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 <john.naylor@enterprisedb.com>
Cc: Nathan Bossart <nathandbossart@gmail.com>, Andres Freund <andres@anarazel.de>, Matthias van de Meent <boekewurm+postgres@gmail.com>, Yura Sokolov <y.sokolov@postgrespro.ru>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2022-11-30T08:53:20Z
Lists: pgsql-hackers 
On Tue, Nov 29, 2022 at 1:36 PM John Naylor
<john.naylor@enterprisedb.com> wrote:
>
> While creating a benchmark for inserting into node128-inner, I found a bug. If a caller deletes from a node128, the slot index is set to invalid, but the child pointer is still valid. Do that a few times, and every child pointer is valid, even if no slot index points to it. When the next inserter comes along, something surprising happens. This function:
>
> /* Return an unused slot in node-128 */
> static int
> node_inner_128_find_unused_slot(rt_node_inner_128 *node, uint8 chunk)
> {
>   int slotpos = 0;
>
>   Assert(!NODE_IS_LEAF(node));
>   while (node_inner_128_is_slot_used(node, slotpos))
>   slotpos++;
>
>   return slotpos;
> }
>
> ...passes an integer to this function, whose parameter is a uint8:
>
> /* Is the slot in the node used? */
> static inline bool
> node_inner_128_is_slot_used(rt_node_inner_128 *node, uint8 slot)
> {
>   Assert(!NODE_IS_LEAF(node));
>   return (node->children[slot] != NULL);
> }
>
> ...so instead of growing the node unnecessarily or segfaulting, it enters an infinite loop doing this:
>
> add     eax, 1
> movzx   ecx, al
> cmp     QWORD PTR [rbx+264+rcx*8], 0
> jne     .L147
>
> The fix is easy enough -- set the child pointer to null upon deletion,

Good catch!

> but I'm somewhat astonished that the regression tests didn't hit this. I do still intend to replace this code with something faster, but before I do so the tests should probably exercise the deletion paths more. Since VACUUM

Indeed, there are some tests for deletion but all of them delete all
keys in the node so we end up deleting the node. I've added tests of
repeating deletion and insertion as well as additional assertions.

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