diff options
Diffstat (limited to '')
| -rw-r--r-- | health/health.d/mysql.conf | 146 |
1 files changed, 146 insertions, 0 deletions
diff --git a/health/health.d/mysql.conf b/health/health.d/mysql.conf new file mode 100644 index 0000000..62cef5a --- /dev/null +++ b/health/health.d/mysql.conf @@ -0,0 +1,146 @@ + +# make sure mysql is running + +template: mysql_last_collected_secs + on: mysql.queries + calc: $now - $last_collected_t + units: seconds ago + every: 10s + warn: $this > (($status >= $WARNING) ? ($update_every) : ( 5 * $update_every)) + crit: $this > (($status == $CRITICAL) ? ($update_every) : (60 * $update_every)) + delay: down 5m multiplier 1.5 max 1h + info: number of seconds since the last successful data collection + to: dba + + +# ----------------------------------------------------------------------------- +# slow queries + +template: mysql_10s_slow_queries + on: mysql.queries + lookup: sum -10s of slow_queries + units: slow queries + every: 10s + warn: $this > (($status >= $WARNING) ? (5) : (10)) + crit: $this > (($status == $CRITICAL) ? (10) : (20)) + delay: down 5m multiplier 1.5 max 1h + info: number of mysql slow queries over the last 10 seconds + to: dba + + +# ----------------------------------------------------------------------------- +# lock waits + +template: mysql_10s_table_locks_immediate + on: mysql.table_locks + lookup: sum -10s absolute of immediate + units: immediate locks + every: 10s + info: number of table immediate locks over the last 10 seconds + to: dba + +template: mysql_10s_table_locks_waited + on: mysql.table_locks + lookup: sum -10s absolute of waited + units: waited locks + every: 10s + info: number of table waited locks over the last 10 seconds + to: dba + +template: mysql_10s_waited_locks_ratio + on: mysql.table_locks + calc: ( ($mysql_10s_table_locks_waited + $mysql_10s_table_locks_immediate) > 0 ) ? (($mysql_10s_table_locks_waited * 100) / ($mysql_10s_table_locks_waited + $mysql_10s_table_locks_immediate)) : 0 + units: % + every: 10s + warn: $this > (($status >= $WARNING) ? (10) : (25)) + crit: $this > (($status == $CRITICAL) ? (25) : (50)) + delay: down 30m multiplier 1.5 max 1h + info: the ratio of mysql waited table locks, for the last 10 seconds + to: dba + + +# ----------------------------------------------------------------------------- +# connections + +template: mysql_connections + on: mysql.connections_active + calc: $active * 100 / $limit + units: % + every: 10s + warn: $this > (($status >= $WARNING) ? (60) : (70)) + crit: $this > (($status == $CRITICAL) ? (80) : (90)) + delay: down 15m multiplier 1.5 max 1h + info: the ratio of current active connections vs the maximum possible number of connections + to: dba + + +# ----------------------------------------------------------------------------- +# replication + +template: mysql_replication + on: mysql.slave_status + calc: ($sql_running <= 0 OR $io_running <= 0)?0:1 + units: ok/failed + every: 10s + crit: $this == 0 + delay: down 5m multiplier 1.5 max 1h + info: checks if mysql replication has stopped + to: dba + +template: mysql_replication_lag + on: mysql.slave_behind + calc: $seconds + units: seconds + every: 10s + warn: $this > (($status >= $WARNING) ? (5) : (10)) + crit: $this > (($status == $CRITICAL) ? (10) : (30)) + delay: down 15m multiplier 1.5 max 1h + info: the number of seconds mysql replication is behind this master + to: dba + + +# ----------------------------------------------------------------------------- +# galera cluster size + +template: mysql_galera_cluster_size_max_2m + on: mysql.galera_cluster_size + lookup: max -2m absolute + units: nodes + every: 10s + info: max cluster size 2 minute + to: dba + +template: mysql_galera_cluster_size + on: mysql.galera_cluster_size + calc: $nodes + units: nodes + every: 10s + warn: $this > $mysql_galera_cluster_size_max_2m + crit: $this < $mysql_galera_cluster_size_max_2m + delay: up 20s down 5m multiplier 1.5 max 1h + info: cluster size has changed + to: dba + +# galera node state + +template: mysql_galera_cluster_state + on: mysql.galera_cluster_state + calc: $state + every: 10s + warn: $this < 4 + crit: $this < 2 + delay: up 30s down 5m multiplier 1.5 max 1h + info: node state (0: undefined, 1: joining, 2: donor/desynced, 3: joined, 4: synced) + to: dba + + +# galera node status + +template: mysql_galera_cluster_status + on: mysql.galera_cluster_status + calc: $wsrep_cluster_status + every: 10s + crit: $mysql_galera_cluster_state != nan AND $this != 0 + delay: up 30s down 5m multiplier 1.5 max 1h + info: node and cluster status (-1: unknown, 0: primary/quorum present, 1: non-primary/quorum lost, 2: disconnected) + to: dba |