Horton
Menu

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

Zhijie Hou (Fujitsu) <houzj.fnst@fujitsu.com>

View thread
From: "houzj.fnst@fujitsu.com" <houzj.fnst@fujitsu.com>
To: "Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>
Cc: Masahiko Sawada <sawada.mshk@gmail.com>, "wangw.fnst@fujitsu.com" <wangw.fnst@fujitsu.com>, Peter Smith <smithpb2250@gmail.com>, Dilip Kumar <dilipbalaut@gmail.com>, "shiy.fnst@fujitsu.com" <shiy.fnst@fujitsu.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Amit Kapila <amit.kapila16@gmail.com>
Date: 2022-12-04T11:18:02Z
Lists: pgsql-hackers 
On Friday, December 2, 2022 7:27 PM Kuroda, Hayato/黒田 隼人 <kuroda.hayato@fujitsu.com> wroteL
> 
> Dear Hou,
> 
> Thanks for making the patch. Followings are my comments for v54-0003 and
> 0004.

Thanks for the comments!

> 
> 0003
> 
> pa_free_worker()
> 
> +       /* Unlink any files that were needed to serialize partial changes. */
> +       if (winfo->serialize_changes)
> +               stream_cleanup_files(MyLogicalRepWorker->subid,
> winfo->shared->xid);
> +
> 
> I think this part is not needed, because the LA cannot reach here if
> winfo->serialize_changes is true. Moreover stream_cleanup_files() is done in
> pa_free_worker_info().

Removed.

> LogicalParallelApplyLoop()
> 
> The parallel apply worker wakes up every 0.1s even if we are in the
> PARTIAL_SERIALIZE mode. Do you have idea to reduce that?

The parallel apply worker usually will wait on the stream lock after entering
PARTIAL_SERIALIZE mode.

> ```
> +                       pa_spooled_messages();
> ```
> 
> Comments are needed here, like "Changes may be serialize...".

Added.

> pa_stream_abort()
> 
> ```
> +                               /*
> +                                * Reopen the file and set the file position to
> the saved
> +                                * position.
> +                                */
> +                               if (reopen_stream_fd)
> +                               {
> +                                       char            path[MAXPGPATH];
> +
> +                                       changes_filename(path,
> MyLogicalRepWorker->subid, xid);
> +                                       stream_fd =
> BufFileOpenFileSet(&MyParallelShared->fileset,
> +
> path, O_RDONLY, false);
> +                                       BufFileSeek(stream_fd, fileno, offset,
> SEEK_SET);
> +                               }
> ```
> 
> MyParallelShared->serialize_changes may be used instead of reopen_stream_fd.

These codes have been removed.

> 
> ```
> +                       /*
> +                        * It is possible that while sending this change to
> parallel apply
> +                        * worker we need to switch to serialize mode.
> +                        */
> +                       if (winfo->serialize_changes)
> +                               pa_set_fileset_state(winfo->shared,
> FS_READY);
> ```
> 
> There are three same parts in the code, can we combine them to common part?

These codes have been slightly refactored.

> apply_spooled_messages()
> 
> ```
> +               /*
> +                * Break the loop if the parallel apply worker has finished
> applying
> +                * the transaction. The parallel apply worker should have closed
> the
> +                * file before committing.
> +                */
> +               if (am_parallel_apply_worker() &&
> +                       MyParallelShared->xact_state ==
> PARALLEL_TRANS_FINISHED)
> +                       goto done;
> ```
> 
> I thnk pfree(buffer) and pfree(s2.data) should not be skippied.
> And this part should be at below "nchanges++;"

buffer, s2.data were allocated in the toplevel transaction's context and it
will be automatically freed soon when handling STREAM COMMIT.

> 
> 0004
> 
> set_subscription_retry()
> 
> ```
> +       LockSharedObject(SubscriptionRelationId, MySubscription->oid, 0,
> +                                        AccessShareLock);
> +
> ```
> 
> I think AccessExclusiveLock should be aquired instead of AccessShareLock.
> In AlterSubscription(), LockSharedObject(AccessExclusiveLock) seems to be
> used.

Changed.

Best regards,
Hou zj

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.