Vinod Kone created MESOS-940:
--------------------------------

             Summary: Slave should checkpoint bootid after recovery instead of 
after registration
                 Key: MESOS-940
                 URL: https://issues.apache.org/jira/browse/MESOS-940
             Project: Mesos
          Issue Type: Bug
    Affects Versions: 0.16.0
            Reporter: Vinod Kone
            Assignee: Vinod Kone
             Fix For: 0.16.0


Doing it after registration means that the slave could keep failing to recover 
in certain scenarios.

Example scenario:

--> A pre 0.16.0 slave was upgraded to 0.16.0
--> After a slave roll it *re-registered* with the master and hence never wrote 
the boot id
--> Now if machine reboots and slave info is incompatible the slave fails 
immediately during recovery.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Reply via email to