Horton

Re: PATCH: logical_work_mem and logical streaming of large in-progress transactions

Amit Kapila <amit.kapila16@gmail.com>

View thread
From: Amit Kapila <amit.kapila16@gmail.com>
To: Dilip Kumar <dilipbalaut@gmail.com>
Cc: Tomas Vondra <tomas.vondra@2ndquadrant.com>, Peter Eisentraut <peter.eisentraut@2ndquadrant.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2019-11-14T06:40:07Z
Lists: pgsql-hackers

Commits

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

  1. Tighten the concurrent abort check during decoding.

  2. Improve hash_create()'s API for some added robustness.

  3. Use HASH_BLOBS for xidhash.

  4. Fix initialization of RelationSyncEntry for streaming transactions.

  5. Remove unused function declaration in logicalproto.h.

  6. Add additional tests to test streaming of in-progress transactions.

  7. Fix inline marking introduced in commit 464824323e.

  8. Add support for streaming to built-in logical replication.

  9. Fix the SharedFileSetUnregister API.

  10. Fix comment in procarray.c

  11. Suppress compiler warning in non-cassert builds.

  12. Extend the BufFile interface.

  13. Mark a few logical decoding related variables with PGDLLIMPORT.

  14. Implement streaming mode in ReorderBuffer.

  15. Extend the logical decoding output plugin API with stream methods.

  16. WAL Log invalidations at command end with wal_level=logical.

  17. Immediately WAL-log subtransaction and top-level XID association.

  18. Allow logical replication to transfer data in binary format.

  19. Only superuser can set sslcert/sslkey in postgres_fdw user mappings

  20. Track statistics for spilling of changes from ReorderBuffer.

  21. Add logical_decoding_work_mem to limit ReorderBuffer memory usage.

  22. logical decoding: process ASSIGNMENT during snapshot build

  23. Emit invalidations to standby for transactions without xid.

On Thu, Nov 14, 2019 at 9:37 AM Dilip Kumar <dilipbalaut@gmail.com> wrote:
>
> On Wed, Nov 13, 2019 at 5:55 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> >
> > On Thu, Oct 3, 2019 at 1:18 PM Dilip Kumar <dilipbalaut@gmail.com> wrote:
> > >
> >
> > As mentioned by me a few days back that the first patch in this series
> > is ready to go [1] (I am hoping Tomas will pick it up), so I have
> > started the review of other patches
> >
> > Review/Questions on 0002-Immediately-WAL-log-assignments.patch
> > -------------------------------------------------------------------------------------------------
> > 1. This patch adds the top_xid in WAL whenever the first time WAL for
> > a subtransaction XID is written to correctly decode the changes of
> > in-progress transaction.  This patch also removes logging and applying
> > WAL for XLOG_XACT_ASSIGNMENT which might have some effect.  As replay
> > of that, it prunes KnownAssignedXids to prevent overflow of that
> > array.  See comments in procarray.c (KnownAssignedTransactionIds
> > sub-module).  Can you please explain how after removing the WAL for
> > XLOG_XACT_ASSIGNMENT, we will handle that or I am missing something
> > and there is no impact of same?
>
> It seems like a problem to me as well.   One option could be that
> since now we are adding the top transaction id in the first WAL of the
> subtransaction we can directly update the pg_subtrans and avoid adding
> sub transaction id in the KnownAssignedXids and mark it as
> lastOverflowedXid.
>

Hmm, I am not sure if we can do that easily because I think in
RecordKnownAssignedTransactionIds, we add those based on the gap via
KnownAssignedXidsAdd and only remove them later while applying WAL for
XLOG_XACT_ASSIGNMENT.  I think if we really want to go in this
direction then for each WAL record we need to check if it has
XLR_BLOCK_ID_TOPLEVEL_XID set and then call function
ProcArrayApplyXidAssignment() with the required information.  I think
this line of attack has WAL overhead both on master whenever
subtransactions are involved and also on hot-standby for doing the
work for each subtransaction separately.  The WAL apply needs to
acquire and release PROCArrayLock in exclusive mode for each
subtransaction whereas now it does it once for
PGPROC_MAX_CACHED_SUBXIDS number of subtransactions which can conflict
with queries running on standby.

The other idea could be that we keep the current XLOG_XACT_ASSIGNMENT
mechanism (WAL logging and apply of same on hot-standby) as it is and
additionally log top_xid the first time when WAL is written for a
subtransaction only when wal_level >= WAL_LEVEL_LOGICAL.  Then use the
same for logical decoding.  The advantage of this approach is that we
will incur the overhead of additional transactionid only when required
especially not with default server configuration.

Thoughts?

-- 
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com