Midnight Crash: The ‘Impossible’ MyBatis‑Plus Pitfalls That Can Break Production

Background and ORM Choice

Rapid data growth and micro‑service fragmentation have pushed Java teams to seek an ORM that balances speed and stability. MyBatis‑Plus emerged as a production‑oriented enhancement over raw MyBatis, offering CRUD shortcuts, logical deletion, auto‑fill, optimistic locking, multi‑tenant support, and code generation.

Production Challenges

In development environments the framework appears flawless, but once deployed the hidden defaults surface, leading to critical failures such as primary‑key conflicts and data inconsistency.

1. Snowflake ID Duplication

Problem: Two container instances generated identical Snowflake IDs, causing primary‑key collisions at 3 AM.

Root Causes:

Machine ID not configured, all instances use the default.

Docker containers clone MAC addresses; the framework extracts the last two bytes and mods by 32, causing collisions when instances exceed 32.

Kubernetes pods receive virtual MACs from CNI plugins, often identical or unavailable, leading to the same workerId.

Clock rollback duplicates timestamps, dataCenterId, workerId, and sequence.

Solution: Provide a unique distributed ID service or switch to database‑generated IDs.

@Slf4j
@Component
public class IdHelper {
    public static long getIdc() {
        try {
            return ZZIdcClient.getInstance().get();
        } catch (Throwable e) {
            log.error("act=getIdc error={}", e.getMessage());
            log.info("k=s act=getIdcByIDHelper");
            return IDHelper.generator(0);
        }
    }
}

Alternative enum for IdType shows built‑in options:

public enum IdType {
    AUTO(0), NONE(1), INPUT(2), ASSIGN_ID(3), ASSIGN_UUID(4),
    @Deprecated ID_WORKER(3), @Deprecated ID_WORKER_STR(3), UUID(4);
    private final int key;
}