Hi Marko, On 2025-01-07 12:22, Marko Mäkelä wrote:
No history length increase issue on 10.6.13 with 4GB log file. BUT quite bad behaviour on 10.6.13 with 256MB log file, i.e. some history length caused by cleanup process still remained at the beginning, and increased history length remained even after everything finished. Also 1-2 seconds stall happened several times. (see the attached graphs)
On 10.5.20, I had no such issue on both 4GB log file and 256MB log file, and its behaviour was quite 10.4.34.
Can you (or anyone) post a test script as well as some details of the test environment that is reproducing this regression? I would need to reproduce this regression in order to analyze and fix it.
I could reproduce 'history list increase' issue with the attached schema and python script, that executes 50% R-C DML and 50% R-R DQL using 64 threads. On 10.5.20, history length at the end of each iteration is quite stable, like < 50. On 10.6.13, history length keeps increasing. My configuration is not that special. innodb_flush_method = O_DIRECT innodb_change_buffering = none innodb_buffer_pool_size = 17179869184 innodb_buffer_pool_instances = 64 Kazuhiko