[
https://issues.apache.org/jira/browse/MESOS-3177?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14791639#comment-14791639
]
Yong Qiao Wang commented on MESOS-3177:
---------------------------------------
Thanks [~cmaloney] for your kindly reply. I have some questions and comments
for your above thinks:
1. As we know, currently roles and weights are not persist in the replicated
log, do you mean that we should persist them?
2. If yes for #1, then I think the initial replicated log for roles and weights
are created when Mesos master starts in the first time, and the content of the
log should be the roles and weights specified by --roles and --weights flag. is
it right?
3. For add a new role "add_role", in code level, there are two places only need
to change:
Add a new HTTP endpoint in master.cpp to add a new item in
{code}
hashmap<std::string, Role*> roles;
{code}
and call allocator to update the RoleSorter;
4. For remove an existing role "remove_role", I think it should ensure the
following things before remove an existing role:
- Kill all tasks which using the resources reserved by this role;
- Shutdown all executors which using the resources reserved by this role;
- Unreserve the dynamically reserved resources for this role;
- Destory the persisted volumn which using the resources reserved by this
role;
- Remove all frameworks which associated with this role?
- Remove the related ACL of this role;
5. Do you mean the authorization rather than authentication in above comments?
[~cmaloney], Welcome your any comments for above thinks of me.
> 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: Thomas Rampelberg
> 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)
