Обсуждение: moving from 6.3.2 to 6.5

Hi

I have a relatively large application developed with php and postgres
using pg 6.3.2. I read the changelog for v.6.5 and i got a little bit
scared, but maybe i understood things in the wrong way.

In this application every "race condition" is placed in a begin->end
block and most of them even have lock's before they're needed just mo
make sure. If I upgrade postgres to 6.5 is there any danger that with
the new locking mechanism my app's database starts getting corrupted or
start's giving out the wrong type of information?

Thanks
--
        Mario Filipe
        mjnf@uevora.pt
        http://neptuno.sc.uevora.pt/~mjnf

Mario Jorge Nunes Filipe wrote:
>
> Hi
>
> I have a relatively large application developed with php and postgres
> using pg 6.3.2. I read the changelog for v.6.5 and i got a little bit
> scared, but maybe i understood things in the wrong way.
>
> In this application every "race condition" is placed in a begin->end
> block and most of them even have lock's before they're needed just mo
> make sure. If I upgrade postgres to 6.5 is there any danger that with
> the new locking mechanism my app's database starts getting corrupted or
> start's giving out the wrong type of information?

    Ok! Mea culpa. I probably didn't read the RELEASE file but i just saw a
message from someone with a similar problem.

    So, correct me if i'm wrong, as long as i use LOCK's i won't be running
into trouble, right ?

    BTW, do i have to install 6.4 to pass from 6.3 to 6.5?

P.S : I'm not thinking about upgrading the production server yetm just
the devel one...

--
        Mario Filipe
        mjnf@uevora.pt
        http://neptuno.sc.uevora.pt/~mjnf

Mario Jorge Nunes Filipe wrote:
>
> Hi
>
> I have a relatively large application developed with php and postgres
> using pg 6.3.2. I read the changelog for v.6.5 and i got a little bit
> scared, but maybe i understood things in the wrong way.
>
> In this application every "race condition" is placed in a begin->end
> block and most of them even have lock's before they're needed just mo
> make sure. If I upgrade postgres to 6.5 is there any danger that with
> the new locking mechanism my app's database starts getting corrupted or
> start's giving out the wrong type of information?

In 6.5.X LOCK TABLE without any option (without IN ... MODE)
does the same things as in prev versions: it stops _any_
(both read and write) concurrent access to locked table.
It's bad for performance of read-only transactions -
I would recommend to use LOCK TABLE IN EXCLUSIVE MODE
to block write access (insert/update/delete/select_for_update)
in update transactions and use SERIALIZABLE mode in read-only
ones.

Vadim