Hey Fred, Logs (master and slave) can be helpful to sched some light on the problem. On 2 Dec 2015 3:01 pm, "Frederic LE BRIS" <[email protected]> wrote:
> Hi, > > I manage a Mesos Cluster 0.23.0 based on .deb from Mesosphere on Ubuntu > 14.04. > > We deployed 3 zookeeper, 3 Mesos-master, and 3 Marathon : HA Mode > > And deployed 6 mesos slaves + slave process on 3 masters mesos. > > So I have the following topology: > > 3 servers : Mesos-master / Marathon > 3 servers : Zookeeper / mesos-slaves. > 3 servers : mesos-slave > > I follow the HA configuration for Mesos-master and marathon. > > The point is when I kill the leader mesos-master, we lost the existing > task on the slave, and the ressources available are lock by the slave even > if the master see no activity on this slave. > > My mesos cluster is in production, so I’m not able te restart from > scratch, so I look for a procedure te re-synchronise the cluster. > > And some way to check that my leaders mesos master are working together as > a leader and two slaves correctly synchronize. > > I guess I miss something, but I’m need some help... > > Regards, > > Fred > > >
