Horton
Menu

Re: Perform streaming logical transactions by background workers and parallel apply

Amit Kapila <amit.kapila16@gmail.com>

View thread
From: Amit Kapila <amit.kapila16@gmail.com>
To: Peter Smith <smithpb2250@gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Masahiko Sawada <sawada.mshk@gmail.com>, "wangw.fnst@fujitsu.com" <wangw.fnst@fujitsu.com>, Dilip Kumar <dilipbalaut@gmail.com>, "shiy.fnst@fujitsu.com" <shiy.fnst@fujitsu.com>, "houzj.fnst@fujitsu.com" <houzj.fnst@fujitsu.com>
Date: 2022-12-06T03:51:16Z
Lists: pgsql-hackers 
On Tue, Dec 6, 2022 at 5:27 AM Peter Smith <smithpb2250@gmail.com> wrote:
>
> Here are my review comments for patch v55-0002
>
...
>
> 3. pa_spooled_messages
>
> Previously I suggested this function name should be changed but that
> was rejected (see [1] #6a)
>
> > 6a.
> > IMO a better name for this function would be
> > pa_apply_spooled_messages();
> Not sure about this.
>
> ~
>
> FYI the reason for the previous suggestion is because there is no verb
> in the current function name, so the reader is left thinking
> pa_spooled_messages "what"?
>
> It means the caller has to have extra comments like:
> /* Check if changes have been serialized to a file. */
> pa_spooled_messages();
>
> OTOH, if the function was called something better -- e.g.
> pa_check_for_spooled_messages() or similar -- then it would be
> self-explanatory.
>

I think pa_check_for_spooled_messages() could be misleading because we
do apply the changes in that function, so probably a comment as
suggested by you is a better option.

> ~
>
> 4.
>
>  /*
> + * Replay the spooled messages in the parallel apply worker if the leader apply
> + * worker has finished serializing changes to the file.
> + */
> +static void
> +pa_spooled_messages(void)
>
> I'm not 100% sure of the logic, so IMO maybe the comment should say a
> bit more about how this works:
>
> Specifically, let's say there was some timeout and the LA needed to
> write the spool file, then let's say the PA timed out and found itself
> inside this function. Now, let's say the LA is still busy writing the
> file -- so what happens next?
>
> Does this function simply return, then the main PA loop waits again,
> then the times out again, then PA finds itself back inside this
> function again... and that keeps happening over and over until
> eventually the spool file is found FS_READY? Some explanatory comments
> might help.
>

No, PA will simply wait for LA to finish. See the code handling for
FS_BUSY state. We might want to slightly improve part of the current
comment to: "If the leader apply worker is busy serializing the
partial changes then acquire the stream lock now and wait for the
leader worker to finish serializing the changes".

>
> 16. apply_spooled_messages
>
> + stream_fd = BufFileOpenFileSet(stream_fileset, path, O_RDONLY, false);
>
> Something still seems a bit odd about this to me (previously also
> mentioned in review [1] #29) but I cannot quite put my finger on it...
>
> AFAIK the 'stream_fd' is the global the LA is using to remember the
> single stream spool file; It corresponds to the LogicalRepWorker's
> 'stream_fileset'. So using that same global on the PA side somehow
> seemed strange to me. The fileset at PA comes from a different place
> (MyParallelShared->fileset).
>

I think 'stream_fd' is specific to apply module which can be used by
apply, tablesync, or parallel worker. Unfortunately, now, the code in
worker.c is a mix of worker and apply module. At some point, we should
separate apply logic to a separate file.

-- 
With Regards,
Amit Kapila.

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Fix invalid memory access during the shutdown of the parallel apply worker.

  2. Fix assertion failure in apply worker.

  3. Use elog to report unexpected action in handle_streamed_transaction().

  4. Use appropriate wait event when sending data in the apply worker.

  5. Allow the logical_replication_mode to be used on the subscriber.

  6. Rename GUC logical_decoding_mode to logical_replication_mode.

  7. Display the leader apply worker's PID for parallel apply workers.

  8. Improve the code to decide and process the apply action.

  9. Document the newly added wait events added by commit 216a784829.

  10. Perform apply of large transactions by parallel workers.

  11. Wake up a subscription's replication worker processes after DDL.

  12. Add copyright notices to meson files

  13. Better document logical replication parameters

  14. Add a common function to generate the origin name.

  15. Harmonize parameter names in storage and AM code.

  16. Avoid using list_length() to test for empty list.

  17. Improve two comments related to a boolean DefElem's value

  18. Fix partition table's REPLICA IDENTITY checking on the subscriber.

  19. Fix data inconsistency between publisher and subscriber.

  20. Fix cache look-up failures while applying changes in logical replication.