A centralized global deadlock detector holds the union of local wait-for graphs. Give an example to explain how a phantom deadlock could be detected if a waiting transaction in a deadlock cycle aborts during the deadlock detection procedure.

What will be an ideal response?

A centralized global deadlock detector holds the union of local wait-for graphs. Give an example to explain how a phantom deadlock could be detected if a waiting transaction in a deadlock cycle aborts during the deadlock detection procedure.

Suppose that at servers X, Y and Z we have:



when U aborts, Y knows first, then X finds out, eventually the global detector finds out, but by then it may be too late (it will have detected a deadlock).