Skip to main content
Nerve Centre KPIs · Audit Profile · Sentiment Settings MySQL-specific health audit for production 5.7 / 8.0 instances. Answers six questions: (1) is access correctly scoped and is performance_schema actually readable; (2) is the instance reachable with a healthy connection pool and low aborted-connect noise; (3) is query latency within band and is the slow-query digest list under control; (4) are replicas streaming with both threads running and Seconds_Behind_Source inside threshold; (5) is disk and buffer-pool capacity healthy; (6) is a recent durable backup in place. Cross-channel area joins query and pool pressure to commerce-sibling checkout traffic to size live revenue at risk.

​
What this audit checks

​
Authentication & access

  • Connection succeeds over the configured connection_string with SSL mode honoured on managed endpoints (RDS / Aurora / PlanetScale / Cloud SQL)
  • Audit user holds SELECT on performance_schema plus REPLICATION CLIENT and PROCESS (else replication and digest checks degrade to skipped)
  • performance_schema is enabled (SHOW VARIABLES LIKE ‘performance_schema’) so events_statements_summary_by_digest returns rows
  • sys schema installed for friendlier digest labels; warn when absent

​
Connection & availability

  • Instance reachable and SHOW GLOBAL STATUS returns within timeout; Uptime > 0 confirms no recent unplanned restart
  • Connection pool saturation (Threads_connected / max_connections) below threshold with headroom for traffic bursts
  • Aborted_connects over 24h below threshold (spikes signal wrong creds, network drops, or max_connect_errors hit)
  • Threads_running not pinned high relative to Threads_connected (sign of query pile-up / stalls)

​
Query performance

  • Query latency p95 (derived from events_statements_summary_by_digest AVG_TIMER_WAIT) within threshold
  • Slow-query rate (Slow_queries delta / Questions delta from SHOW GLOBAL STATUS) below threshold over a 15m window
  • Top-10 slow digests reviewed; no new digest entering the list with high rows_examined / rows_returned ratio
  • InnoDB deadlocks (SHOW ENGINE INNODB STATUS) at zero over the last 5m

​
Replication & lag

  • Both Replica_IO_Running and Replica_SQL_Running are Yes on every replica (SHOW REPLICA STATUS / SHOW SLAVE STATUS on < 8.0.22)
  • Seconds_Behind_Source within lag threshold and not null on any active replica
  • Active replica count matches expected topology; no replica in BROKEN or STOPPED state
  • Binlog backlog ahead of the slowest replica position below threshold (growing backlog = replica falling behind)

​
Storage & capacity

  • Database disk usage percent below threshold with runway before the volume fills
  • InnoDB buffer pool hit rate (1 - Innodb_buffer_pool_reads / Innodb_buffer_pool_read_requests) above threshold
  • InnoDB dirty pages percent below threshold to avoid checkpoint pressure and write storms
  • InnoDB free pages above the floor (> 1% of total pages) so the pool is not starved

​
Backups & durability

  • Last successful backup age (mysqldump / Percona XtraBackup / RDS snapshot) within freshness threshold
  • binlog_expire_logs_seconds / expire_logs_days set so point-in-time recovery has a usable window
  • Binary logging enabled (log_bin = ON) where PITR or replication is required
  • innodb_flush_log_at_trx_commit = 1 and sync_binlog = 1 for durable-by-default writes (warn on relaxed settings)

​
Cross-channel: revenue protection

  • QPS spike with no matching order spike (sibling = bigcommerce/shopify/adobe orders_per_15m flat while mysql.qps surges = bot / scraper load)
  • Pool saturation during traffic burst (sibling = commerce checkout volume rising while Threads_connected / max_connections crosses threshold = lost orders)
  • Slow queries co-occurring with checkout drop (mysql.slow_query in same 5m window as sibling.checkout_step_completion_rate falling > 5pp)
  • Inventory-table row drift vs ecom inventory count (mysql merchant-owned schema row count vs sibling.product_inventory by sku = oversell risk)

​
Severity thresholds

SignalWarnCritical
connection_error_rate0.51
query_p95_ms200500
replication_lag_sec1030
disk_usage_pct8090
slow_query_count520

​
Data sources

  • GET mysql://{host}:{port}/{database} :: SHOW VARIABLES - Instance config: version, max_connections, performance_schema, buffer-pool size, binlog settings
  • GET mysql://{host}:{port}/{database} :: SHOW GLOBAL STATUS - Live counters: Questions, Slow_queries, Threads_connected/running, Aborted_connects, Uptime, InnoDB buffer-pool reads
  • GET mysql://{host}:{port}/{database} :: performance_schema.events_statements_summary_by_digest - Per-digest latency (p50/p95/p99), rows_examined, rows_returned for slow-query analysis
  • GET mysql://{host}:{port}/{database} :: performance_schema.processlist - Live connection / thread inventory and pool occupancy by app_name
  • GET mysql://{host}:{port}/{database} :: SHOW REPLICA STATUS - Replica thread state and Seconds_Behind_Source (SHOW SLAVE STATUS on < 8.0.22)
  • GET mysql://{host}:{port}/{database} :: SHOW ENGINE INNODB STATUS - InnoDB deadlocks, buffer-pool dirty/free pages, checkpoint pressure