Ask Your Question
0

Install PackStack in two nodes (Controller and Compute)

asked 2018-09-22 10:43:13 -0500

soheil08 gravatar image

I want to install distributed OpenStack using PackStack. PackStack by default needs 3 nodes, Controller, Network and Compute. I want to merge controller with the network node but I couldn't find any instruction for that. My plan is to create a virtual machine for the Controller and the compute node will be installed on bare metal I tried inserting Controller IP address as Network host in the PackStack answer file, but it was not successful. Should I do anything else for that?

In addition, I want to know is it necessary to have two NIC in the Controller and Compute machine? and if yes which NIC address should I set for Hosts in the PackStack answer file? Is it Ok to the both NIC in machine bind to the same IP range or it should be different?

edit retag flag offensive close merge delete

2 answers

Sort by » oldest newest most voted
0

answered 2018-09-24 11:47:56 -0500

itlinux gravatar image

Are you using Virtual like (VirtualBox, VMware-Fusion etc?) or is this a bm ? Anyhow, the process should be pretty simple.

1) if you are using VMs create a new linux base with 2 nics, one to local only one to nat out. 2) the packstack config will have options for COMPUTE and CONTROLLER, use the ip of the compute there and the controller for the appropriate field. Once that's done you can run packstack with the option --answer-file filename. Make sure you gen the file first. packstack --gen-answer-file myanswerfile.txt for example. Or you can add a date etc so you know when that answer file has been gen. Let me know if you have any questions.

edit flag offensive delete link more
0

answered 2018-09-22 18:16:52 -0500

updated 2018-09-22 19:38:56 -0500

By default, Packstack needs one node with a single NIC. See here.

You can then add a compute node. While the instructions say that you need two NICs, I remember faintly that it was possible with a single NIC. You could also work with VLANs or create extra NICs of the type eth0:0 on the compute host.

In case of several NICs, I would use different IP address ranges for the NICs, to simplify troubleshooting. A single range probably works as well, but I am not sure about the ramifications.

Once you have added your compute node, the first node still has both controller and compute roles. If you don’t want that, you can easily disable nova-compute: Either uninstall it, or use openstack compute service set --disable.

edit flag offensive delete link more

Comments

Thanks, What do you mean by saying Once you have added your compute node, the first node still has both controller and compute roles. I suppose at the first we have at least one controller and one compute node. So what do you mean by saying the first node?

soheil08 gravatar imagesoheil08 ( 2018-09-23 00:06:07 -0500 )edit

Do you know any instruction about how to config PackStack to have two NIC?

soheil08 gravatar imagesoheil08 ( 2018-09-23 00:06:53 -0500 )edit

You normally install Packstack on a single server, which functions as both controller and compute node. Let's call it node1. You then add a compute node. If you don't want the compute service on node1, you can then disable it.

Bernd Bausch gravatar imageBernd Bausch ( 2018-09-23 00:16:49 -0500 )edit

The instructions to add a compute node tell you how to work with two NICs.

Bernd Bausch gravatar imageBernd Bausch ( 2018-09-23 00:17:58 -0500 )edit

I got it, Thanks!

soheil08 gravatar imagesoheil08 ( 2018-09-23 01:14:47 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2018-09-22 10:43:13 -0500

Seen: 262 times

Last updated: Sep 24 '18