_check_updated(rd))) + continue; + + rrdset_debug(st, "%s: setting last_collected_value (old: " COLLECTED_NUMBER_FORMAT ") to last_collected_value (new: " COLLECTED_NUMBER_FORMAT ")", rrddim_name(rd), rd->collector.last_collected_value, rd->collector.collected_value); + + rd->collector.last_collected_value = rd->collector.collected_value; + + switch(rd->algorithm) { + case RRD_ALGORITHM_INCREMENTAL: + if(unlikely(!first_entry)) { + rrdset_debug(st, "%s: setting last_calculated_value (old: " NETDATA_DOUBLE_FORMAT ") to " + "last_calculated_value (new: " NETDATA_DOUBLE_FORMAT ")" + , rrddim_name(rd) + , rd->collector.last_calculated_value + rd->collector.calculated_value + , rd->collector.calculated_value); + + rd->collector.last_calculated_value += rd->collector.calculated_value; + } + else { + rrdset_debug(st, "THIS IS THE FIRST POINT"); + } + break; + + case RRD_ALGORITHM_ABSOLUTE: + case RRD_ALGORITHM_PCENT_OVER_ROW_TOTAL: + case RRD_ALGORITHM_PCENT_OVER_DIFF_TOTAL: + rrdset_debug(st, "%s: setting last_calculated_value (old: " NETDATA_DOUBLE_FORMAT ") to " + "last_calculated_value (new: " NETDATA_DOUBLE_FORMAT ")" + , rrddim_name(rd) + , rd->collector.last_calculated_value + , rd->collector.calculated_value); + + rd->collector.last_calculated_value = rd->collector.calculated_value; + break; + } + + rd->collector.calculated_value = 0; + rd->collector.collected_value = 0; + rrddim_clear_updated(rd); + + rrdset_debug(st, "%s: END " + " last_collected_value = " COLLECTED_NUMBER_FORMAT + " collected_value = " COLLECTED_NUMBER_FORMAT + " last_calculated_value = " NETDATA_DOUBLE_FORMAT + " calculated_value = " NETDATA_DOUBLE_FORMAT + , rrddim_name(rd) + , rd->collector.last_collected_value + , rd->collector.collected_value + , rd->collector.last_calculated_value + , rd->collector.calculated_value + ); + } + + spinlock_unlock(&st->data_collection_lock); + rrdset_push_metrics_finished(&stream_buffer, st); + + // ALL DONE ABOUT THE DATA UPDATE + // -------------------------------------------------------------------- + + if(unlikely(st->rrd_memory_mode == RRD_MEMORY_MODE_MAP)) { + // update the memory mapped files with the latest values + + rrdset_memory_file_update(st); + + for(dim_id = 0, rda = rda_base; dim_id < rda_slots ; ++dim_id, ++rda) { + rd = rda->rd; + if(unlikely(!rd)) continue; + rrddim_memory_file_update(rd); + } + } + + for(dim_id = 0, rda = rda_base; dim_id < rda_slots ; ++dim_id, ++rda) { + rd = rda->rd; + if(unlikely(!rd)) continue; + + dictionary_acquired_item_release(st->rrddim_root_index, rda->item); + rda->item = NULL; + rda->rd = NULL; + } + + rrdcontext_collected_rrdset(st); + + store_metric_collection_completed(); +} + +time_t rrdset_set_update_every_s(RRDSET *st, time_t update_every_s) { + if(unlikely(update_every_s == st->update_every)) + return st->update_every; + + internal_error(true, "RRDSET '%s' switching update every from %d to %d", + rrdset_id(st), (int)st->update_every, (int)update_every_s); + + time_t prev_update_every_s = (time_t) st->update_every; + st->update_every = (int) update_every_s; + + // switch update every to the storage engine + RRDDIM *rd; + rrddim_foreach_read(rd, st) { + for (size_t tier = 0; tier < storage_tiers; tier++) { + if (rd->tiers[tier].db_collection_handle) + storage_engine_store_change_collection_frequency( + rd->tiers[tier].db_collection_handle, + (int)(st->rrdhost->db[tier].tier_grouping * st->update_every)); + } + } + rrddim_foreach_done(rd); + + return prev_update_every_s; +} + +// ---------------------------------------------------------------------------- +// compatibility layer for RRDSET files v019 + +#define RRDSET_MAGIC_V019 "NETDATA RRD SET FILE V019" +#define RRD_ID_LENGTH_MAX_V019 200 + +struct avl_element_v019 { + void *avl_link[2]; + signed char avl_balance; +}; +struct avl_tree_type_v019 { + void *root; + int (*compar)(void *a, void *b); +}; +struct avl_tree_lock_v019 { + struct avl_tree_type_v019 avl_tree; + pthread_rwlock_t rwlock; +}; +struct rrdset_map_save_v019 { + struct avl_element_v019 avl; // ignored + struct avl_element_v019 avlname; // ignored + char id[RRD_ID_LENGTH_MAX_V019 + 1]; // check to reset all - update on load + void *name; // ignored + void *unused_ptr; // ignored + void *type; // ignored + void *family; // ignored + void *title; // ignored + void *units; // ignored + void *context; // ignored + uint32_t hash_context; // ignored + uint32_t chart_type; // ignored + int update_every; // check to reset all - update on load + long entries; // check to reset all - update on load + long current_entry; // NEEDS TO BE UPDATED - FIXED ON LOAD + uint32_t flags; // ignored + void *exporting_flags; // ignored + int gap_when_lost_iterations_above; // ignored + long priority; // ignored + uint32_t rrd_memory_mode; // ignored + void *cache_dir; // ignored + char cache_filename[FILENAME_MAX+1]; // ignored - update on load + pthread_rwlock_t rrdset_rwlock; // ignored + size_t counter; // NEEDS TO BE UPDATED - maintained on load + size_t counter_done; // ignored + union { // + time_t last_accessed_time_s; // ignored + time_t last_entry_s; // ignored + }; // + time_t upstream_resync_time; // ignored + void *plugin_name; // ignored + void *module_name; // ignored + void *chart_uuid; // ignored + void *state; // ignored + size_t unused[3]; // ignored + size_t rrddim_page_alignment; // ignored + uint32_t hash; // ignored + uint32_t hash_name; // ignored + usec_t usec_since_last_update; // NEEDS TO BE UPDATED - maintained on load + struct timeval last_updated; // NEEDS TO BE UPDATED - check to reset all - fixed on load + struct timeval last_collected_time; // ignored + long long collected_total; // ignored + long long last_collected_total; // ignored + void *rrdfamily; // ignored + void *rrdhost; // ignored + void *next; // ignored + long double green; // ignored + long double red; // ignored + struct avl_tree_lock_v019 rrdvar_root_index; // ignored + void *variables; // ignored + void *alarms; // ignored + unsigned long memsize; // check to reset all - update on load + char magic[sizeof(RRDSET_MAGIC_V019) + 1]; // check to reset all - update on load + struct avl_tree_lock_v019 dimensions_index; // ignored + void *dimensions; // ignored +}; + +void rrdset_memory_file_update(RRDSET *st) { + if(!st->db.st_on_file) return; + struct rrdset_map_save_v019 *st_on_file = st->db.st_on_file; + + st_on_file->current_entry = st->db.current_entry; + st_on_file->counter = st->counter; + st_on_file->usec_since_last_update = st->usec_since_last_update; + st_on_file->last_updated.tv_sec = st->last_updated.tv_sec; + st_on_file->last_updated.tv_usec = st->last_updated.tv_usec; +} + +const char *rrdset_cache_filename(RRDSET *st) { + if(!st->db.st_on_file) return NULL; + struct rrdset_map_save_v019 *st_on_file = st->db.st_on_file; + return st_on_file->cache_filename; +} + +const char *rrdset_cache_dir(RRDSET *st) { + if(!st->db.cache_dir) + st->db.cache_dir = rrdhost_cache_dir_for_rrdset_alloc(st->rrdhost, rrdset_id(st)); + + return st->db.cache_dir; +} + +void rrdset_memory_file_free(RRDSET *st) { + if(!st->db.st_on_file) return; + + // needed for memory mode map, to save the latest state + rrdset_memory_file_update(st); + + struct rrdset_map_save_v019 *st_on_file = st->db.st_on_file; + __atomic_sub_fetch(&rrddim_db_memory_size, st_on_file->memsize, __ATOMIC_RELAXED); + netdata_munmap(st_on_file, st_on_file->memsize); + + // remove the pointers from the RRDDIM + st->db.st_on_file = NULL; +} + +void rrdset_memory_file_save(RRDSET *st) { + if(!st->db.st_on_file) return; + + rrdset_memory_file_update(st); + + struct rrdset_map_save_v019 *st_on_file = st->db.st_on_file; + if(st_on_file->rrd_memory_mode != RRD_MEMORY_MODE_SAVE) return; + + memory_file_save(st_on_file->cache_filename, st->db.st_on_file, st_on_file->memsize); +} + +bool rrdset_memory_load_or_create_map_save(RRDSET *st, RRD_MEMORY_MODE memory_mode) { + if(memory_mode != RRD_MEMORY_MODE_SAVE && memory_mode != RRD_MEMORY_MODE_MAP) + return false; + + char fullfilename[FILENAME_MAX + 1]; + snprintfz(fullfilename, FILENAME_MAX, "%s/main.db", rrdset_cache_dir(st)); + + unsigned long size = sizeof(struct rrdset_map_save_v019); + struct rrdset_map_save_v019 *st_on_file = (struct rrdset_map_save_v019 *)netdata_mmap( + fullfilename, size, ((memory_mode == RRD_MEMORY_MODE_MAP) ? MAP_SHARED : MAP_PRIVATE), 0, false, NULL); + + if(!st_on_file) return false; + + time_t now_s = now_realtime_sec(); + + st_on_file->magic[sizeof(RRDSET_MAGIC_V019)] = '\0'; + if(strcmp(st_on_file->magic, RRDSET_MAGIC_V019) != 0) { + netdata_log_info("Initializing file '%s'.", fullfilename); + memset(st_on_file, 0, size); + } + else if(strncmp(st_on_file->id, rrdset_id(st), RRD_ID_LENGTH_MAX_V019) != 0) { + netdata_log_error("File '%s' contents are not for chart '%s'. Clearing it.", fullfilename, rrdset_id(st)); + memset(st_on_file, 0, size); + } + else if(st_on_file->memsize != size || st_on_file->entries != st->db.entries) { + netdata_log_error("File '%s' does not have the desired size. Clearing it.", fullfilename); + memset(st_on_file, 0, size); + } + else if(st_on_file->update_every != st->update_every) { + netdata_log_error("File '%s' does not have the desired granularity. Clearing it.", fullfilename); + memset(st_on_file, 0, size); + } + else if((now_s - st_on_file->last_updated.tv_sec) > (long)st->update_every * (long)st->db.entries) { + netdata_log_info("File '%s' is too old. Clearing it.", fullfilename); + memset(st_on_file, 0, size); + } + else if(st_on_file->last_updated.tv_sec > now_s + st->update_every) { + netdata_log_error("File '%s' refers to the future by %zd secs. Resetting it to now.", fullfilename, (ssize_t)(st_on_file->last_updated.tv_sec - now_s)); + st_on_file->last_updated.tv_sec = now_s; + } + + if(st_on_file->current_entry >= st_on_file->entries) + st_on_file->current_entry = 0; + + // make sure the database is aligned + bool align_last_updated = false; + if(st_on_file->last_updated.tv_sec) { + st_on_file->update_every = st->update_every; + align_last_updated = true; + } + + // copy the useful values to st + st->db.current_entry = st_on_file->current_entry; + st->counter = st_on_file->counter; + st->usec_since_last_update = st_on_file->usec_since_last_update; + st->last_updated.tv_sec = st_on_file->last_updated.tv_sec; + st->last_updated.tv_usec = st_on_file->last_updated.tv_usec; + + // link it to st + st->db.st_on_file = st_on_file; + + // clear everything + memset(st_on_file, 0, size); + + // set the values we need + strncpyz(st_on_file->id, rrdset_id(st), RRD_ID_LENGTH_MAX_V019); + strcpy(st_on_file->cache_filename, fullfilename); + strcpy(st_on_file->magic, RRDSET_MAGIC_V019); + st_on_file->memsize = size; + st_on_file->entries = st->db.entries; + st_on_file->update_every = st->update_every; + st_on_file->rrd_memory_mode = memory_mode; + + if(align_last_updated) + last_updated_time_align(st); + + // copy the useful values back to st_on_file + rrdset_memory_file_update(st); + + __atomic_add_fetch(&rrddim_db_memory_size, st_on_file->memsize, __ATOMIC_RELAXED); + return true; +} |