Re: parallel mode and parallel contexts

On Thu, Mar 19, 2015 at 11:13 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> Here is yet another version of this patch. In addition to the fixes
> mentioned above, this version includes some minor rebasing around
> recent commits, and also better handling of the case where we discover
> that we cannot launch workers after all. This can happen because (1)
> dynamic_shared_memory_type=none, (2) the maximum number of DSM
> segments supported by the system configuration are already in use, or
> (3) the user creates a parallel context with nworkers=0. In any of
> those cases, the system will now create a backend-private memory
> segment instead of a dynamic shared memory segment, and will skip
> steps that don't need to be done in that case. This is obviously an
> undesirable scenario. If we choose a parallel sequential scan, we
> want it to launch workers and really run in parallel. Hopefully, in
> case (1) or case (3), we will avoid choosing a parallel plan in the
> first place, but case (2) is pretty hard to avoid completely, as we
> have no idea what other processes may or may not be doing with dynamic
> shared memory segments ... and, in any case, degrading to non-parallel
> execution beats failing outright.

I see that you're using git format-patch to generate this. But the
patch is only patch 1/4. Is that intentional? Where are the other
pieces?

I think that the parallel seqscan patch, and the assessing parallel
safety patch are intended to fit together with this patch, but I can't
find a place where there is a high level overview explaining just how
they fit together (I notice Amit's patch has an "#include
"access/parallel.h", which is here, but that wasn't trivial to figure
out). I haven't been paying too much attention to this patch series.

--
Peter Geoghegan