question about updating to mdbx 0.39.10 (0.13.8)

[JkLondon]

libmdbx released new version (0.13.8) some time ago
but there are some issues with growing inmem db size.
PR with this db #17284
Also have done some measurements of InMem DB:
Average size: 0.48 MB mdbx-go 0.39.10 (libmdbx 0.13.8)
Average size: 0.25 MB mdbx-go 0.39.9 (libmdbx 0.13.7)

Also in closed PR I've introduced testState memfix (4GB db when test is running -> 4MB)

Moreover
SaveNoSync now causing some issues

./build/bin/mdbx_chk -vvv ~/data/chiado32_full/nodes/eth69
mdbx_chk v0.13.7-0-g566b0f93 (2025-07-30T11:44:04+03:00, T-7777cbdf5aa4c1ce85ff902a4c3e6170edd42495)
Running for /.../chiado32_full/nodes/eth69 in 'read-only' mode with verbosity level 3 (of 0..9)...
 ~ meta[2] used-bytes (1228800) beyond filesize (1097728), skip it
 ~ meta[2] used-bytes (1228800) beyond filesize (1097728), skip it
 ~ filesize mismatch (expect 16777216b/1024p, have 1097728b/67p)
 ! filesize should be rounded-up to system allocation granularity 65536
 ! error -30419 (MDBX_WANNA_RECOVERY: Database should be recovered, but this could NOT be done automatically for now since it opened in read-only mode)
 ! mdbx_env_open() failed, error -30419, MDBX_WANNA_RECOVERY: Database should be recovered, but this could NOT be done automatically for now since it opened in read-only mode
Please run ./build/bin/mdbx_chk in the read-write mode (with '-w' option).

That was fixed in 0.13.8
erthink/libmdbx@43cb913

bench of mdbx-go (old - 0.13.7 new 0.13.8)

➜  mdbx git:(v0.39.9) ✗ benchstat bench-old.txt bench-new.txt 
goos: darwin
goarch: arm64
pkg: github.com/erigontech/mdbx-go/mdbx
cpu: Apple M3 Max
                         │ bench-old.txt │            bench-new.txt            │
                         │    sec/op     │   sec/op     vs base                │
Cursor-16                   106.4n ±  2%   112.3n ± 5%   +5.59% (p=0.001 n=10)
Cursor_Renew/1-16           32.72n ±  8%   32.72n ± 0%        ~ (p=0.985 n=10)
Cursor_Renew/2-16           32.29n ±  1%   32.53n ± 0%   +0.74% (p=0.022 n=10)
Cursor_Renew/3-16           115.8n ±  1%   110.2n ± 3%   -4.79% (p=0.000 n=10)
Cursor_Renew/4-16           40.23n ±  2%   39.37n ± 0%   -2.14% (p=0.000 n=10)
Cursor_Set_OneKey-16        63.18n ±  1%   62.80n ± 0%   -0.59% (p=0.002 n=10)
Cursor_Set_Sequence-16      112.5n ±  1%   110.1n ± 0%   -2.05% (p=0.000 n=10)
Cursor_Set_Random-16        555.1n ± 13%   483.0n ± 3%  -12.97% (p=0.000 n=10)
Errno_Error-16              209.0n ±  2%   192.7n ± 0%   -7.80% (p=0.000 n=10)
Txn_abort-16                172.8n ±  1%   172.3n ± 1%        ~ (p=0.404 n=10)
Txn_commit-16               50.31µ ±  6%   47.54µ ± 4%   -5.51% (p=0.011 n=10)
Txn_ro-16                   205.5n ±  3%   198.6n ± 4%   -3.33% (p=0.037 n=10)
Txn_unmanaged_abort-16      168.7n ±  2%   164.1n ± 2%   -2.67% (p=0.006 n=10)
Txn_unmanaged_commit-16     170.2n ±  1%   163.9n ± 4%   -3.70% (p=0.001 n=10)
Txn_unmanaged_ro-16         153.8n ±  2%   149.8n ± 1%   -2.63% (p=0.000 n=10)
Txn_renew-16                85.92n ±  1%   85.41n ± 0%   -0.59% (p=0.019 n=10)
Txn_Put_append-16           183.6n ±  1%   181.5n ± 0%   -1.14% (p=0.002 n=10)
Txn_Put_append_noflag-16    217.6n ±  1%   212.8n ± 1%   -2.23% (p=0.001 n=10)
Txn_Get_OneKey-16           62.17n ±  3%   58.64n ± 1%   -5.68% (p=0.000 n=10)
Txn_Get_Sequence-16         148.0n ±  2%   147.3n ± 1%   -0.47% (p=0.013 n=10)
Txn_Get_Random-16           480.5n ±  9%   463.9n ± 2%   -3.45% (p=0.023 n=10)
geomean                     167.7n         163.1n        -2.72%

                         │ bench-old.txt │            bench-new.txt            │
                         │     B/op      │    B/op     vs base                 │
Cursor-16                   16.00 ± 0%     16.00 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/1-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/2-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/3-16           16.00 ± 0%     16.00 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/4-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Set_OneKey-16        0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Set_Sequence-16      0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Set_Random-16        0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Errno_Error-16              320.0 ± 0%     320.0 ± 0%       ~ (p=1.000 n=10) ¹
Txn_abort-16                80.00 ± 0%     80.00 ± 0%       ~ (p=1.000 n=10) ¹
Txn_commit-16               248.0 ± 0%     248.0 ± 0%       ~ (p=1.000 n=10) ¹
Txn_ro-16                   240.0 ± 0%     240.0 ± 0%       ~ (p=1.000 n=10) ¹
Txn_unmanaged_abort-16      80.00 ± 0%     80.00 ± 0%       ~ (p=1.000 n=10) ¹
Txn_unmanaged_commit-16     80.00 ± 0%     80.00 ± 0%       ~ (p=1.000 n=10) ¹
Txn_unmanaged_ro-16         80.00 ± 0%     80.00 ± 0%       ~ (p=1.000 n=10) ¹
Txn_renew-16                0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Put_append-16           8.000 ± 0%     8.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Put_append_noflag-16    0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Get_OneKey-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Get_Sequence-16         0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Get_Random-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
geomean                                ²               +0.00%                ²
¹ all samples are equal
² summaries must be >0 to compute geomean

                         │ bench-old.txt │            bench-new.txt            │
                         │   allocs/op   │ allocs/op   vs base                 │
Cursor-16                   1.000 ± 0%     1.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/1-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/2-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/3-16           1.000 ± 0%     1.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Renew/4-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Set_OneKey-16        0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Set_Sequence-16      0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Cursor_Set_Random-16        0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Errno_Error-16              6.000 ± 0%     6.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_abort-16                1.000 ± 0%     1.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_commit-16               3.000 ± 0%     3.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_ro-16                   2.000 ± 0%     2.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_unmanaged_abort-16      1.000 ± 0%     1.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_unmanaged_commit-16     1.000 ± 0%     1.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_unmanaged_ro-16         1.000 ± 0%     1.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_renew-16                0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Put_append-16           1.000 ± 0%     1.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Put_append_noflag-16    0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Get_OneKey-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Get_Sequence-16         0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Txn_Get_Random-16           0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
geomean                                ²               +0.00%                ²
¹ all samples are equal
² summaries must be >0 to compute geomean

TODO:

• bench with @sudeepdino008 tool
• check sizes of synced node in diff regimes

FYI @AskAlexSharov

[JkLondon]

After small research and the fact that it would only increase size of inmem dbs in our erigon (which we don't have many)
have decided that we should upgrade mdbx to next version

[AskAlexSharov]

@JkLondon merged PR about new mdbx version. But plz help me to investigate this one #17405 - need to figure out why benchmarks are unstable (degrading) and then set dbSyncBytesThreshold to ~10mb (choose by bench). And release #17405 to main

[VBulikov]

VBulikov added Imp2 as Importance

[AskAlexSharov]

released 0.13.7 with SafeNoSync assert fix - it's enough for now. closing.