Clientadmin timeout

I have a cluster with three brokers version 2.8.0.
every ~7 days the “admin client” gets a kind of stacked and when I try to create(or modify) a topic I get a timeout exception then if I try to delete it I get notExistsExceptin. if I check via the Kafka tool I can see the topic only in the zookeeper.
if I try to do the process via the command line I get the same behavior.
cannot find mutch logs about it.
is anybody knows what’s going on? workaround?

to overcome it, I need to stop all brokers at the same time(very painful) and start them one by one.