[
https://issues.apache.org/jira/browse/MESOS-3066?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14637173#comment-14637173
]
Joseph Wu commented on MESOS-3066:
----------------------------------
The object would be used to communicate between operator and the master
(through the HTTP API). The master just reads and stores the object (since the
master doesn't perform any of the maintenance mode transitions).
> Replicated registry does not have a representation of maintenance schedules
> ---------------------------------------------------------------------------
>
> Key: MESOS-3066
> URL: https://issues.apache.org/jira/browse/MESOS-3066
> Project: Mesos
> Issue Type: Task
> Components: master, replicated log
> Reporter: Joseph Wu
> Assignee: Joseph Wu
> Labels: mesosphere
>
> In order to persist maintenance schedules across failovers of the master, the
> schedule information must be kept in the replicated registry.
> This means adding an additional key in src/master/registry.proto. The status
> of each individual slave's maintenance will also be persisted in this way.
> {code}
> message Maintenance {
> message HostStatus {
> required string hostname = 1;
> // True if the slave is deactivated for maintenance.
> // False if the slave is draining in preparation for maintenance.
> required bool is_down = 2;
> }
> message Schedule {
> // The set of affected slave(s).
> repeated HostStatus hosts = 1;
> // Interval in which this set of slaves is expected to be down for.
> optional Unavailability interval = 2;
> }
> message Schedules {
> repeated Schedule schedules;
> }
> optional Schedules schedules = 1;
> }
> {code}
> Note: There can be multiple SlaveID's attached to a single hostname.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
