On Wed, Aug 01, 2018 at 02:06:44PM +0300, Andrey Borodin wrote:
> But adding LSNs, and whole regular PageHeader is quite easy in this
> patch. Do you think we should really go that way?
>
> Putting SLRUs into usual shared buffers and WAL-logging looks like a good idea, but a lot of work.
Back at PgCon in Ottawa this year, I pitched the idea to moving
components off of SLRU and on to the buffer cache. That idea would
involve adding regular page headers to SLRU blocks, and transitioning
various components like multixact, commit timestamp, clog, subtrans and
others, to being first class residents in the buffer cache. The goal is
to provide performance benefits along with resiliency.
I believe Heikki also mentioned wanting support for stamping LSNs to
ensure durability in clog. I agree with the idea of incorporating the
PageHeader at the beginning of the SLRU page to maintain checksums and
we can iteratively add support for tracking LSNs if necessary.
--
Shawn Debnath
