how to auto launch new task instance when mesos-slave stopped?

问题

1. • version Info and command line args

     mesos-master & mesos-slave version 1.1.0
     marathon version 1.4.3
     docker server version 1.28
     

   • mesos-master's command line args:

     --zk=zk://ip1:2181,ip2:2181,ip3:2181/mesos \
     --port=5050 \
     --log_dir=/var/log/mesos \
     --hostname=ip1 \
     --quorum=2 \
     --work_dir=/var/lib/mesosmaster
     

   • mesos-slave's command line args:

     --master=zk://ip1:2181,ip2:2181,ip3:2181/mesos \
     --log_dir=/var/log/mesos --containerizers=docker,mesos \
     --executor_registration_timeout=10mins --hostname=ip1 \
     --recovery_timeout=1mins \
     --resources=ports:[25000-65000] \
     --work_dir=/var/lib/mesos  
     

2. operation

   1. In marathon web UI,run a app via docker image,and task state is "Unknown". (Because I doesn't add health check)
      

   2. reboot the machine which runs the task.
   3. Now,the expected result what I thought is that the task will be killed and marathon will create a new task,but now (see below picture)
      

The State of Task became "Unscheduled". the the task can not be killed only after about 15mins. Finally,what I want is the new task will auto launch and the old task auto be killed ( or removed or expunged)

回答1:

It seems like you need to configure unreachableStrategy for your tasks. The guide here explains it. I will be playing with it soon so will post an example here as well.

来源：https://stackoverflow.com/questions/44113232/how-to-auto-launch-new-task-instance-when-mesos-slave-stopped