rolfm at denison
Jan 21, 2009, 9:15 AM
Views: 7064
Permalink
|
Well, we've got the first error message from a story version being
corrupted thanks to our new unique key index on story instance. The
scenerio:
A person checked in a story. The check in failed in some manner, as it
is still on their workflow. If you go to the desk, there is a button
that claims it can be checked out. Attempting to check it out results
in this error in the logs
Unable to execute SQL statement: DBD::Pg::st execute failed: ERROR:
duplicate key violates unique constraint "udx_story__story_instance"\n
[.for Statement "INSERT INTO story_instance (id, name, description,
story__id, version, usr__id, primary_oc__id, slug, cover_date, note,
checked_out)VALUES
(NEXTVAL('seq_story_instance'), ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)" with
ParamValues: 6='1', 3='7616', 7='Overview', 9=undef, 2='',
8='2007-06-20 19:15:00', 1='Overview', 4='11', 10='1', 5='1029'] at /
usr/local/bricolage/lib/Bric/Util/DBI.pm line 1087, <GEN1620> line 68.\n
[/usr/local/bricolage/lib/Bric/Util/DBI.pm:1088]
Which makes sense. Go to the story instance table, though, and we see
this:
id | name | description | story__id | version | usr__id |
slug | primary_oc__id | cover_date | note | checked_out
...
73198 | Overview | | 7616 | 7 | 1366 |
Overview | 1 | 2007-06-20 19:15:00 | | f
73304 | Overview | | 7616 | 8 | 1366 |
Overview | 1 | 2007-06-20 19:15:00 | | f
90836 | Overview | | 7616 | 9 | 1312 |
Overview | 1 | 2007-06-20 19:15:00 | | f
101200 | Overview | | 7616 | 10 | 1312 |
Overview | 1 | 2007-06-20 19:15:00 | | f
101212 | Overview | | 7616 | 11 | 1312 |
Overview | 1 | 2007-06-20 19:15:00 | | f
101799 | Overview | | 7616 | 11 | 1312 |
Overview | 1 | 2007-06-20 19:15:00 | | t
Now we actually have lots of stories that look like this - 435 to be
exact. Most of the time it means that the latest version is being
worked on, and hasn't been checked in yet. But you've got the
database creating a new record with the same version when a story is
checked out, and then incrementing that on checkin. Maybe that is
part of the problem here, and it would be better to increment the
version number on the initial check out - maybe it would help prevent
this sort of thing. But I'm not sure why the checkin partially works,
but leaves a version on her workspace and doesn't increment the
version number.
What do other people think?
-Matt
|
Story version corruption
