Re: Doubt about AccessExclusiveLock in ALTER TABLE .. SET ( .. );

On Thu, Jul 30, 2015 at 10:46 PM, Michael Paquier <michael(dot)paquier(at)gmail(dot)com>
wrote:
>
> This patch size has increased from 16k to 157k because of the output
> of the isolation tests you just added. This is definitely too large
> and actually the test coverage is rather limited. Hence I think that
> they should be changed as follows:
> - Use only one table, the locks of tables a and b do not conflict, and
> that's what we want to look at here.
> - Check the locks of all the relation parameters, by that I mean as
> well fillfactor and user_catalog_table which still take
> AccessExclusiveLock on the relation
> - Use custom permutations. I doubt that there is much sense to test
> all the permutations in this context, and this will reduce the
> expected output size drastically.
>

Ok.

> On further notice, I would recommend not to use the same string name
> for the session and the query identifiers. And I think that inserting
> only one tuple at initialization is just but fine.
>
> Here is an example of such a spec:
> setup
> {
> CREATE TABLE a (id int PRIMARY KEY);
> INSERT INTO a SELECT generate_series(1,100);
> }
> teardown
> {
> DROP TABLE a;
> }
> session "s1"
> step "b1" { BEGIN; }
> # TODO add one query per parameter
> step "at11" { ALTER TABLE a SET (fillfactor=10); }
> step "at12" { ALTER TABLE a SET
(autovacuum_vacuum_scale_factor=0.001); }
> step "c1" { COMMIT; }
> session "s2"
> step "b2" { BEGIN; }
> step "wx1" { UPDATE a SET id = id + 10000; }
> step "c2" { COMMIT; }
>
> And by testing permutations like for example that it is possible to
> see which session is waiting for what. Input:
> permutation "b1" "b2" "at11" "wx1" "c1" "c2"
> Output where session 2 waits for lock taken after fillfactor update:
> step b1: BEGIN;
> step b2: BEGIN;
> step at11: ALTER TABLE a SET (fillfactor=10);
> step wx1: UPDATE a SET id = id + 10000; <waiting ...>
> step c1: COMMIT;
> step wx1: <... completed>
> step c2: COMMIT;
>
> Be careful as well to not include incorrect permutations in the
> expected output file, the isolation tester is smart enough to ping you
> about that.
>

Changed the isolation tests according your suggestions.

Regards,

--
Fabrízio de Royes Mello
Consultoria/Coaching PostgreSQL
>> Timbira: http://www.timbira.com.br
>> Blog: http://fabriziomello.github.io
>> Linkedin: http://br.linkedin.com/in/fabriziomello
>> Twitter: http://twitter.com/fabriziomello
>> Github: http://github.com/fabriziomello