Cassandra Compaction Backlog Hidden Latency
warningRising pending compaction tasks silently degrade read performance by forcing queries to scan more SSTables. This storage debt accumulates invisibly until read latency spikes occur.
Track cassandra.compaction.tasks.pending metric continuously. Alert when pending tasks trend upward without dropping or when count exceeds baseline by 50%. Cross-reference with grafana_database_backend_wait_time_seconds and grafana_storage_server_poller_query_time_seconds for storage layer confirmation.
When compaction backlog grows: (1) Verify disk I/O capacity isn't saturated during peak write periods, (2) Review compaction strategy configuration (SizeTiered vs. LeveledCompaction), (3) Temporarily throttle write throughput if cluster cannot keep pace, (4) Scale disk IOPS or add nodes if sustained high write load exceeds cluster capacity.