Thank you very much Dmitry, I understand now and it makes sense, very logic.

Greetings and thanks again.

Walter.

On Sun, Jan 6, 2013 at 7:02 PM, Dmitry Kuzmenko <kdv@...> wrote:

> **
>
>
> Hello, W!
>
> Monday, January 7, 2013, 2:33:30 AM, you wrote:
>
> WO> Hello everybody, I have a question:
>
> WO> - Transaction 510 inserts a row
> WO> - Transaction 510 commit
>
> ok, data committed.
>
> WO> - Transaction 526 start
> WO> - Transaction 526 updates the row inserted by transaction 510
>
> well, row "locked" by update made by transaction 526.
>
> WO> - Transaction 535 start
> WO> - Transaction 535 updates the row inserted by transaction 510
>
> no, transaction 535 will get error "deadlock", because
> version created by transaction 526 is not committed or rolled back
> yet.
>
> WO> - Transaction 542 start
> WO> - Transaction 542 updates the row inserted by transaction 510
>
> NO. Same reason that I noted before.
>
> WO> 1. Should be the order for the following commits: 526, 535, 542?
>
> NO.
>
> WO> 2. Or in some cases it could be 542, 535, 526?
>
> NO.
>
> WO> If the answer is 1. it means that if transaction 526 hangs up will
> avoid
> WO> transactions 535 and 542 to finish. Am I right?
>
> No transaction can overwrite version, created by any transaction that
> modified some record and has not committed or rolled back changes.
>
> --
> Dmitry Kuzmenko, www.ib-aid.com
>
>
>


[Non-text portions of this message have been removed]