Horton

Re: [PoC] Improve dead tuple storage for lazy vacuum

Masahiko Sawada <sawada.mshk@gmail.com>

View thread View on postgr.es
From: Masahiko Sawada <sawada.mshk@gmail.com>
To: John Naylor <johncnaylorls@gmail.com>
Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>, Nathan Bossart <nathandbossart@gmail.com>, Andres Freund <andres@anarazel.de>, Matthias van de Meent <boekewurm+postgres@gmail.com>, Yura Sokolov <y.sokolov@postgrespro.ru>
Date: 2024-03-20T13:30:06Z
Lists: pgsql-hackers 
On Wed, Mar 20, 2024 at 3:48 PM John Naylor <johncnaylorls@gmail.com> wrote:
>
> On Thu, Mar 14, 2024 at 12:06 PM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
> >
> > On Thu, Mar 14, 2024 at 1:29 PM John Naylor <johncnaylorls@gmail.com> wrote:
> > > Locally (not CI), we should try big inputs to make sure we can
> > > actually go up to many GB -- it's easier and faster this way than
> > > having vacuum give us a large data set.
> >
> > I'll do these tests.
>
> I just remembered this -- did any of this kind of testing happen? I
> can do it as well.

I forgot to report the results. Yes, I did some tests where I inserted
many TIDs to make the tidstore use several GB memory. I did two cases:

1. insert 100M blocks of TIDs with an offset of 100.
2. insert 10M blocks of TIDs with an offset of 2048.

The tidstore used about 4.8GB and 5.2GB, respectively, and all lookup
and iteration results were expected.

>
> > Thank you. I've incorporated all the comments above. I've attached the
> > latest patches, and am going to push them (one by one) after
> > self-review again.
>
> One more cosmetic thing in 0001 that caught my eye:
>
> diff --git a/src/backend/access/common/Makefile
> b/src/backend/access/common/Makefile
> index b9aff0ccfd..67b8cc6108 100644
> --- a/src/backend/access/common/Makefile
> +++ b/src/backend/access/common/Makefile
> @@ -27,6 +27,7 @@ OBJS = \
>   syncscan.o \
>   toast_compression.o \
>   toast_internals.o \
> + tidstore.o \
>   tupconvert.o \
>   tupdesc.o
>
> diff --git a/src/backend/access/common/meson.build
> b/src/backend/access/common/meson.build
> index 725041a4ce..a02397855e 100644
> --- a/src/backend/access/common/meson.build
> +++ b/src/backend/access/common/meson.build
> @@ -15,6 +15,7 @@ backend_sources += files(
>    'syncscan.c',
>    'toast_compression.c',
>    'toast_internals.c',
> +  'tidstore.c',
>    'tupconvert.c',
>    'tupdesc.c',
>  )
>
> These aren't in alphabetical order.

Good catch. I'll fix them before the push.

While reviewing the codes again, the following two things caught my eyes:

in check_set_block_offset() function, we don't take a lock on the
tidstore while checking all possible TIDs. I'll add
TidStoreLockShare() and TidStoreUnlock() as follows:

+           TidStoreLockShare(tidstore);
            if (TidStoreIsMember(tidstore, &tid))
                ItemPointerSet(&items.lookup_tids[num_lookup_tids++],
blkno, offset);
+           TidStoreUnlock(tidstore);

---
Regarding TidStoreMemoryUsage(), IIUC the caller doesn't need to take
a lock on the shared tidstore since dsa_get_total_size() (called by
RT_MEMORY_USAGE()) does appropriate locking. I think we can mention it
in the comment as follows:

-/* Return the memory usage of TidStore */
+/*
+ * Return the memory usage of TidStore.
+ *
+ * In shared TidStore cases, since shared_ts_memory_usage() does appropriate
+ * locking, the caller doesn't need to take a lock.
+ */

What do you think?

Regards,

--
Masahiko Sawada
Amazon Web Services: https://aws.amazon.com

Commits

  1. radixtree: Fix SIGSEGV at update of embeddable value to non-embeddable.

  2. Get rid of anonymous struct

  3. Teach radix tree to embed values at runtime

  4. Teach TID store to skip bitmap for small numbers of offsets

  5. Use bump context for TID bitmaps stored by vacuum

  6. Fix alignment of stack variable

  7. Use TidStore for dead tuple TIDs storage during lazy vacuum.

  8. Rethink create and attach APIs of shared TidStore.

  9. Fix inconsistent function prototypes with function definitions.

  10. Fix a calculation in TidStoreCreate().

  11. Fix potential integer handling issue in radixtree.h.

  12. Add TIDStore, to store sets of TIDs (ItemPointerData) efficiently.

  13. Fix link error for test_radixtree module on Windows

  14. Blind attempt to fix ODR violations

  15. Fix incorrect format specifier for int64

  16. Fix redefinition of typedefs

  17. Add template for adaptive radix tree

  18. Fix signedness error in 9f225e992 for gcc

  19. Introduce helper SIMD functions for small byte arrays

  20. Optimize vacuuming of relations with no indexes.

  21. Add bound check before bsearch() for performance

  22. Allocate consecutive blocks during parallel seqscans