This course is for people who already know what Apache Cassandra is and now have to operate it. It is not an install guide, not a CQL primer, and not a local cluster lab. You will learn how an operator reads a cluster. Each module ties a symptom to a layer: token ownership and replication, commitlog, memtable, SSTable, compaction, tombstones, hinted handoff, and read repair. The goal is a decision, not a tour of the architecture. Monitoring is treated as an operator tool. You will see which signals matter for latency, pending work, dropped messages, disk, compaction backlog, and tombstones; how labels and a scrape job change what an alert can say; and how to write alert rules that name the next action. Prometheus and Grafana appear as the place those signals land. Demonstrations use diagrams, configuration excerpts, screenshots, and simulated series, not a cluster you must run.
Repair, compaction, and tombstones get their own decision path. You will compare full and incremental repair, see what an overdue repair looks like, and learn why a growing tombstone count is a read risk rather than a cosmetic metric. Backup and restore stay at the concept and checklist level: snapshots, Medusa full and incremental backups, and the case where the backup job succeeded and the restore did not. Seven incidents close the course. Disk grows without a matching write rate. A node is down. Repair is overdue. Tombstones rise. Latency moves after a workload change. Prometheus stops scraping the exporter. A backup completes and the restore fails. For each one you leave with a line in a checklist or a short report.
