Horton

Re: Parallel worker hangs while handling errors.

vignesh C <vignesh21@gmail.com>

View thread View on postgr.es
From: vignesh C <vignesh21@gmail.com>
To: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Robert Haas <robertmhaas@gmail.com>
Date: 2020-07-28T05:35:05Z
Lists: pgsql-hackers

Attachments

Thanks for your comments Bharath.

On Mon, Jul 27, 2020 at 10:13 AM Bharath Rupireddy
<bharath.rupireddyforpostgres@gmail.com> wrote:
> 1. Do we need "worker" as a function argument in
> update_parallel_worker_sigmask(BackgroundWorker *worker,.... ? Since
> MyBgworkerEntry is a global variable, can't we have a local variable
> instead?

Fixed, I have moved the worker check to the caller function.

> 2. Instead of update_parallel_worker_sigmask() serving only for
> parallel workers, can we make it generic, so that for any bgworker,
> given a signal it unblocks it, although there's no current use case
> for a bg worker unblocking a single signal other than a parallel
> worker doing it for SIGUSR1 for this hang issue. Please note that we
> have BackgroundWorkerBlockSignals() and
> BackgroundWorkerUnblockSignals().

Fixed. I have slightly modified the changes to break into
BackgroundWorkerRemoveBlockSignal & BackgroundWorkerAddBlockSignal.
This maintains the consistency similar to
BackgroundWorkerBlockSignals() and BackgroundWorkerUnblockSignals().

> If okay, with the BackgroundWorkerUpdateSignalMask() function, please
> note that we might have to add it in bgworker.sgml as well.

Included the documentation.

Attached the updated patch for the same.

Regards,
Vignesh
EnterpriseDB: http://www.enterprisedb.com

Commits

  1. Centralize setup of SIGQUIT handling for postmaster child processes.

  2. Accept SIGQUIT during error recovery in auxiliary processes.

  3. Avoid lockup of a parallel worker when reporting a long error message.