The following bug has been logged on the website:
Bug reference: 15149
Logged by: Dmitry Chirkin
Email address: dmitry.chirkin@gmail.com
PostgreSQL version: 10.3
Operating system: Debian
Description:
After upgrading to 10.3 I have found strange issue with
pg_try_advisory_lock
Here is my simplified workflow which is executed every minute:
1. TX start
2. select pg_try_advisory_lock(42) as lock; // we have several workers so we
don't want to start more than one simultaneously.
if lock is true -> continue, else -> rollback and finish.
3. Make a query via oracle_fdw, usually return 70-80 rows.
4. insert on conflict update in table1.
5. run some analyze query with several CTE's -> insert some artifacts into
table2.
6. run another one analyze query with several CTE's -> inserts some
artifacts into table 2.
7. TX commit.
So, few days ago i've added step 6 and on step 7 i'm getting error: Invalid
cache id: 42
On 10.2 everything working fine but on 10.3 - i'm getting error.
I'm not sure it's possible to provide good steps to reproduce with plain SQL
queries because of fdw but maybe this would be enough to understand what is
the reason for issue.
Thank you!