After a loss of connection to zookeeper the mesos marathon framework we run becomes inactive.
We run a 3 node zookeeper 3.4.12 cluster in AWS. When we rollingly replace the nodes of zookeeper this happens. Marathon 1.5.8 (docker) is running together with mesos-master 1.5.0 and Chronos on a different 3 node cluster. The chronos framework is/stays registered after the zookeeper connection loss.
I will attach logs of marathon, zookeeper and mesos.
Can you see what is going wrong here? Why doesn't the marathon framework stay active?