Deadlock with one table - PostgreSQL is doing it right

Поиск
Список
Период
Сортировка
От Hans Schou
Тема Deadlock with one table - PostgreSQL is doing it right
Дата
Msg-id CAApBw36hmDPXTN44Fxx=CCgTgf+-iDsBzNOTtnF-QaCfAOSncQ@mail.gmail.com
обсуждение исходный текст
Ответы Re: Deadlock with one table - PostgreSQL is doing it right  ("Rakesh Kumar" <rakeshkumar464@mail.com>)
Re: Deadlock with one table - PostgreSQL is doing it right  (Jeremy Finzel <finzelj@gmail.com>)
Список pgsql-general
Hi

FYI - if it has any interest

During my preparation for describing what happens when two processes update the same row in a table, I came across that PostgreSQL is doing right and Oracle is doing it wrong.

The situation is a process which get a deadlock, but because it is a script, it sends a commit anyway. This is bad behavior by humans but that's how they are.

After both processes commit's the table should be:
 i |  n
---+---
 1 | 11
 2 | 21
in Oracle it is:
 i |  n
---+---
 1 | 11
 2 | 22

PostgreSQL: https://youtu.be/rH-inFRMcvQ
Oracle: https://youtu.be/l2IGoaWql64

PostgreSQL:
A
select * from t;
begin;
update t set n=n+1 where i=2;

B
begin;
update t set n=n+1 where i=1;
update t set n=n+1 where i=2;

A
update t set n=n+1 where i=1;

B
commit;

A
commit;

best regards
hans

В списке pgsql-general по дате отправления:

Предыдущее
От: Olivier Macchioni
Дата:
Сообщение: Dynamic Enum?
Следующее
От: Dave Cramer
Дата:
Сообщение: Re: psycopg2 and java gssapi questions