On Fri, Aug 7, 2020 at 9:33 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Amit Kapila <amit.kapila16@gmail.com> writes:
> > On Sat, Aug 1, 2020 at 1:53 AM Andres Freund <andres@anarazel.de> wrote:
> >> We could also just use pg_class.relpages. It'll probably mostly be
> >> accurate enough?
>
> > Don't we need the accurate 'number of blocks' if we want to invalidate
> > all the buffers? Basically, I think we need to perform BufTableLookup
> > for all the blocks in the relation and then Invalidate all buffers.
>
> Yeah, there is no room for "good enough" here. If a dirty buffer remains
> in the system, the checkpointer will eventually try to flush it, and fail
> (because there's no file to write it to), and then checkpointing will be
> stuck. So we cannot afford to risk missing any buffers.
>
Today, again thinking about this point it occurred to me that during
recovery we can reliably find the relation size and after Thomas's
recent commit c5315f4f44 (Cache smgrnblocks() results in recovery), we
might not need to even incur the cost of lseek. Why don't we fix this
first for 'recovery' (by following something on the lines of what
Andres suggested) and then later once we have a generic mechanism for
"caching the relation size" [1], we can do it for non-recovery paths.
I think that will at least address the reported use case with some
minimal changes.
[1] -
https://www.postgresql.org/message-id/CAEepm%3D3SSw-Ty1DFcK%3D1rU-K6GSzYzfdD4d%2BZwapdN7dTa6%3DnQ%40mail.gmail.com
--
With Regards,
Amit Kapila.