Tag Info

New answers tagged

1

I was recently facing the same question and, aside from the URL quoted above did not find much more. With Maas and Curtin's developer's help, I was able to come up with a working scenario that I described in this blog post. I hope that it can help you out an others facing the same question.


0

Make sure you've set the power management in the MAAS UI so MAAS knows how to power on/off the VM's. See here for more information: How to configure MAAS to be able to boot virtual machines The other problem I see is that you seem to be running the installer as root. As per our documentation you'll want to run the installer as a normal non-root user. More ...


0

This was addressed @ https://github.com/Ubuntu-Solutions-Engineering/openstack-installer/issues/631 Basically boiled down to not running the installer as a non-root user.


1

I posted the solution here: Adding two single Cluster Controller to a MaaS Region Controller


0

I've resolved that upgraded MaaS with its own last release and following this procedure. - On Region Controller run: $: cat /var/lib/maas/secret copy the key - On Cluster Controller run: $: sudo dpkg-reconfigure maas-cluster-controller and adding the key copied from Region Controller, then after few second we'll see the new Cluster Controller ...


0

If using power address 'qemu:///system', then add maas to the libvirtd group. I.e.: # grep ^libvirtd /etc/group libvirtd:x:112:maas # Reboot after applying this change as existing processes won't be in the new extra group.


1

Meanwhile I have a better understanding of how networking is arranged in maas and its pretty cool. So to answer my own question: I recommend against separating IPMI and PXE traffic. Its more efficient to just share the RAC traffic on eth0. All server can boot PXE default on eth0 too. Besides, you don't need an extra ethernet port/switch for just the RAC, ...


1

iptables does not know about interface aliasing and so you would only refer to eth0 and eth1 in any iptables rules. Assuming you have the default policy of ACCEPT for the FORWARD chain, and already have /proc/sys/net/ipv4/ip_forward set to 1, then all you should need is: sudo iptables -t nat -A PREROUTING -i eth0 -d 168.XXX.XXX.72 -j DNAT --to-destination ...


1

Instead of disabling PXE boot altogether, just move the PXE boot a notch lower in the boot priority and then the systems will boot normally and will only start re-install when needed.


0

sudo apt-get purge maas sudo apt-get install maas The above resolved the issue. Maas is working as intended.


0

One thing to keep in mind is that this can be caused by problems with the actual image being deployed; we have seen regressions like this in the daily cloud-images which we make available for testing (for instance, in bug 1391354).


1

Ok, I tried to establish "juju environments" to the node where I have install maas controller node. Then when you try to power juju bootstrap, it randomly chooses one of the node where it gets ssh from maas envrionment that you have already provided. So bootstrap of juju is going to be absolutely another node after you run juju bootstrap then you can add ...


1

Depending on which installer you are using, MAAS provides different mechanisms to customize an installation. This answer assumes you are using the curtin installer (also known as fast-path installer) which became the default in MAAS 1.7. MAAS uses cloud-init as part of its installation process. There are multiple ways to customize cloud-init behaviour, ...


1

You install juju on the machine where you expect to drive the deployments from. It can even be more than one machine, and then you would share the environment file with other people. It's kind of like installing openssh-client on the machines where you want to administer your network from. It's where you run ssh from. Same with juju. You will run the juju ...


2

You are right that you need to "see" the machine in the juju status output before you can use the deploy --to syntax. In your case, what you need to do is "reserve" the machine first via add-machine. If you want to reserve a specific machine, you run the command like this: juju add-machine maas:netNode.maas For any random machine, use: juju add-machine ...


2

At the Vancouver OpenStack Summit, it was indicated the plan is for Kilo to be supported by the Landscape Autopilot in July. Of course, Kilo packages are available today, and you can build an Autopilot-less OpenStack yourself, just without the enhanced user experience of the guided setup.


2

rmustakos is correct and giving good advice, thank you! From my perspective since you have 8 nodes, MAAS is worthwhile. You will need one node for MAAS, and on that node you can create some VMs for the Juju bootstrap node, leaving 7 nodes for cloud or other projects. You will need to add some network cards, it is required to have some nodes with multiple ...



Top 50 recent answers are included