Re: parallel restore

On Mon, 2008-12-29 at 18:42 -0500, Andrew Dunstan wrote:
> Attached is the latest parallel restore patch. I think this is getting
> fairly close.
>
> Includes help text, docco and some extra error checking.

Very brief review.

Hopefully the --truncate-before-load option works in both parallel mode
and data-only mode?

Few typos on docs.

No README or comments explain how the patch works. This part of the code
was probably most opaque already, so its really needed unfortunately.

* I'm particularly interested in error handling, for example if one
thread hits a deadlock, gets accidentally killed by user, hits bug in
custom add-in code etc..

* Earlier bugs with pre/post data were related to missing objects or
putting them in wrong groups. Documenting that will help identify
errors.

Few starnge names, sorry:

work_is_being_done --> work_in_progress
_inhibit_data?? --> _skip_data??
prestored --> parallel_restored
restore_one_te --> restore_TocEntry

Would like ability to increase/decrease number of parallel threads as
restore progresses. Experience says you always pick the wrong number
and/or situation changes while in progress.

--
Simon Riggs www.2ndQuadrant.com
PostgreSQL Training, Services and Support