ansible/roles/prometheus/templates/postgres.rules.yml.j2
2021-12-31 14:44:50 +01:00

219 lines
6.9 KiB
Django/Jinja

---
{{ ansible_managed | comment }}
{% macro raw(string) -%}
{{ "{{" }} {{ string }} {{ "}}" }}
{%- endmacro %}
groups:
- name: postgres.rules
rules:
- alert: PostgresqlDown
expr: pg_up == 0
for: 0m
labels:
severity: critical
annotations:
summary: Serveur PostgreSQL down
- alert: PostgresqlRestarted
expr: time() - pg_postmaster_start_time_seconds < 60
for: 0m
labels:
severity: critical
annotations:
summary: Serveur PostgreSQL redémarré
- alert: PostgresqlExporterError
expr: pg_exporter_last_scrape_error > 0
for: 0m
labels:
severity: critical
annotations:
summary: Erreur dans l'exporter PostgreSQL
- alert: PostgresqlReplicationLag
expr:
pg_replication_lag > 30
and
ON(instance) pg_replication_is_replica == 1
for: 0m
labels:
severity: critical
annotations:
summary: >-
La réplication PostgreSQL lag ({{ raw('$value') }} > 30s)
(base de données {{ raw('$labels.datname') }} )
- alert: PostgresqlTableNotVaccumed
expr:
time() - pg_stat_user_tables_last_autovacuum
> 60 * 60 * 24
for: 0m
labels:
severity: warning
annotations:
summary: >-
Le démon autovacuum n'a pas été lancé depuis 24h
(base de données {{ raw('$labels.datname') }} )
- alert: PostgresqlTableNotAnalyzed
expr:
time() - pg_stat_user_tables_last_autoanalyze
> 60 * 60 * 24
for: 0m
labels:
severity: warning
annotations:
summary: >-
Table non-analysée depuis 24h
(base de données {{ raw('$labels.datname') }})
- alert: PostgresqlTooManyConnections
expr:
(
sum by (datname)
(pg_stat_activity_count{datname!~"template.*|postgres"})
) * 100
> pg_settings_max_connections * 80
for: 2m
labels:
severity: warning
annotations:
summary: >-
PostgreSQL a trop de connexions
({{ raw('$value | printf "%.1f"') }} > 80%)
(base de données {{ raw('$labels.datname') }})
- alert: PostgresqlDeadLocks
expr: increase(pg_stat_database_deadlocks{datname!~"template.*|postgres"}[1m]) > 5
for: 0m
labels:
severity: warning
annotations:
summary: >-
PostgreSQL a plus de 5 deadlocks.
(base de données {{ raw('$labels.datname') }} )
- alert: PostgresqlSlowQueries
expr: pg_slow_queries > 0
for: 2m
labels:
severity: warning
annotations:
summary: >-
Présence de requêtes lentes (slow-queries)
(base de données {{ raw('$labels.datname') }} )
- alert: PostgresqlHighRollbackRate
expr:
(
rate(pg_stat_database_xact_rollback{datname!~"template.*"}[3m]) /
rate(pg_stat_database_xact_commit{datname!~"template.*"}[3m])
) * 100
> 20
for: 0m
labels:
severity: warning
annotations:
summary: >-
PostgreSQL a un taux de retour en arrière (rollback) élevé
(base de données {{ raw('$labels.datname') }}, valeur {{ raw('$value | printf "%.1f"') }} %)
- alert: PostgresqlWaleReplicationStopped
expr: rate(pg_xlog_position_bytes[1m]) == 0
for: 0m
labels:
severity: critical
annotations:
summary: >-
Réplication de PostgreSQL WALE stoppée
(base de données {{ raw('$labels.datname') }} )
- alert: PostgresqlHighRateStatementTimeout
expr: rate(postgresql_errors_total{type="statement_timeout"}[1m]) > 3
for: 0m
labels:
severity: critical
annotations:
summary: >-
Beaucoup de requêtes PostgreSQL sont timeout
(base de données {{ raw('$labels.datname') }}, valeur {{ raw('$value | printf "%.1f"') }} )
- alert: PostgresqlHighRateDeadlock
expr: increase(postgresql_errors_total{type="deadlock_detected"}[1m]) > 1
for: 0m
labels:
severity: critical
annotations:
summary: >-
PostgreSQL a un fort taux de deadlock
(base de données {{ raw('$labels.datname') }}, valeur {{ raw('$value | printf "%.1f"') }} )
# - alert: PostgresqlReplicationLagBytes
# expr:
# (pg_xlog_position_bytes and pg_replication_is_replica == 0)
# - GROUP_RIGHT(instance) (pg_xlog_position_bytes and pg_replication_is_replica == 1)
# > 1e+09
# for: 0m
# labels:
# severity: critical
# annotations:
# summary: La réplication Postgresql a des octets de retard (instance {{ raw('$labels.name') }}, value {{ raw('$value') }} )
- alert: PostgresqlTooManyDeadTuples
expr:
(
(pg_stat_user_tables_n_dead_tup > 10000)
/ (pg_stat_user_tables_n_live_tup + pg_stat_user_tables_n_dead_tup)
) >= 0.1 unless ON(instance) (pg_replication_is_replica == 1)
for: 2m
labels:
severity: warning
annotations:
summary: >-
Les tuples morts PostgreSQL sont trop volumineux
(base de données {{ raw('$labels.datname') }}, valeur {{ raw('$value | printf "%.1f"') }} )
- alert: PostgresqlSplitBrain
expr: count(pg_replication_is_replica == 0) != 1
for: 0m
labels:
severity: critical
annotations:
summary: >-
Split Brain : trop de bases de données PostgreSQL primaires en mode lecture-écriture
(base de données {{ raw('$labels.datname') }}, valeur {{ raw('$value') }} )
- alert: PostgresqlPromotedNode
expr:
pg_replication_is_replica
and
changes(pg_replication_is_replica[1m]) > 0
for: 0m
labels:
severity: warning
annotations:
summary: >-
Le serveur de secours PostgreSQL a été promu comme nœud principal
(base de données {{ raw('$labels.datname') }}, valeur {{ raw('$value') }})
- alert: PostgresqlTooManyLocksAcquired
expr:
(
(sum (pg_locks_count))
/ (pg_settings_max_locks_per_transaction * pg_settings_max_connections)
) * 100 > 20
for: 2m
labels:
severity: critical
annotations:
summary: >-
Trop de deadlocks acquis sur la base de données.
Si cette alerte se produit fréquemment, nous devrons peut-être augmenter
le paramètre postgres max_locks_per_transaction
(Valeur = {{ raw('$value | printf "%.1f"') }} )
...