CephFS Backend for Hadoop

[Patrizio Bassi]

2017-07-26 9:17 GMT+02:00 Mark Shuttleworth <mark at ubuntu.com>:

> On 26/07/17 07:14, Patrizio Bassi wrote:
>
> Deploying hadoop via juju in an openstack tenant requires a separate model
> (as far as i could design it).
> So we may use the new juju 2.2 cross model relation to relate the hadoop
> charms to the openstack ceph units.
>
> does it sound feasible?
>
>
> Yes, that sounds feasible. I'm not sure how Ceph identity / permissions
> will work in that case (i.e. who has access to which data, how Ceph will
> correlate tenants in OpenStack both through Cinder and through a direct
> relationship). In principle though, as long as the networking is arranged
> so that IP addresses and routes enable traffic to flow between your tenant
> network and your Ceph network, and as long as both sets of machines can see
> the Juju controller, they can exchange messages and traffic.
>
> Mark
>


Dear Mark,

On relation join event we may create a new ceph storage pool dedicated to
the incoming unit (i.e. prefixed with the controller/model/unit/charm name
by default). Can cephx proto

Regarding networking openstack neutron by default block traffic from tenant
VM to the admin network which it required to access the same ceph mon/osd.
It requires changing neutron or implement an external nat for instance (our
solution at the moment)

Patrizio
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/juju/attachments/20170727/aecabffa/attachment.html>