Обсуждение: deadlocks problem

On Fri, 16 May 2003, Alexander Goldybin wrote:

> we are using PostgreSQL 7.3.2 for some monthes and are very glad with
> overall perfomance and features of pgSQL !

> But in the last time we experiencing some troubles with deadlocks in
> some tables :-(((. The problem occures from time to time if multiple
> transactions do UPDATE the same table. The found solution (or better
> to say "workaround") for this problem consists of removing all the
> foreign keys in this table. The question now - is there any other
> solution, because this workaround damages the referential integrity of
> the database ? We have tested the transaction isolation level
> SERIALIZEABLE, but with no success. Please help us. Thank you very
> much in advance !!!

You may wish to see recent posts on -hackers re-pointing to a patch that
eliminates one of the locking cases for the foreign keys. It sounds like
it'll be in 7.3.3, but you can try it now and see if it helps the cases
you're working with.

On Fri, 16 May 2003, Alexander Goldybin wrote:

> To Whom It May Concern
> --------------------------------------
>
> Hi,
>
> we are using PostgreSQL 7.3.2 for some monthes and are very glad with
> overall perfomance and features of pgSQL !
> But in the last time we experiencing some troubles with deadlocks in
> some tables :-(((. The problem occures from time to time if multiple
> transactions do UPDATE the same table. The found solution (or better to
> say "workaround") for this problem consists of removing all the foreign
> keys in this table. The question now - is there any other solution,
> because this workaround damages the referential integrity of the
> database
> ? We have tested the transaction isolation level SERIALIZEABLE, but with
> no success. Please help us. Thank you very much in advance !!!

To add to what was previously posted, the patch for 7.3.2 is found here:

ftp://ftp.postgresql.org/pub/patches/v7.3.2/