Hi,
On 2018-10-22 20:18:32 +0200, Juan José Santamaría Flecha wrote:
> When you say corruption, do you mean that the data was corrupted
> > afterwards? Or just that the standby spewed errors until you restarted
> > with adjusted limits?
> >
>
> We didn't do a proper low level check to verify that the database was
> corrupted, but the symtopms took us to belive so: after a restart the
> system crashed almost immediately with and 'invalid memory alloc request
> size', no matter the user limits.
>
> There is a backup still available if any test can be of interest.
Yea, I think looking at that would be worthwhile. That an FD leak or
such could cause replay to get stuck is not that suprrising, but it
shouldn't cause corruption. That hints that something further is borked.
Greetings,
Andres Freund