MySQL/MariaDB — InnoDB, indexing, replication, JSON, query tuning.
Scanned 9/5/2026
Install to Claude Code
npx -y skills add Dev-Toolbelt/dev-team-agents --skill mysql --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Mysql?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/dev-toolbelt-mysql)More formats (shields.io, HTML) on the badges page.
---
name: mysql
description: MySQL/MariaDB — InnoDB, indexing, replication, JSON, query tuning.
---
# MySQL Reference
## When to Load
Load when the project uses MySQL or MariaDB (signal: `mysql`, `mysql2`, `mysql:` in docker-compose, `DATABASE_URL` with `mysql://`).
## CLI Connection
```bash
mysql -h "$DB_HOST" -P "${DB_PORT:-3306}" -u "$DB_USER" -p"$DB_PASSWORD" "$DB_NAME"
# non-interactive:
mysql -h "$DB_HOST" -u "$DB_USER" -p"$DB_PASSWORD" "$DB_NAME" -e "SELECT VERSION();"
```
Key inspection commands:
```sql
SHOW ENGINE INNODB STATUS; -- lock waits, undo log
SHOW FULL PROCESSLIST; -- running queries
EXPLAIN FORMAT=JSON SELECT ...; -- detailed query plan
SELECT * FROM performance_schema.events_statements_summary_by_digest
ORDER BY sum_timer_wait DESC LIMIT 10; -- slow query digest
```
## InnoDB Storage Engine
- Default engine since MySQL 5.5; always use InnoDB (not MyISAM) for new tables
- Row-level locking; supports foreign keys and transactions
- Clustered primary key index: all secondary indexes contain the PK value
- Large PK → large secondary indexes; prefer `INT AUTO_INCREMENT` or compact UUID
## Index Behavior
| Scenario | Recommendation |
|----------|---------------|
| Composite index | Most selective column first; match query column order |
| Covering index | Include all queried columns: `CREATE INDEX idx ON t(a, b) INCLUDE (c)` |
| Prefix index | For long VARCHAR: `CREATE INDEX ON t(col(20))` |
| `EXPLAIN` | Always run before merging slow queries; check `type` column (prefer `ref` or `range`, avoid `ALL`) |
## Transactions & Isolation
Default isolation: `REPEATABLE READ` (unlike PostgreSQL's `READ COMMITTED`).
- MVCC via undo logs (not tuple versioning like Postgres)
- Long transactions grow the undo log → monitor with `SHOW ENGINE INNODB STATUS`
- Use `READ COMMITTED` for OLAP or reporting queries to reduce lock contention
## Replication
| Type | Consistency | Use when |
|------|-------------|----------|
| Async (default) | Eventual | Most setups; risk of data loss on primary failure |
| Semi-sync | Near-sync | One replica ACKs before commit returns |
| GTID-based | Position-independent | Failover without manual binlog coordinates |
- Never write to a replica; use `read_only = ON` on replicas
- Replica lag: monitor `Seconds_Behind_Master` or `performance_schema.replication_connection_status`
## JSON Type (MySQL 5.7.8+)
- Store as `JSON` column for validation; queried with `->` (returns JSON) and `->>` (returns string)
- Index generated columns: `ALTER TABLE t ADD COLUMN col VARCHAR(255) GENERATED ALWAYS AS (json_col->>'$.key') VIRTUAL, ADD INDEX idx(col)`
- Avoid deep nesting; normalize frequently queried fields
## Common Gotchas
| Gotcha | Fix |
|--------|-----|
| Silent truncation in strict mode OFF | Enable `sql_mode = STRICT_ALL_TABLES` |
| `GROUP BY` without `ONLY_FULL_GROUP_BY` silently picks random non-aggregated values | Enable `ONLY_FULL_GROUP_BY` |
| `utf8` charset is actually 3-byte (no emoji) | Use `utf8mb4` + `utf8mb4_unicode_ci` |
| `TINYINT(1)` returned as boolean in some ORMs | Use `TINYINT(1)` deliberately for booleans |
| Deadlocks on concurrent `INSERT ... ON DUPLICATE KEY UPDATE` | Use sparingly; prefer application-level locking |
## Useful Diagnostics
```sql
SHOW ENGINE INNODB STATUS; -- lock waits, undo log
SHOW FULL PROCESSLIST; -- running queries
SELECT * FROM performance_schema.events_statements_summary_by_digest
ORDER BY sum_timer_wait DESC LIMIT 10; -- slow queries
EXPLAIN FORMAT=JSON SELECT ...; -- detailed plan
```
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!