Notes for the problems that show up after launch: bad plans, awkward migrations, index debt, vacuum pressure, replica lag, and the small decisions that make PostgreSQL easier to operate.
MariaDB ships a real thread pool in the community server — MySQL charges extra for one. Here is how pool-of-threads works, when it saves you under a connection storm, and when it quietly serializes you.
A decommissioned standby left its replication slot behind, and over one quiet weekend the slot pinned 214 GB of WAL until the primary ran out of disk and PANIC'd. Here is the mechanism, the monitoring queries, and the circuit breaker that caps the damage.
An UPDATE that touches no indexed column can skip index maintenance entirely, if the page has room. How HOT updates work, why fillfactor matters, and how to measure the ratio in pg_stat_user_tables.
MariaDB's system-versioned tables keep row history automatically — no triggers, no shadow tables. How FOR SYSTEM_TIME works, what storage growth really costs, and the limitations worth knowing first.
Group Replication promises consensus-backed failover; classic async promises simplicity. Field notes on what GR actually guarantees, how flow control stalls write bursts, and when async plus an orchestrator is still the better architecture.
COUNT(*) is slow on big tables in both InnoDB and PostgreSQL, for related but different MVCC reasons. Here are the honest options: index-only counting, estimates, and counter tables.
Someone forgot the WHERE clause. You can restore last night's backup and replay a day of binlog, or you can generate the inverse transaction and apply it in minutes. Here is how MariaDB flashback actually works.
The query cache lives on in MariaDB long after MySQL deleted it, and on a busy many-core server it is often the top contention point. Here is how to prove it with counters and what to run instead.
The planner picked a nested loop that ran 400,000 times because n_distinct was off by a factor of 30. Raising one column's statistics target fixed the estimate — and the plan — in a single ANALYZE.
The dashboard said SYNCHRONIZED while the readable secondary served twenty-minute-old data. Send queue and redo queue are different numbers; here is what each one means, what synchronous commit really guarantees, and how to measure RPO.
The app said the database was down; the database said CPU was twelve percent and forty-three sessions were queued behind one sleeping connection holding a lock. How I walk a blocking chain to its head and decide whether to kill or wait.
Semi-synchronous replication promises that a commit is acknowledged by at least one replica before the client sees success. The fine print — silent fallback to async, wait points, and what an ack actually means — is where production incidents live.