On 09/03/2010 16:51, Tom Lane wrote:
> Andrea Suisani<sickpig@opinioni.net> writes:
>> I'm experiencing something weird. here the session's log
>> that involves the prob I mentioned
>
>> 2010-03-08 12:59:41 CET p@c[3189]ERROR: duplicate key value violates unique constraint "pg_type_typname_nsp_index"
>> 2010-03-08 12:59:41 CET p@c[3189]STATEMENT: create table check_unfitted_strata_col as select * from
sampling.sample_gen_designlimit 0
>
>> those sql commands could be executed in separate
>> postgres back-ends almost simultaneously (isolation level
>> is read committed), but I always thought wrapping the CREATE TABLE
>> command inside a transaction avoid any misbehavior (a part from the
>> fact that the second process will wait for the first to commit or rollback)
>
> Well, it does avoid any "misbehavior", it's just not producing exactly
> the error message you expect. If you have two sessions trying to create
> the same table name at exactly the same time, this is a likely result.
I'm not 100% sure but it's quite unlikely that they
try to create the table at exactly the same time.
With "almost simultaneously" I meant to indicate
some kind of time overlapping between the two session.
> They both have to create pg_type entries for the table's row type,
> so they can collide on that as well as on the table name itself.
take for granted the "at exactly same time" hypothesis, it does maks sense.
just after I've sent the email I've checked to which table pg_type_typname_nsp_index
belongs to and found that pg_type's the "owner"
> There have been some unexplained reports of similar error messages in
> cases where it didn't seem that any concurrent creation could be going
> on. But I don't see any reason to think there's anything very
> mysterious here, if you do have sessions trying to create the same table
> concurrently.
I will dig deeper and I'll try to monitor the
process for the next few days trying to get more infos.
I'll keep you posted
thanks
Andrea