Re: 2 new contributions to pgdg - Mailing list pgsql-pkg-debian

Re: To Jeremy Finzel 2018-11-30 <20181130102855.GA20278@msg.df7cb.de>
> It would be nice if you could have a look at the pglogical-ticker
> instabilities mentioned earlier. The tests seem to have a few race
> conditions that trigger occasionally.

Poking around a bit more, the pg-fact-loader tests are also unstable.
The two failures I've seen are:

12:33:26 Dropping cluster 11/regress ...
12:33:26 **** regression.diffs ****
12:33:26 --- /tmp/adt-run.TMZdfu/build.UnY/pg-fact-loader-1.5.2/expected/16_1_2_features.out    2018-11-29
20:17:47.000000000+0000
 
12:33:26 +++ /tmp/adt-run.TMZdfu/build.UnY/pg-fact-loader-1.5.2/results/16_1_2_features.out    2018-11-30
11:33:15.707006177+0000
 
12:33:26 @@ -235,18 +235,15 @@
12:33:26  \! psql contrib_regression -c ' SELECT fact_loader.worker() INTO try2;'
12:33:26  SELECT 1
12:33:26  SELECT pg_sleep(2);
12:33:26 -COMMIT
12:33:26   pg_sleep 
12:33:26  ----------
12:33:26   
12:33:26  (1 row)
12:33:26  
12:33:26  SELECT * FROM try1;
12:33:26 - worker 
12:33:26 ---------
12:33:26 - t
12:33:26 -(1 row)
12:33:26 -
12:33:26 +ERROR:  relation "try1" does not exist
12:33:26 +LINE 1: SELECT * FROM try1;
12:33:26 +                      ^
12:33:26  SELECT * FROM try2;
12:33:26   worker 
12:33:26  --------
12:33:26 @@ -292,6 +289,7 @@
12:33:26  $BODY$
12:33:26  LANGUAGE plpgsql;
12:33:26  UPDATE fact_loader.fact_tables SET enabled = FALSE;
12:33:26 +COMMIT
12:33:26  BEGIN;  --Keep the same transaction time to make these tests possible
12:33:26  INSERT INTO fact_loader.fact_tables (fact_table_relid, enabled, priority, use_daily_schedule,
daily_scheduled_time,daily_scheduled_tz, daily_scheduled_proid)
 
12:33:26  VALUES ('test_fact.daily_customers_fact', TRUE, 10, TRUE, now() + interval '1 second', 'America/Chicago',
'test_fact.daily_customers_fact_merge'::REGPROC);
12:33:26

12:30:28 Dropping cluster 9.5/regress ...
12:30:29 **** regression.diffs ****
12:30:29 --- /tmp/adt-run.Z3qsII/build.i9p/pg-fact-loader-1.5.2/expected/16_1_2_features.out    2018-11-29
20:17:47.000000000+0000
 
12:30:29 +++ /tmp/adt-run.Z3qsII/build.i9p/pg-fact-loader-1.5.2/results/16_1_2_features.out    2018-11-30
11:30:20.065918575+0000
 
12:30:29 @@ -227,15 +227,16 @@
12:30:29  \! psql contrib_regression -c 'BEGIN; SELECT fact_loader.worker() INTO try1; SELECT pg_sleep(2); COMMIT;' &
12:30:29  SELECT pg_sleep(1);
12:30:29  NOTICE:  table "process_queue" does not exist, skipping
12:30:29 +COMMIT
12:30:29   pg_sleep
12:30:29  ----------
12:30:29
12:30:29  (1 row)
12:30:29
12:30:29  \! psql contrib_regression -c ' SELECT fact_loader.worker() INTO try2;'
12:30:29 +NOTICE:  table "process_queue" does not exist, skipping
12:30:29  SELECT 1
12:30:29  SELECT pg_sleep(2);
12:30:29 -COMMIT
12:30:29   pg_sleep
12:30:29  ----------
12:30:29
12:30:29 @@ -250,7 +251,7 @@
12:30:29  SELECT * FROM try2;
12:30:29   worker
12:30:29  --------
12:30:29 - f
12:30:29 + t
12:30:29  (1 row)
12:30:29
12:30:29  --Daily schedule test - with range format suggestions included!!!
12:30:29

Maybe pg_sleep(2) isn't enough?

Christoph