On 5 March 2014 09:28, Simon Riggs <simon@2ndquadrant.com> wrote:
> So that returns us to solving the catalog consistency problem in
> pg_dump and similar applications.
No answer, guys, and time is ticking away here.
I'd like to get a communal solution to this rather than just punting
the whole patch.
If we have to strip it down to the bar essentials, so be it. For me,
the biggest need here is to make VALIDATE CONSTRAINT take only a
ShareUpdateExclusiveLock while it runs. Almost everything else needs a
full AccessExclusiveLock anyway, or doesn't run for very long so isn't
a critical problem. (Perhaps we can then wrap ADD CONSTRAINT ... NOT
VALID and VALIDATE into a single command using the CONCURRENTLY
keyword so it runs two transactions to complete the task).
Validating FKs on big tables can take hours and it really isn't
acceptable for us to lock out access while we do that. FKs are
*supposed* to be a major reason people use RDBMS, so keeping them in a
state where they are effectively unusable is a major debilitating
point against adoption of PostgreSQL.
If there are issues with pg_dump we can just document them.
Guide me with your thoughts.
-- Simon Riggs http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services