RabbitMQ does not crash, it blocks publishers. Queue depth, consumer counts, memory and disk alarms, cluster partitions — what to watch and what to alert on.
Renewal that succeeded without a reload, an intermediate expiring before the leaf, certificates on ports no browser visits — what to monitor beyond expiry.
Connection saturation, idle-in-transaction, autovacuum debt and transaction ID wraparound — the PostgreSQL metrics that matter, and how to collect them.
Accepted versus handled connections is the one number that says nginx is dropping traffic. What stub_status gives you, what it cannot, and what to alert on.
Which Redis metrics predict an incident — memory, eviction, hit ratio, persistence, replication — what a bad value looks like, and how to collect them.
