The JusDB Database Engineering Blog

Perform zero-downtime MySQL schema changes with pt-online-schema-change. Covers how it works, throttle settings, max-load protection, and when to use gh-ost instead.

Diagnose and fix PostgreSQL query plan cache issues. Covers generic vs custom plans, plan_cache_mode setting, DEALLOCATE, statistics targets, and pg_stat_statements variance detection.

Configure the MySQL slow query log and analyze it with pt-query-digest to find top offenders by total time. Includes EXPLAIN diagnosis and performance_schema alternative.

Harden MySQL user permissions with host-scoped grants, validate_password plugin, account locking, and privilege auditing. Essential for SOC2 and PCI-DSS compliance.

Systematically diagnose MySQL replication lag: single-threaded applier, large transactions, missing replica indexes. Enable parallel replication and use pt-heartbeat.

Understand InnoDB crash recovery mechanics and handle edge cases with innodb_force_recovery. Emergency dump procedures and durability settings to prevent future crashes.

Day-to-day MongoDB replica set operations: adding/removing members, triggering elections, diagnosing replication lag, and managing oplog size.

Systematically diagnose and fix MySQL 'Too many connections' errors. Learn to identify connection leaks, tune wait_timeout, and implement ProxySQL pooling.

Configure MySQL Orchestrator for automated failover and replication topology management. Step-by-step guide covering installation, Raft mode, and failover testing.

Error 1213 fires 400 times per hour and your application retries hide it. Learn to read SHOW ENGINE INNODB STATUS deadlock output, identify lock ordering violations, and eliminate the patterns causing them.

PostgreSQL's pg_stat_activity view exposes every connection's state — but understanding what 'idle in transaction' means and why it's dangerous requires knowing the full lifecycle of a PostgreSQL backend process.

Hatchet transforms raw MongoDB JSON logs into actionable insights — surfacing slow operations, connection storms, and namespace hotspots that are impossible to spot in raw log files.