Notes for the problems that show up after launch: bad plans, awkward migrations, index debt, vacuum pressure, replica lag, and the small decisions that keep PostgreSQL, MySQL, MariaDB, and SQL Server easier to operate.
Find the queries dragging your Postgres down, read their plans, and fix them in order of impact.
Read the guide →Which indexes to add, which to drop, and how to spot redundant index debt before it costs you writes.
Read the guide →How autovacuum works, when it falls behind, and how to keep bloat from eating your storage and your buffer cache.
Read the guide →Sizing pools, PgBouncer modes, and how to stop idle-in-transaction sessions from exhausting max_connections.
Read the guide →Track lag, WAL throughput, and replication slots before they turn into stale reads or a full disk.
Read the guide →
Backups, restores, PITR, archive gaps, and recovery drills for PostgreSQL.
AWS RDS, Azure Database for PostgreSQL, Google Cloud SQL, AlloyDB, and managed PostgreSQL monitoring notes.
Connection storms, pool saturation, PgBouncer behavior, and max_connections pressure.
CPU, memory, disk, temp files, checkpoints, buffers, and workload pressure.
Index design, index debt, reindexing, constraint indexes, and evidence-driven DDL.
Lock waits, deadlocks, blocking chains, idle transactions, and transaction hygiene.
Galera clustering, InnoDB and Aria internals, replication, optimizer behavior, and MariaDB operations notes.
InnoDB internals, replication, locking, query performance, and MySQL operations notes.
HNSW, IVFFlat, recall, embedding search, and production RAG performance notes.
Replica lag, WAL growth, failover readiness, hot standby behavior, and replication slots.
Query plans, EXPLAIN ANALYZE, planner regressions, pagination, joins, and statistics.
Wait statistics, tempdb, Query Store, Always On availability groups, and SQL Server operations notes.
Autovacuum, dead tuples, bloat, wraparound, MVCC, freeze age, and visibility.
Our 8.0 config had innodb_flush_neighbors=1, inherited verbatim from a 5.7 my.cnf written for spinning disks in 2015. On NVMe it was pure write amplification: 18% of checkpoint flush IOPS were pages nobody asked to write. Setting it to 0 smoothed…
The analytics service stored 64-bit hashes in a BIGINT UNSIGNED column, and the PostgreSQL port mapped it to bigint — which is why roughly half the hash space exploded on insert with bigint out of range. Field notes on what…
The failover took ninety seconds. Bringing the old primary back as a standby was the expensive part — rsync estimated four hours for 1.1 TB. pg_rewind walked the WAL history to the divergence point and copied only the blocks that…
ERROR: invalid page header in block 488291 of relation orders_pkey at 03:12 on a Saturday. Before touching REINDEX or reaching for a backup, pageinspect let us read the damaged page itself, prove the heap was clean, and turn a weekend-long…
A CREATE INDEX CONCURRENTLY killed by a lock timeout left behind an INVALID index: invisible to the planner, invisible in \d, and still eating write I/O on every insert. We found eleven of them during a storage audit — 40…
Every morning at 06:05 the first report query failed with 'MySQL server has gone away' and succeeded on retry. The pool held connections idle past the default 8-hour wait_timeout, and the server had closed them silently overnight. The fix was…
After a switch flap, the entire application fleet got 'Host is blocked because of many connection errors' — all at once, because every app server sits behind one NAT address. TRUNCATE performance_schema.host_cache unblocked us; understanding the error counter kept it…
The deploy to a new subnet failed with Access denied for user 'app'@'10.0.14.7' even though the GRANT looked correct — because MySQL identity includes the client host and the most specific match wins. The PostgreSQL port failed differently: a pg_hba.conf…
The 5.7-to-8.0 upgrade went clean — until Monday, when a PHP billing worker started throwing error 2059 and a Perl cron from 2016 answered with 1251. The default authentication plugin changed, and nobody had read that line of the release…
Our cross-region replica link peaked at 310 Mbps during batch windows and lag followed like a tide. Turning on binlog_transaction_compression cut the stream to 115 Mbps for a 3% CPU cost — and then Debezium stopped reading us entirely, which…