High cpu and memory shortage issues in mirror container when space is configured with drop-oldest policy for mirror and distributed transaction are used when redolog reach its capacity.
Taking several heap dumps its shown that the instances of MultiSourceSingleFileReliableAsyncTargetProcessLog takes more and more memory.
(wait time was configured to 5 seconds, and operations threshold to 1000, but in heap itself more packets are accumulated)
current work around:
configure the space with global order replication
e.g <prop key="cluster-config.groups.group.repl-policy.processing-type">global-order</prop>
Attaching 2 heap dumps of the mirror