[kolla-ansible] Correct way to add/remove nodes.

Eddie Yen missile0407 at gmail.com
Wed Sep 4 01:02:46 UTC 2019


OK, I think I found the probably answer about removing controller.

Refer from Oracle. I know there're different between kollacli and
kolla-ansible, but I think the few mechanism are the same.

To remove the controller that having a problem (the controller already dead
caused by hardware failure for example):
1. Remove the node from inventory.
2. Do kolla-ansible reconfigure. Then check if all information has updated.

To add the new controller, just do the same as addition from previous mail.

For now I have not enough machines to try this, but this is what I approach.
Please correct me if there's something wrong.


Many thanks,
Eddie.

Eddie Yen <missile0407 at gmail.com> 於 2019年9月3日 週二 下午3:51寫道:

> Hi,
>
> I wanna know the correct way to add/remove nodes since I can't find the
> completely document or tutorial about this part.
>
> Here's what I know for now.
>
> For addition:
> 1. Install OS and setting up network on new servers.
> 2. Add new server's information into /etc/hosts and inventory file
> 3. Do bootstrapping to these servers by using bootstrap-servers with
> --limit option
> 4. (For Ceph OSD node) Add disk label to the disks that will become OSD.
> 5. Deploy again.
>
>
> For deletion (Compute):
> 1. Do migration if there're VMs exist on target node.
> 2. Set nova-compute service down on target node. Then remove the service
> from nova cluster.
> 3. Disable all Neutron agents on target node and remove from Neutron
> cluster.
> 4. Using kolla-ansible to stop all containers on target node.
> 5. Cleanup all containers and left settings by using cleanup-containers
> and cleanup-host script.
>
>
> For deletion (Ceph OSD node):
> 1. Remove all OSDs on target node by following Ceph tutorial.
> 2. Using kolla-ansible to stop all containers on target node.
> 3. Cleanup all containers and left settings by using cleanup-containers
> and cleanup-host script.
>
>
>
> Now I'm not sure about Controller if there's one controller down and want
> to add another one into HA cluster. My thought is that add into cluster
> first, then delete the informations about corrupted controller. But I have
> no clue about the details. Only about Ceph controller (mon, rgw, mds,. etc)
>
> Does anyone has experience about this?
>
>
> Many thanks,
> Eddie.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20190904/4e8f943b/attachment.html>


More information about the openstack-discuss mailing list