diff options
| author | Emmanuel Vasilakis <mrzammler@mm.st> | 2021-09-19 14:11:17 +0300 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2021-09-19 14:11:17 +0300 |
| commit | b87473c481283bbd936661fb15c3bd63fa74dbca (patch) | |
| tree | 62743e2236957dbb5ed3a0c1076ce6f89b76098d /health/health.c | |
| parent | 686e5782d109fe79f2c5ec6aa44850f3e449b679 (diff) | |
Use sqlite to store the health log and alert configurations. (#11399)
* Rebased
* use sql health log if it exists
* store alert config in sqlite
* move unlock before loop
* fix warnings
* remove hash message
* check return from counting health log
* remove check of hostname when reading log
* try to create the health log table to catch accidental removals of it
* fix warnings, cast values, report config_hash_id
* use snprintfz, add info logging
* remove unnecessary strdup and free
* check if stored config hash is null
* return if prepare statement fails
* replace with static variables
* remove replace info, free edit_command
* remove setting cfg entries to NULL
* change uuid_copy
* check return of uuid_parse, and exit if its not valid
* also free cfg
* use address
* removed health_alarm_entry_sql2json and sql_health_alarm_log_select_all
* remove check for is_valid_alarm_id
* replace lengths with GUID_LEN
* use uuid_unparse_lower_fix
* removed web api endopoint to get alert config
* check for non null values for name, chart and family
* include a date_updated field in alert_hash
* for config hash, digest NULL string if value to digest is null
* Use empty string instead of null
Diffstat (limited to 'health/health.c')
| -rw-r--r-- | health/health.c | 9 |
1 files changed, 7 insertions, 2 deletions
diff --git a/health/health.c b/health/health.c index ccd2603176..5782e54f0f 100644 --- a/health/health.c +++ b/health/health.c @@ -662,6 +662,8 @@ void *health_main(void *ptr) { int min_run_every = (int)config_get_number(CONFIG_SECTION_HEALTH, "run at least every seconds", 10); if(min_run_every < 1) min_run_every = 1; + int cleanup_sql_every_loop = 7200 / min_run_every; + time_t now = now_realtime_sec(); time_t hibernation_delay = config_get_number(CONFIG_SECTION_HEALTH, "postpone alarms during hibernation for seconds", 60); @@ -716,6 +718,9 @@ void *health_main(void *ptr) { host->health_delay_up_to = 0; } + if(likely(!host->health_log_fp) && (loop == 1 || loop % cleanup_sql_every_loop == 0)) + sql_health_alarm_log_cleanup(host); + rrdhost_rdlock(host); // the first loop is to lookup values from the db @@ -956,7 +961,7 @@ void *health_main(void *ptr) { if(likely(!rrdcalc_isrepeating(rc))) { ALARM_ENTRY *ae = health_create_alarm_entry( - host, rc->id, rc->next_event_id++, now, rc->name, rc->rrdset->id, + host, rc->id, rc->next_event_id++, rc->config_hash_id, now, rc->name, rc->rrdset->id, rc->rrdset->family, rc->classification, rc->component, rc->type, rc->exec, rc->recipient, now - rc->last_status_change, rc->old_value, rc->value, rc->status, status, rc->source, rc->units, rc->info, rc->delay_last, @@ -1006,7 +1011,7 @@ void *health_main(void *ptr) { if(unlikely(repeat_every > 0 && (rc->last_repeat + repeat_every) <= now)) { rc->last_repeat = now; ALARM_ENTRY *ae = health_create_alarm_entry( - host, rc->id, rc->next_event_id++, now, rc->name, rc->rrdset->id, + host, rc->id, rc->next_event_id++, rc->config_hash_id, now, rc->name, rc->rrdset->id, rc->rrdset->family, rc->classification, rc->component, rc->type, rc->exec, rc->recipient, now - rc->last_status_change, rc->old_value, rc->value, rc->old_status, rc->status, rc->source, rc->units, rc->info, rc->delay_last, |