Hello, Peter.
> The simple answer is: I don't know. I could probably come up with a
> better answer than that, but it would take real effort, and time.
I remember you had an idea about using the LP_REDIRECT bit in btree
indexes as some kind of “recently dead” flag (1).
Is this idea still in progress? Maybe an additional bit could provide
a space for a better solution.
> I think that you could do a better job of explaining and promoting the
> problem that you're trying to solve here. Emphasis on the problem, not
> so much the solution.
System I am working on highly depends on the performance of reading
from standby. In our workloads queries on standby are sometimes
10-100x slower than on primary due to absence of LP_DEAD support.
Other users have the same issues (2). I believe such functionality is
great optimization for read replicas with both analytics and OLTP
(read-only) workloads.
> You must understand that this whole area is *scary*. The potential for
> serious data corruption bugs is very real. And because the whole area
> is so complicated (it is at the intersection of 2-3 complicated
> areas), we can expect those bugs to be hidden for a long time. We
> might never be 100% sure that we've fixed all of them if the initial
> design is not generally robust. Most patches are not like that.
Moved to “Waiting for Author” for now.
[1]:
https://www.postgresql.org/message-id/flat/CAH2-Wz%3D-BoaKgkN-MnKj6hFwO1BOJSA%2ByLMMO%2BLRZK932fNUXA%40mail.gmail.com#6d7cdebd68069cc493c11b9732fd2040
[2]: https://www.postgresql.org/message-id/flat/20170428133818.24368.33533%40wrigleys.postgresql.org
Thanks,
Michail.