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>
Date: 2024-05-08T02:53:34Z
Lists: pgsql-hackers

Attachments

On Wed, May 1, 2024 at 4:29 PM John Naylor <johncnaylorls@gmail.com> wrote:
>
> On Thu, Apr 25, 2024 at 8:36 AM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
> >
> > On Mon, Apr 15, 2024 at 6:12 PM John Naylor <johncnaylorls@gmail.com> wrote:
>
> > > - RT_KEY_GET_SHIFT is not covered for key=0:
> > >
> > > https://anarazel.de/postgres/cov/16-vs-HEAD-2024-04-14/src/include/lib/radixtree.h.gcov.html#L803
> > >
> > > That should be fairly simple to add to the tests.
> >
> > There are two paths to call RT_KEY_GET_SHIFT():
> >
> > 1. RT_SET() -> RT_KEY_GET_SHIFT()
> > 2. RT_SET() -> RT_EXTEND_UP() -> RT_KEY_GET_SHIFT()
> >
> > In both cases, it's called when key > tree->ctl->max_val. Since the
> > minimum value of max_val is 255, RT_KEY_GET_SHIFT() is never called
> > when key=0.
>
> Ah, right, so it is dead code. Nothing to worry about, but it does
> point the way to some simplifications, which I've put together in the
> attached.

Thank you for the patch. It looks good to me.

+       /* compute the smallest shift that will allowing storing the key */
+       start_shift = pg_leftmost_one_pos64(key) / RT_SPAN * RT_SPAN;

The comment is moved from RT_KEY_GET_SHIFT() but I think s/will
allowing storing/will allow storing/.

>
> > > - RT_DELETE: "if (key > tree->ctl->max_val)" is not covered:
> > >
> > > https://anarazel.de/postgres/cov/16-vs-HEAD-2024-04-14/src/include/lib/radixtree.h.gcov.html#L2644
> > >
> > > That should be easy to add.
> >
> > Agreed. The patch is attached.
>
> LGTM
>
> > > - TidStoreCreate* has some memory clamps that are not covered:
> > >
> > > https://anarazel.de/postgres/cov/16-vs-HEAD-2024-04-14/src/backend/access/common/tidstore.c.gcov.html#L179
> > > https://anarazel.de/postgres/cov/16-vs-HEAD-2024-04-14/src/backend/access/common/tidstore.c.gcov.html#L234
> > >
> > > Maybe we could experiment with using 1MB for shared, and something
> > > smaller for local.
> >
> > I've confirmed that the local and shared tidstore with small max sizes
> > such as 4kB and 1MB worked. Currently the max size is hard-coded in
> > test_tidstore.c but if we use work_mem as the max size, we can pass
> > different max sizes for local and shared in the test script.
>
> Seems okay, do you want to try that and see how it looks?

I've attached a simple patch for this. In test_tidstore.sql, we used
to create two local tidstore and one shared tidstore. I thought of
specifying small work_mem values for these three cases but it would
remove the normal test cases. So I created separate tidstore for this
test. Also, the new test is just to check if tidstore can be created
with such a small size, but it might be a good idea to add some TIDs
to check if it really works fine.

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