Add a new Compute node

Latest response

Hi ,

Using the packstack and answer file I installed two nodes - control and compute ones .

Now I want to add a new compute nodes .

Would it be enough to edit the answer file - remove existing compute node and add a new one ? How the control node will be impacted in this case ?

Regards ,

Vladimir Berezovski

Responses

Two options exist, hand install nova or modify packstack's answer file. The manual install isn't very complicated.

In broad terms, you would want to install the Nova packages on the new compute node, copy the /etc/nova/nova.conf from an existing node with the appropriate local mods, then start the service. nova-manage should see the new node.

Just modifying the packstack answer file you'd want to make sure you disable all the other modules, including Cinder install, otherwise you could wind up trying to recreate services located other places.

In both cases, you are looking for local options to modify (IP address, NIC names, etc).

The easiest way is to modify your answer file and add details of the new node to "CONFIG_NOVA_COMPUTE_HOSTS=". Then configure "EXCLUDE_SERVERS=" with the ip of all other nodes which are already installed. This will make sure that when you rerun packstack it will touch only the new compute node and no system in the EXCLUDE_SERVERS= will be touched.

See details at https://review.openstack.org/#/c/35913/

If I go with packstack scenario - regarding existing and new compute nodes it's clear .What about the control node - if it will remain intact and bew packstack rerun will be touch only new compute node(s) ?

If you configure details of controller server in "EXCLUDE_SERVERS=" before running packstack, it will not touch controller server. Only the new compute node will be provisioned.

If you don't configure details of controller server in "EXCLUDE_SERVERS=", I think it will re-apply puppet modules which is harm less unless you have done manual configuration changes which may disappear.