Обсуждение: Unique transaction ID

Apart from now() function, is there some kind of sequence that advances
with each transaction started? If not: is it possible to implement that
using sequences?

Robson.

On 1/4/06, Robert Osowiecki <robson@cavern.pl> wrote:
> Apart from now() function, is there some kind of sequence that advances
> with each transaction started? If not: is it possible to implement that
> using sequences?
>
> Robson.
>
>

it exists a transaction id called XID, you can't get it via sql
though, you need to make a c function and invoke GetTopTransactionId()
(or GetCurrentTransactionId() in versions prior to 8.0)

--
regards,
Jaime Casanova
(DBA: DataBase Aniquilator ;)

Jaime Casanova napisał(a):

> it exists a transaction id called XID, you can't get it via sql
>
>though, you need to make a c function and invoke GetTopTransactionId()
>(or GetCurrentTransactionId() in versions prior to 8.0)
>
>
I've read a little about problems with overlapping XID and how they were
resolved. I don't understand everything but it seems that XID could be
decreased after vacuum, could it?

Robson.

On 1/4/06, Robert Osowiecki <robson@cavern.pl> wrote:
> Jaime Casanova napisał(a):
>
> > it exists a transaction id called XID, you can't get it via sql
> >
> >though, you need to make a c function and invoke GetTopTransactionId()
> >(or GetCurrentTransactionId() in versions prior to 8.0)
> >
> >
> I've read a little about problems with overlapping XID and how they were
> resolved. I don't understand everything but it seems that XID could be
> decreased after vacuum, could it?
>
> Robson.
>
>

NO!!!

if for overlapping you are refering to the wraparound problem you just
have to execute VACUUM at least every billion transactions or if you
are in 8.1.x let autovaccum keep you safe...

http://www.postgresql.org/docs/current/static/maintenance.html#VACUUM-FOR-WRAPAROUND

--
regards,
Jaime Casanova
(DBA: DataBase Aniquilator ;)