Bug description

Under certain conditions, a deleted tuple isn't purged from a secondary index of a vinyl space.

Tarantool 3.3.0-entrypoint-199-g5325abd3441e
Target: Linux-x86_64-Debug
Build options: cmake . -DCMAKE_INSTALL_PREFIX=/home/vlad/src/tarantool/tarantool/build/debug/install -DENABLE_BACKTRACE=TRUE
Compiler: GNU-13.2.0
C_FLAGS: -fexceptions -funwind-tables -fasynchronous-unwind-tables -fno-common -msse2 -Wformat -Wformat-security -Werror=format-security -fstack-protector-strong -fPIC -fmacro-prefix-map=/home/vlad/src/tarantool/tarantool=. -std=c11 -Wall -Wextra -Wno-gnu-alignof-expression -fno-gnu89-inline -Wno-cast-function-type -Werror -g -ggdb -O0
CXX_FLAGS: -fexceptions -funwind-tables -fasynchronous-unwind-tables -fno-common -msse2 -Wformat -Wformat-security -Werror=format-security -fstack-protector-strong -fPIC -fmacro-prefix-map=/home/vlad/src/tarantool/tarantool=. -std=c++11 -Wall -Wextra -Wno-invalid-offsetof -Wno-gnu-alignof-expression -Wno-cast-function-type -Werror -g -ggdb -O0

Steps to reproduce

Run the following script:

os.execute('rm -rf [0-9]*')

box.cfg{log_level = 'warn'}

local s = box.schema.space.create('test', {engine = 'vinyl'})
s:create_index('primary')
s:create_index('secondary', {parts = {2, 'unsigned'}})

s:replace{1, 2}

box.begin()
s:replace{1, 1}
s:replace{1, 1}
box.commit()

box.snapshot()

print(string.format('primary: count = %d, len = %d',
                    s.index.primary:count(), s.index.primary:len()))
print(string.format('secondary: count = %d, len = %d',
                    s.index.secondary:count(), s.index.secondary:len()))
os.exit(0)

Actual behavior

Output:

primary: count = 1, len = 1
secondary: count = 1, len = 2

Note that discrepancy between len and count for the secondary index. They should be the same because the index is compacted.

$ $ tt cat 512/1/00000000000000000004.run
   • Running cat with files: [512/1/00000000000000000004.run]

• Result of cat: the file "512/1/00000000000000000004.run" is processed below •
---
HEADER:
  lsn: 6
  type: REPLACE
BODY:
  tuple: [1, 1]
---
HEADER:
  lsn: 4
  type: INSERT
BODY:
  tuple: [2, 1]
---
HEADER:
  type: ROWINDEX
BODY:
  data: !!binary AAAAAAAAAAo=
...

Tuple {1, 2} wasn't purged from the secondary index run file although it was deleted from the primary index.

Expected behavior

count and len should be the same.

Notes

• The issue doesn't seem to be critical because it doesn't result in invalid select results. The only side-effect is piling up garbage in secondary indexes.

• The issue results in vinyl-luatest/select_consistency_test.lua failures. To increase the probability of a failure, apply the patch below to the test.

  test patch

  diff --git a/test/vinyl-luatest/select_consistency_test.lua b/test/vinyl-luatest/select_consistency_test.lua
  index 22feac904a6f..f5ae01aad852 100644
  --- a/test/vinyl-luatest/select_consistency_test.lua
  +++ b/test/vinyl-luatest/select_consistency_test.lua
  @@ -73,11 +73,11 @@ g.test_select_consistency = function(cg)
               local WRITE_FIBERS = 20
               local READ_FIBERS = 5
               local MAX_TX_STMTS = 10
  -            local MAX_KEY = 100 * 1000
  +            local MAX_KEY = 1000
               local MAX_VAL = 1000
               local MAX_MULTIKEY_COUNT = 10
               local PADDING_SIZE = 100
  -            local DUMP_COUNT = 5
  +            local DUMP_COUNT = 1
   
               local s = box.space.test
   

  test failure report

  Tarantool version is 3.3.0-entrypoint-199-g5325abd3441e
  Started on Fri Nov 15 16:36:58 2024
      vinyl.select_consistency.defer_deletes:false.test_select_consistency ... (2.929s) fail
  ...tarantool/test/vinyl-luatest/select_consistency_test.lua:241: expected: 658, actual: 666
      vinyl.select_consistency.defer_deletes:true.test_select_consistency ... (2.917s) fail
  ...tarantool/test/vinyl-luatest/select_consistency_test.lua:241: expected: 671, actual: 672
  =========================================================
  
  Failed tests:
  -------------
  
  1) vinyl.select_consistency.defer_deletes:false.test_select_consistency
  ...tarantool/test/vinyl-luatest/select_consistency_test.lua:241: expected: 658, actual: 666
  stack traceback:
          ...tarantool/test/vinyl-luatest/select_consistency_test.lua:241: in function <...tarantool/test/vinyl-luatest/select_consistency_test.lua:217>
          ...tarantool/test/vinyl-luatest/select_consistency_test.lua:217: in function 'vinyl.select_consistency.defer_deletes:false.test_select_consistency'
  artifacts:
          master -> /tmp/t/artifacts/master-DlSbSoaS-fdB
  2) vinyl.select_consistency.defer_deletes:true.test_select_consistency
  ...tarantool/test/vinyl-luatest/select_consistency_test.lua:241: expected: 671, actual: 672
  stack traceback:
          ...tarantool/test/vinyl-luatest/select_consistency_test.lua:241: in function <...tarantool/test/vinyl-luatest/select_consistency_test.lua:217>
          ...tarantool/test/vinyl-luatest/select_consistency_test.lua:217: in function 'vinyl.select_consistency.defer_deletes:true.test_select_consistency'
  artifacts:
          master -> /tmp/t/artifacts/master-jg7jA8UlxS33
  Ran 2 tests in 6.762 seconds, 0 succeeded, 2 failed
  
  =========================================================
  Failed tests:
  
  vinyl.select_consistency.defer_deletes:false.test_select_consistency
  vinyl.select_consistency.defer_deletes:true.test_select_consistency