On E, 2005-08-22 at 02:14 +0300, Hannu Krosing wrote:
> Probably nonInVacuumXmin needs more care, i.e. initialising and setting
> it outside GetSnapshotData, at trx start and/or end. I'm too sleepy now
> to investigate further (it's 2:10 am here).
The attached patch works now as advertized so that concurrent non-vacuum
transactions don't interfere with vacuums OldestXmin calculations.
I fixed also the "VACUUM FULL case it manages to invoke *both*
full_vacuum_rel and lazy_vacuum_rel " bug.
But I could not find the breakage (from your Aug 17 email) with
> > In my patch I specifically exclude TruncateSUBTRANS from using the
> > inVacuum flag
>
> You missed vac_truncate_clog, though.
Has somethoing changed in vac_truncate_clog or am I just too
lazy/stupid ?
--
Hannu Krosing <hannu@skype.net>