Re: Race conditions during parallel worker (unclean) exit

Sami Imseih <samimseih(dot)pg(at)gmail(dot)com> wrote:

> > > It seems cleaner to introduce a separate message for this case, maybe
> > > "PqMsg_ParallelAborted", and keep the existing
> > > "ERROR: lost connection to parallel worker" message.
> >
> > I'm not sure this is worth adjusting the FE/BE protocol. I'd prefer regular
> > ERROR message (PqMsg_ErrorResponse), but only in some minimalistic form. The
> > worker would only send the message, w/o doing any error handling itself.
>

> I was thinking of PqMsg_ParallelAborted because we have precedent already
> with PqMsg_Progress which was introduced in f1889729dd3a as the 'P' message
> for parallel index progress reported, but later converted to a macro in
> a99cc6c6b4b.

I realize now that new message type for communication between background
worker and the leader backend actually shouldn't need the protocol version to
be bumped - see commit f1889729dd3a. However, when adding a new message type,
I'd use the opportunity to stop using PqMsg_Terminate in the background
worker: per documentation, this one should be sent by frontend to backend.

A new message, e.g. PqMsg_WorkerExit, would contain status code. The worker
would set zero status code when sending the message at the end of
ParallelWorkerMain(), and non-zero in ParallelWorkerShutdown(). Does that make
sense to you?

--
Antonin Houska
Web: https://www.cybertec-postgresql.com