[
https://issues.apache.org/jira/browse/MESOS-3177?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14934901#comment-14934901
]
Adam B commented on MESOS-3177:
-------------------------------
The tricky part with your use case description, [~air], is that a framework
registers with a single role, so you're either asking if
a) a framework can register with a new, unrecognized role
b) a framework can change its role after registration
c) a framework can create an additional role and use it.
And none of those are going to be recommended/possible after just this ticket.
a) Since the total number of roles affects each role's fair share of the
cluster, it should be up to the cluster admin to add new roles, not the
frameworks. We could add a flag/ACL to allow frameworks to register with
unrecognized roles, but it would be disabled by default. (I can also imagine
typos wreaking havoc.)
b) Even with the progress we've already made on updating the simpler fields in
FrameworkInfo upon reregistration, updating a framework's role is the most
complex. See
https://cwiki.apache.org/confluence/display/MESOS/Design+doc:+Updating+Framework+Info
c) Even assuming that we've enabled an ACL to allow (specific) frameworks to
create new roles, we still have a long way to go before we can support
multi-role frameworks. See MESOS-1763
> Make Mesos own configuration of roles/weights
> ---------------------------------------------
>
> Key: MESOS-3177
> URL: https://issues.apache.org/jira/browse/MESOS-3177
> Project: Mesos
> Issue Type: Improvement
> Components: master, slave
> Reporter: Cody Maloney
> Assignee: Yong Qiao Wang
> Labels: mesosphere
>
> All roles and weights must currently be specified up-front when starting
> Mesos masters currently. In addition, they should be consistent on every
> master, otherwise unexpected behavior could occur (You can have them be
> inconsistent for some upgrade paths / changing the set).
> This makes it hard to introduce new groups of machines under new roles
> dynamically (Have to generate a new master configuration, deploy that, before
> we can connect slaves with a new role to the cluster).
> Ideally an administrator can manually add / remove / edit roles and have the
> settings replicated / passed to all masters in the cluster by Mesos.
> Effectively Mesos takes ownership of the setting, rather than requiring it to
> be done externally.
> In addition, if a new slave joins the cluster with an unexpected / new role
> that should just work, making it much easier to introduce machines with new
> roles. (Policy around whether or not a slave can cause creation of a new
> role, a given slave can register with a given role, etc. is out of scope, and
> would be controls in the general registration process).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
