From: | Hannu Krosing <hannu(at)skype(dot)net> |
---|---|
To: | Rod Taylor <pg(at)rbt(dot)ca> |
Cc: | PostgreSQL Development <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: Constraint Exclusion and Partition Locking |
Date: | 2006-03-04 19:29:44 |
Message-ID: | 1141500584.3772.15.camel@localhost.localdomain |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Ühel kenal päeval, L, 2006-03-04 kell 13:16, kirjutas Rod Taylor:
> Adding a new partition is fairly simple, particularly if you don't care
> about writing RULEs to direct data into it and can rely on a bulk loader
> to figure that part out.
>
> Removing a partition, on the other hand, is currently impossible to do
> without blocking selects against the table as a whole.
>
> The reason for this is that when planning to see if a partition needs to
> be read an AccessShareLock is taken and held for the entire transaction.
>
> If that lock was not there a partition (inherited table) could be
> dropped without any interruptions to normal query processing.
>
> Is it possible to release the AccessShareLock lock on
> "measurement_yy04mm03" after planning has determined that the structure
> will not be used for this query in a "READ COMMITTED" transaction?
> Serializable obviously would have problems with the partitions
> constraint definition changing.
Ar you sure ? I suspect that the change will not be visible to a
serialisable trx.
-------------
Hannu
From | Date | Subject | |
---|---|---|---|
Next Message | Stefan Kaltenbrunner | 2006-03-04 19:37:01 | EXPLAIN and HashAggregate |
Previous Message | Michael Paesold | 2006-03-04 18:42:29 | Re: Not so happy with psql's new multiline behavior |