citus icon indicating copy to clipboard operation
citus copied to clipboard

columnar_first_row test fails when run in parallel with columnar_query test

Open JelteF opened this issue 3 years ago • 2 comments

When running columnar_first_row and columnar_query in parallel sometimes columnar_first_row fails with the following error:

SELECT row_count, first_row_number FROM columnar.stripe a 
WHERE a.storage_id = columnar.get_storage_id('col_table_1'::regclass)
 ORDER BY stripe_num;
- row_count | first_row_number
----------------------------------------------------------------------
-        10 |                1
-        12 |           300001
-      1000 |           450001
-      1000 |           451001
-       350 |           452001
-(5 rows)
-
+ERROR:  could not open relation with OID 17223
 VACUUM FULL col_table_1;

Example of failed test: https://app.circleci.com/pipelines/github/citusdata/citus/26106/workflows/75d00ea9-23f8-4bff-a927-bced19e1f81b/jobs/736713

For reference @jeff-davis had these thoughts on this issue in #6184:

It seems that col_table_1 is only referenced in columnar_first_row_number.sql so I still don't understand the diagnosis. How does columnar_query.sql conflict with it?

JelteF avatar Aug 18 '22 10:08 JelteF

For reference Relation with OID 17223 is a table created and dropped in the parallel run test columnar_query, and referenced in columnar.stripe's storage_id column.

naisila avatar Aug 19 '22 08:08 naisila

If we vacuum full the tables in columnar_query test the issue doesn't seem to persist. I will open a draft PR for this now.

naisila avatar Aug 19 '22 08:08 naisila