Re: Race in "tablespace" test on Windows

On Thu, Nov 13, 2014 at 8:46 AM, Noah Misch <noah(at)leadboat(dot)com> wrote:
>
> On Tue, Nov 11, 2014 at 10:21:26AM +0530, Amit Kapila wrote:
> > On Sat, Nov 8, 2014 at 10:34 AM, Noah Misch <noah(at)leadboat(dot)com> wrote:
> > > Here is a briefer command sequence exhibiting the same problem:
> > >
> > > To make this work as well on Windows as it does elsewhere, DROP
TABLESPACE
> > > would need to wait for other backends to close relevant unlinked
files.
> > > Perhaps implement "wait_unlinked_files(const char *dirname)" to poll
> > unlinked,
> > > open files until they disappear. (An attempt to open an unlinked file
> > reports
> > > ERROR_ACCESS_DENIED. It might be tricky to reliably distinguish this
> > cause
> > > from other causes of that error, but it should be possible.)
> >
> > I think the proposed mechanism can work but the wait can be very long
> > (untill the backend holding descriptor executes another command).
>
> The DROP TABLESPACE could send a catchup interrupt.
>

Yeah, that can work.

> > Can we think of some other solution like in Drop Tablespace instead of
> > checking if directory is empty, check if there is no object that belongs
> > to database/cluster, then allow to forcibly delete that directory
someway.
>
> I'm not aware of a way to forcibly delete the directory. One could rename
> files to the tablespace top-level directory just before unlinking them.
Since
> DROP TABLESPACE never removes that directory, their continued presence
there
> would not pose a problem. (Compare use of the rename-before-unlink trick
in
> RemoveOldXlogFiles().) That adds the overhead of an additional system
call to
> every unlink, which might be acceptable. It may be possible to rename
after
> unlink, as-needed in DROP TABLESPACE.
>

Right. I think we can discuss further about which approach is better,
once someone decides to work on this issue.

With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com