Neil Conway created MESOS-6232:

             Summary: "Unreachable" metric incremented for non-partition aware 
tasks during framework failover
                 Key: MESOS-6232
             Project: Mesos
          Issue Type: Bug
          Components: master
            Reporter: Neil Conway
            Assignee: Neil Conway

Suppose a task is started by a non-partition aware framework. Then the master 
fails over and the framework fails to reregister with the new master; then the 
agent where the task is running is partitioned. Because we don't have the 
{{FrameworkInfo}} for the task's framework, we can't determine whether it is 
{{PARTITION_AWARE}} or not, so we can't determine whether to increment the 
{{tasks_lost}} or {{tasks_unreachable}} master metric.

This message was sent by Atlassian JIRA

Reply via email to