We have a simple mariadb replication with a master and a bunch of slaves. without any prior alert or warning, randomly in our slaves some indexes ( some of them are not present in master) become corrupted. This happened for us after migrating from we receive a bunch of error log each time.
it happened for us several times we upgraded one of our slaves to mariadb 10.6 but it happened again.
mariadb hung when it want to insert (or update ) a row because it cannot renew its index Btree I think. and the major error is because of SEMAPHOR timeout.
I have attached our error logs.
some times it says you hit a bug, some times it says the index is corupted.
we have enough resources ( more than 180 G of ram and SSD Disk and 56 core CPU )
the command that we used for creating the indexes is :
alter table foo add index bar_idx(bar), algorithm =inplace, lock=none;