RE: [HACKERS] [6.5.2] potentially major bug?

> > >
> > > Hmmm,isn't there old transaction running somewhere ?
> >
> > not that we are aware of ... there is a daemon running that is
> doing COPY
> > INs to the table ... how does something like that deal with a
> > vacuum? Will the vacuum wait for the COPY IN to end and/or
> prevent a COPY
> > IN from starting?
> >
>
> If a transaction read/write the target table it would be blocked
> by vacuum.
> But vacuum couldn't know what tables other backends would read/write in
> their running transactions. In MVCC old transaction have to see
> old deleted
> tuples in SERIALIZABLE isolation level and so vacuum doesn't remove the
> tuples which old transactions may see.
>

For example,if you type

begin;
select .. from .. ;

and leave your seat,any vacuum won't be able to remove tuples
deleted after the 'select .. from ..'.

Regards.

Hiroshi Inoue
Inoue(at)tpf(dot)co(dot)jp