> On Oct 25, 2026, at 17:32, Álvaro Herrera <[email protected]> wrote: > > On 2026-Sep-03, Kirill Reshke wrote: > >>> Potential issue remaining >>> =================== >>> >>> Currently, find_all_inheritors() sorts all inheritors purely by OID. >>> In most cases this works, because users typically create a >>> partitioned table first and then add partitions later, so the >>> partitioned table’s OID is smaller than those of its partitions. > > Hmm, I suppose I just added one more possibility for this kind of > deadlock failure with the 2ebd6849743 fix; maybe we should apply this > methodology to that code also. > > -- > Álvaro Herrera PostgreSQL Developer — https://www.EnterpriseDB.com/ > Are you not unsure you want to delete Firefox? > [Not unsure] [Not not unsure] [Cancel] > http://smylers.hates-software.com/2008/01/03/566e45b2.html
Thanks for pointing out that, looks like ALTER INDEX ATTACH PARTITION has the same deadlock issue. PFA v5: * 0001 - just added one line comment based on v4-0001 * 0002 - fixed the ALTER INDEX ATTACH PARTITION by leveraging 0001’s mechanism * 0003 - based on v4-0002, added a new test for 0002. Without 0002, the test will fail Best regards, -- Chao Li (Evan) HighGo Software Co., Ltd. https://www.highgo.com/
v5-0001-Prevent-deadlock-between-ALTER-TABLE-recursion-an.patch
Description: Binary data
v5-0002-Prevent-deadlocks-during-partitioned-index-attach.patch
Description: Binary data
v5-0003-Add-isolation-tests-for-ATTACH-PARTITION-lock-ord.patch
Description: Binary data
