Controller high availability

billwear · 8 February 2023 23:56

High availability is built into MAAS: region and rack controllers balance the load and execute failover as part of normal operations. This article will help you understand how to take advantage of these built-in features.

How to put MAAS in HA mode

You only need to install multiple rack controllers to achieve real high availability. Once that’s done, you automatically gain highly-available BMC control and highly-available DHCP. MAAS is constantly trying to answer three questions:

How many racks is each region managing?
How many connections does a given rack controller have?
How many regions – and region “worker processes” are running right now?

With just one rack, most of the logic can’t function. On the other hand, when you have multiple racks (and especially multiple regions), MAAS will continuously balance the load.

Every time a rack controller connects to a region controller to do something, MAAS checks whether racks and regions are balanced. If the ratio for one rack-region connection is above a moderate threshold, compared to other connections, MAAS will re-balance. This activity includes balancing not only discrete region controllers, but also re-distributing connections so that no single worker process has an uneven share of the load.

Rebalancing is also done at various other opportune times. For example, if a network change happens (like toggling DHCP or changing a VLAN), MAAS will also re-balance the load. And MAAS can maintain primary and secondary rack designations, so that faster, more nuanced load-balancing can occur.

How to enable highly-available BMC

You can also enable HA for BMC control (node power cycling) just by adding a second rack controller. MAAS will automatically identify which rack controller is responsible for a BMC, continuously balancing the connections.

How to enable highly-available DHCP services

You can enable highly-availalbe DHCP services by using MAAS-managed DHCP, and adding rack controllers. This DHCP HA affects the way MAAS manages nodes, including enlistment, commissioning and deployment. It enables primary and secondary DHCP instances to serve the same VLAN. This VLAN replicates all lease information is between rack controllers, so there’s a bit of performance boost for large networks.

MAAS DHCP automatically creates failover peers, using mostly standard parameters:

failover peer "failover-partner" {
     primary;
     address dhcp-primary.example.com;
     peer address dhcp-secondary.example.com;
     max-response-delay 60;
     max-unacked-updates 10;
     mclt 3600;
     split 255;
     load balance max seconds 3;
}
failover peer "failover-partner" {
     secondary;
     address dhcp-secondary.example.com;
     peer address dhcp-primary.example.com;
     max-response-delay 60;
     max-unacked-updates 10;
     load balance max seconds 3;
}

Note that the only difference from a standard 50/50 split (split 128) is that the primary DHCP server answers any requests that it can (split 255), within the maximum response delay of 60 seconds and an unacknowledged update count of 10. In this sense, highly-available MAAS DHCP fails over only when absolutely necessary.

If you are enabling DHCP for the first time after adding a second rack controller, please read Enabling DHCP. On the other hand, if you have already enabled DHCP on your initial rack controller, you’ll need to reconfigure DHCP to get optimum results.

To reconfigure DHCP after adding a new rack controller:

Select Subnets.
Select the appropriate VLAN.
Choose Reconfigure DHCP.
Confirm that you can see the second rack controller under Secondary controller.
Select Reconfigure DHCP.

To reconfigure DHCP after adding a new rack controller, use the following sequence of commands:

vid=$(maas maas subnets read | jq -r '.[] | select(.cidr == "10.0.0.0/24") | .vlan.vid')
fabric_name=$(maas maas subnets read | jq -r '.[] | select(.cidr == "10.0.0.0/24") | .vlan.fabric')
query=".[] | select(.name == \"$fabric_name\") | .id"
fabric_id=$(maas maas fabrics read | jq "$query")
maas maas ipranges create type=reserved start_ip=10.0.0.3 end_ip=10.0.0.49
maas maas ipranges create type=dynamic start_ip=10.0.0.50 end_ip=10.0.0.99
maas maas vlan update ${fabric_id} ${vid} primary_rack=$(hostname) dhcp_on=true

Be sure to substitute the sample values for those of your own environment.

How to configure multiple region endpoints

MAAS will automatically discover and track all reachable region controllers in a single cluster of rack controllers It will also attempt to automatically connect to them if the one in use becomes inaccessible. Administrators can alternatively specify multiple region-controller endpoints for a single rack controller by adding entries to /var/snap/maas/current/rackd.conf. For example:

MAAS will automatically discover and track all reachable region controllers in a single cluster of rack controllers It will also attempt to automatically connect to them if the one in use becomes inaccessible. Administrators can alternatively specify multiple region-controller endpoints for a single rack controller by adding entries to /etc/maas/rackd.conf. For example:

.
.
.
maas_url:
  - http://<ip 1>:<port>/MAAS/
  - http://<ip 2>:<port>/MAAS/
.
.
.

The setup of highly-available DHCP is now complete. Note that, for HA purposes, DHCP provisioning will take into account multiple DNS services when there is more than one region controller on a single region.

How to make region controllers highly available

Implementing highly-available region control is possible when you learn:

How to enable highly-available PostgreSQL
How to enable highly-available API services
How to set up load balancing with HAProxy (optional)

Load balancing is optional, but is highly recommended.

How to enable highly-available PostgreSQL

MAAS stores all state information in the PostgreSQL database. It is therefore recommended to run it in HA mode. Configuring HA for PostgreSQL is external to MAAS. You will, therefore, need to study the PostgreSQL documentation↗ and implement the variant of HA that makes you feel most comfortable.

Each region controller uses up to 40 connections to PostgreSQL in high load situations. Running two region controllers requires no modifications to the max_connections in postgresql.conf. More than two region controllers require that max_connections be adjusted to add 40 more connections per added region controller.