Hello! I have reviewed all available documentation several times, but am running into difficulty understanding the intended MaaS setup for my configuration.
I have two datacenter deployments in separate countries. I have two management servers, one in each datacenter - each server should be responsible for all machines in that datacenter.
I have one MaaS instance (region+rack) in one datacenter, which works well for managing all machines there.
When adding the second datacenter deployment, I initially installed MaaS as a region+rack controller, connecting to the same Postgresql database and using the same MaaS URL. This resulted in having two region+rack controllers.
After reviewing the documentation, I uninstalled the second region+rack controller and reinstalled it as a single rack controller, connected to my main MaaS instance. In the logs of the second MaaS controller, I can see that it’s receiving requests to check the BMC status of machines in datacenter 1 - despite being in a separate AZ, fabric, etc.
My question is: what is the intended configuration for MaaS to control servers in several (two+) datacenters? Ideally, I have a single dashboard and API URL, utilizing AZs and fabrics to differentiate between different datacenters.
There seems to be no documentation on multi-region deployments (only on HA deployments, which this is not). Has this setup been considered in MaaS?
The current status of my setup leads me to believe that the intended setup is to have two entirely separate MaaS instances in the two datacenters. I must be missing something!
Any insight or advice is very welcome, and I welcome further questions about my setup. Thanks in advance.
A single MAAS installation is supposed to manage only a specific location, for this reason you won’t find any official documentation for multi regions setup.
I know there are people using the same MAAS for multiple locations. They might be using a VPN with proper networking setup for example but I can’t tell exactly what they did and what challenges they might be facing
I too am interested in seeing how this distributed setup is supposed to be done. Not finding much in the way of existing documentation.
I wonder if the term “region” here is quite overloaded or can have various interpretations of what it means. In some cases of hybrid cloud - “region” can mean a single logical collection several “datacenters”, i.e. physically separate facilities, that are regionally proximate, and share very good, private connectivity (bandwidth, redundancy).
The “rack” could be a physical rack, but could also be something more broad - like a row, room, cage, or some other physical grouping that is typically part of some network domain (collection of vlans, etc).
I would hope that the documentation go into more of these kinds of concerns for those who are looking at MaaS for more than just a homelab or single server room or two.
In my case, we are looking at a hybrid cloud solution. I’m looking at coming up with a solution for managing BM/VM provisioning across potentially dozens of sites, and hundreds of servers.
Since we also use public cloud, I’d love to be able to host the higher layers (managing “control plane” concerns; is that regiond?) on some VM in the cloud, assuming some private connectivity (e.g. site-to-site VPN from site to cloud) to the individual sites, where we may have one or more “rackd” controllers running managing the “data plane” concerns.
The real reason here is to hopefully have some unified/centralized view of the multiple sites. Is that even achievable, or is it completely intended that each location be a separate “instance” (i.e. regiond + rackd), that are independently managed & operated?
Thank you for sharing your use case. A single MAAS installation is composed by one or more regions and one or more racks. You can use multiple regions and racks to achieve HA and scalability.
A single MAAS installation is supposed to manage only a single datacenter. There are some people that are using the regions as a central control plane for multiple datacenters where they deploy the racks, but this is not the intended usage and it’s definitely not supported
I think this highlights a significant shortfall in the design of MaaS, and is holding it back from serving larger deployments. A lack of ability to unify several datacenter locations under a single API means that MaaS cannot suitably be utilised in anything more than a small single region deployment.
We are using MaaS to manage our hardware which will span many datacenter locations around the world. Requiring a new MaaS instance for each location makes automation and management very tricky, and makes it impossible to easily manage all machines under a unified dashboard.
How could I make a feature suggestion to consider adding multi-datacenter support to MaaS? Just a way to link together several MaaS instances under a unified API and dashboard would be good enough to suit this use case.
