RDS ACUUtilization suddenly spiked. General query log shows all internal RDS queries, not our application.
Does anyone know what changed mid-August with RDS serverless v2? We had very low usage and then suddenly very high. The server was getting slammed with all these queries that are from RDS. Like I said, this database server doesn't get a lot of usage. It's entirely internal queries from Amazon. I turned off Enhanced Monitoring and Performance Insights and I'm still not getting below 50% ACUUtilization. It's bizarre.
2025-09-04T15:20:26.433693Z 12 Query set local oscar_local_only_replica_host_status=1; 2025-09-04T15:20:26.433810Z 12 Query SELECT durable_lsn, current_read_point, server_id, last_update_timestamp FROM information_schema.replica_host_status; 2025-09-04T15:20:26.433975Z 12 Query set local oscar_local_only_replica_host_status=0 2025-09-04T15:20:26.436987Z 15 Query SET @@sql_log_bin=off ; 2025-09-04T15:20:26.437091Z 15 Query INSERT INTO mysql.rds_heartbeat2(id, value) values (1,1756999226436) ON DUPLICATE KEY UPDATE value = 1756999226436 ; 2025-09-04T15:20:26.441757Z 15 Query SET @@sql_log_bin=on 2025-09-04T15:20:26.443213Z 15 Query COMMIT 2025-09-04T15:20:26.445964Z 10 Query SELECT /*+ MAX_EXECUTION_TIME(5000) */ @@global.innodb_read_only, @@global.max_connections, @@global.autocommit, @@global.sql_mode,
- Newest
- Most votes
- Most comments
Refer this base on my understanding:
What changed recently: • Internal RDS health checks and replication monitoring have become more aggressive. These include: o rds_heartbeat2 inserts to monitor cluster health o replica_host_status queries for replication sync o Global variable reads for performance tuning • Passive ACU usage has increased, even during idle periods. Multiple users have reported that clusters now maintain higher baseline ACU levels, even when no user queries are running. • RDS Proxy and connection pooling may be contributing. If you're using RDS Proxy, idle connections can keep the cluster “awake,” preventing scale-down and triggering internal activity.
What you can consider:
- Cap the minimum ACU Set a lower bound (e.g., 0.5 ACU) to force scale-down during idle periods.
- Disable RDS Proxy Proxy mismanagement of idle connections has been linked to inflated ACU usage.
- Monitor internal queries with general logs Keep tracking patterns to see if they correlate with maintenance windows or replication syncs.
Sudden ACU spikes often come from a change in workload pattern (query plan change, new connection pool behavior, or background tasks). A few steps you can try:
Enable Performance Insights and check “load by waits” to see if one query started consuming more ACUs.
Review connection pooling - too many short-lived connections can trigger scaling.
Compare query logs before/after the spike to identify new or slower queries.
If Aurora version/parameters changed, verify flags like max_parallel_workers_per_gather.
Meanwhile, set sensible min/max capacity to avoid runaway costs while investigating.
