On Mon, Jul 15, 2024 at 12:47 PM Adrian Klaver <adrian.kla...@aklaver.com> wrote:
> On 7/15/24 09:21, Ron Johnson wrote: > > On Mon, Jul 15, 2024 at 11:37 AM Adrian Klaver > > <adrian.kla...@aklaver.com <mailto:adrian.kla...@aklaver.com>> wrote: > > > > > > I don't think it is entirely coincidental that 1210 is the only shown > > user_id with a modified_on value that is in proximity to the delete > > error. > > > > > > I don't think so either. > > > > My suspicion is that actions are not happening in the exact order > > you think they are. > > > > > > modified_on is CURRENT_TIMESTAMP or NOW() or somesuch. I'm not sure, > > because I'm not privy to the code. > > > > But I'm printing the system time in bash before every statement. > > That is why I wrote 'Time travel?'. > > I suspect the modified_on time in the table is not accurately > representing when the row is modified. > That JBDC code is pretty slow... > > > > > I would think that combining DELETE FROM > > rel_group_user; and DELETE FROM public.access_user; in a single > > transaction would be a good start to fixing this. > > > > > > That is in fact what I'm working on now. There are 26 tables, and they > > must be done in a specific order when deleting, and the reverse while > > inserting. > > > > postgres_fdw would make this easier... > > It can't be installed? > Less bureaucratic overhead to write a script.
