On Mar 5, 2007, at 4:16 PM, Perrin Harkins wrote:
I'm no Postgres expert, but I don't think a SELECT should block an
UPDATE in an MVCC database system unless the first statement is a
SELECT FOR UPDATE or similar. Maybe if you use an isolation level of
REPEATABLE READ.
Yeah.
I think the system was configured to select for update by default.
There were a bunch of fixes possible:
change default isolation level
use the locking db handle for that call
"SELECT FOR SHARE"
migrate the call
i ended up being lazy, and just moved the call a few lines down, so
no issues would occur for now. not the best fix, but it works and
frees me some time for other important issues ;)
i also put a request in with rosedb , so that people can "FOR UPDATE"
or "FOR SHARE" explicitly, which I would have rather done. I tried
to patch things myself, but it started touching so many internals of
rose that I figured jon would yell at me for submitting a patch
before asking :)
// Jonathan Vanasco
| - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - -
| FindMeOn.com - The cure for Multiple Web Personality Disorder
| Web Identity Management and 3D Social Networking
| - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - -
| RoadSound.com - Tools For Bands, Stuff For Fans
| Collaborative Online Management And Syndication Tools
| - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - -
