You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
it seems like a race condition; only happens when multiple workers are consuming 'update-records' queue
if I run it with a single worker, errors disappeared
what is also interesting, is the fact that the record gets created (so one of the workers succeeded) -- and it only affects deleted records; I checked a few and they were all deleted from the import pipeline
import_pipeline=> select * from change_log where key = 'deleted' and oldvalue = '2017arXiv170902057K';
id | created | key | newvalue | oldvalue
--------+----------------------------+---------+----------+---------------------
336520 | 2017-12-11 21:44:40.231005 | deleted | | 2017arXiv170902057K
(1 row)
import_pipeline=> select * from change_log where key = 'deleted' and oldvalue = '2017arXiv170902017E';
id | created | key | newvalue | oldvalue
--------+----------------------------+---------+----------+---------------------
239759 | 2017-12-11 21:41:56.536064 | deleted | | 2017arXiv170902017E
(1 row)
import_pipeline=> select * from change_log where key = 'deleted' and oldvalue = '2017arXiv170902037B';
id | created | key | newvalue | oldvalue
--------+----------------------------+---------+----------+---------------------
668949 | 2017-12-11 21:53:51.495622 | deleted | | 2017arXiv170902037B
(1 row)
import_pipeline=> select * from change_log where key = 'deleted' and oldvalue = '2017arXiv170901605M';
id | created | key | newvalue | oldvalue
--------+----------------------------+---------+----------+---------------------
635200 | 2017-12-11 21:52:55.371403 | deleted | | 2017arXiv170901605M
this is a very curious error, it only affected set of ~700K bibcodes that were submitted to fulltext pipeline
this is the stacktrace:
it seems like a race condition; only happens when multiple workers are consuming 'update-records' queue
if I run it with a single worker, errors disappeared
what is also interesting, is the fact that the record gets created (so one of the workers succeeded) -- and it only affects deleted records; I checked a few and they were all deleted from the import pipeline
The text was updated successfully, but these errors were encountered: