The RQG testing with thread-handling=pool-of-threads on
origin/10.6 04477bd9364e7a753918c730b89e7012b1926e6a 2023-11-08T12:00:29+01:00
showed no issues related to the use of thread-handling=pool-of-threads.
System variable settings tried

* --mysqld=--thread-handling=pool-of-threads --mysqld=--thread_pool_size=1

* --mysqld=--thread-handling=pool-of-threads --mysqld=--thread_pool_max_threads=1

* --mysqld=--thread-handling=pool-of-threads --mysqld=--thread_pool_dedicated_listener=1

* --mysqld=--thread-handling=pool-of-threads

There were initially many tests which failed because of
1. failures within the RQG code used for detection of deadlocks/hangs of the DB server
This was mostly fixed by corresponding modifications within the RQG code.
The left over cases happen without setting thread-handling=pool-of-threads too.
Some of them might be real server hangs, other might be false alarm by RQG.
2. In case the DB server runs under the tool "rr" than it happens frequent that certain sessions
get starting at some point of time somehow no attention by the DB server at all and similar.
Caused by that RQG assumes to have met a server hang.
This seems to be some incompatibility between how the tool "rr" acts and the requirements
of the feature "Pool of threads" regarding the OS.
Nearly all of these cases were no more replayed when not using thread-handling=pool-of-threads.