On Fri, Nov 08, 2019 at 06:44:08PM -0800, Daniel Wood wrote:
> I repro'ed on PG11 and PG10 STABLE but several months old.
> I looked at 6d05086 but it doesn't address the core issue.
>
> DropRelFileNodeBuffers prevents the checkpoint from writing all
> needed dirty pages for any REDO's that exist BEFORE the truncate.
> If we crash after a checkpoint but before the physical truncate then
> the REDO will need to replay the operation against the dirty page
> that the Drop invalidated.
I am beginning to look at this thread more seriously, and I'd like to
first try to reproduce that by myself. Could you share the steps you
used to do that? This includes any manual sleep calls you may have
added, the timing of the crash, manual checkpoints, debugger
breakpoints, etc. It may be possible to extract some more generic
test from that.
--
Michael