This is my first post, I am brand new to all of this.
My home lab is via a Netgate SG-1100 LAN 192.168.127.0/24 (DHCP), OPT 10.3.0.0/21 (No DHCP)…
The Region+Rack (with test_db) is a Virtualbox instance on OPT, gateway from LAN to OPT works…
The servers are 4 x Proliant G7, configured to 20.04 LXD pods…
Some additional Virtualbox PXE capacity to provide infrastructure pieces…
Not sure what to do, where to start with JuJu, how to use this capacity.
QUESTION: Install and initialize JuJu inside a slice of MAAS, or outside from a Virtualbox VM?
I do not seem to understand how this would work with MAAS.
Thanks for reading!
I work in manufacturing, and although I have no formal IT training I would love to learn this technology.
1 Like
To get started:
- Configure a machine in MAAS with at least one network address in your LAN so that it can reach internet and you can reach it from your local client. This machine will be your juju controller.
- Perform the juju bootstrap in such a way that this machine will be used. (juju help bootstrap).
- Once the controller is up. Configure your other machines as you like in Maas and try to perform ‘juju add-model’ and ‘juju add-machine mymachine’ where mymachine is the instance name of the machines you have in MAAS.
- Try then ‘juju deploy cs:~erik-lonroth/tiny-bash’ that will perhaps work with you pods, but I don’t know exactly how that would work. You will probably discover and learn from the above and feel free to come back with your discovery.
Wielding juju is often a combination of knowing some about your cloud and juju itself to be useful. So don’t get disencouraged by the steep learning curve while learning this if you are new to both juju and maas.
3 Likes
I tried a few things, and it ended by installing Juju on the MAAS VirtualBox with --local, and after connecting to the MAAS and configuring the credentials bootstrapping the controller using an LXD instance inside MAAS.
For those who are considering using MAAS - what I learned is that you must use good, enterprise-grade hardware for servers as MAAS needs full control over cycling power - the Proliant G7s I use work well with Ipmi. Another is isolating a non-DHCP network, and maybe disabling IP6 entirely to simplify your overview and understanding of the subnet. I put myself though a lot of pain trying to deal with a mixed IP4/IP6 network, which I think is unnecessary, IP4 works and sufficient.
On MAAS 2.8 the LXD VM is still in beta. I would also recommend installing MAAS on Bionic not on Focal, as several of my attempts on Focal failed (this simply means I could not make it work). The G7 are deployed to Focal, with NIC1 used for PXE only, the IPMI NIC is configured with static IP, and one NIC is bridged. This is the bridge I used while initializing LXD and made available on the network, instead of allowing LXD to create a local one. Ubuntu is testing LXD with a dir for pool but I am a reckless renegade and configured with zfs.
Ubuntu states MAAS and Juju are made for each other, and I kept an eye on this technology in the last few years. I did not want to use KVM, and with LXD this looks really really attractive to me. The four servers are able to provide about 20-30 LXD instances. Once you do all the right things, in the correct sequence, spinning up a brand new MAAS cloud takes only a few hours!
But again I work on a factory production floor, I do this on my spare time and I get impressed easily 
4 Likes
