On Monday, April 24, 2023 5:50 PM Yu Shi (Fujitsu) <shiy.f...@fujitsu.com> wrote: > > On Fri, Apr 21, 2023 1:48 PM Yu Shi (Fujitsu) <shiy.f...@fujitsu.com> wrote: > > > > I wrote a patch to dump rel state in wait_for_subscription_sync() as > suggested. > > Please see the attached patch. > > I will try to add some debug logs in code later. > > > > Please see the attached v2 patch. > > I added some debug logs when invalidating syncing table states and updating > table_states_not_ready list. I also adjusted the message level in the tests > which > failed before.
Just a reference. I think similar issue has been analyzed in other thread[1] and the reason seems clear that the table state cache invalidation got lost due to a race condition. The fix is also being discussed there. [1] https://www.postgresql.org/message-id/flat/711a6afe-edb7-1211-cc27-1bef8239eec7%40gmail.com Best Regards, Hou zj