Re: idle in transaction - Mailing list pgsql-general

From Tom Lane
Subject Re: idle in transaction
Date
Msg-id 5390.1008030631@sss.pgh.pa.us
Whole thread Raw
In response to idle in transaction  ("Rich Ryan" <rich@usedcars.com>)
Responses Need to vacuum twice for effectiveness?
List pgsql-general
"Rich Ryan" <rich@usedcars.com> writes:
> If I kill pid 4168, the idle in transaction process, then the vacuum and all
> the other waiting processes will proceed fine.

You've got a client that is sitting with an open transaction block ---
ie, BEGIN, do something, go to sleep without issuing COMMIT/ROLLBACK.

In 7.1 the "do something" could just be a SELECT; that's still
sufficient to block VACUUM.  I think that 7.0 might have been a little
laxer about that, and would not hold a lock for a SELECT-only
transaction.

            regards, tom lane

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: Using Cursor in PostgreSQL 7.2
Next
From: "Paul Laub"
Date:
Subject: Re: referential integrity on existing table