> 31 июля 2021 г., в 13:51, Peter Geoghegan <pg@bowt.ie> написал(а):
>
> maybe use rr for this?
Thanks, Peter!
I'll give it a try.
Currently I observe that during PrepareTransaction() there might be a moment when the transaction is not reported by
GetLockConflicts()with relevant lock.
I observe that PrepareTransaction() for violating xid is somewhere between
CallXactCallbacks(XACT_EVENT_PREPARE);
and
PostPrepare_PgStat();
And GetLockConflicts() fails to detect this transaction.
Thanks!
Best regards, Andrey Borodin.