Your range query lists five orders, then six on its rerun, because a new order arrived. Which anomaly is this?

• A phantom read
• A non-repeatable read
• A deadlock

Your transaction reads a stock count of 12, another transaction commits 9, and your next read gets 9. Which anomaly is this?

• A dirty read of an uncommitted 9
• A lost update
• A non-repeatable read

A dirty read sees data that was never committed.

Two transactions interleave to produce a phantom read. Which level would have prevented it?

• Read committed, which is enough for new rows
• Read uncommitted, the fastest choice
• Serializable, which forbids phantoms

Which level is the lowest one that forbids dirty reads?

• Read uncommitted
• Read committed
• Repeatable read

Which workload fits a weaker isolation level, and why?

• A rough analytics count, as a stale total harms nobody
• A bank balance, because money is simple
• A seat booking, because the last seat rarely sells twice

Pat claims that, by the SQL standard's definitions, repeatable read also forbids phantoms, since both involve rereading. What is wrong?

• Nothing, the two anomalies are the same
• Repeatable read forbids dirty reads only
• Phantoms are new rows, not changed rows, and need serializable

A booking system runs at read committed. Each booking checks that no booking row exists for the seat, then inserts one. Sometimes the last seat sells twice. Which level fixes it, and what does it forbid that the current one permits?

• Read uncommitted, which is faster and looser
• Serializable, which forbids the phantom that sneaks the second booking in
• No level can help, since seats are not rows