[MariaDB developers] Re: [PATCH 0/2] Suggestion for smaller fix to XA parallel replication performance, MDEV-31949

Andrei Elkin <andrei.elkin@mariadb.com> writes:

In normal cases, like that of not so many duplicate xids, the size of maybe_active_xid must remain constrained because of "continuous" relocation of elements from the tail to replace garbage (normally found in head indexes).

Nevertheless as the array can be dynamically re-allocated to reflect some some spike in the load, I suggest in

origin/review__knielsen_xa_sched_minimal_fix

to take care of that.

I think your suggestion in review__knielsen_xa_sched_minimal_fix is fine. I'm not sure exactly how large the array can grow in pathological cases. As you wrote, in most normal cases we would expect it to not grow that much. But shrinking it if necessary surely cannot hurt. Thanks, - Kristian.