HashiStack

Void's fleet and desire for services makes it impractical to ask every contributor to understand how the Ansible playbooks are laid out and how the fleet is architected. It is also brittle from a services perspective to host important things in only one machine with no fallback capability. To solve these problems Void uses the Hashicorp stack to enable dynamic workloads scheduled across the fleet, and to enable updates to be decoupled from package updates leading to a more stable infrastructure ecosystem.

More information about the individual services can be found on their specific pages. The remainder of this document speaks in general terms about the architecture of Void's cluster.

The Global Namespace

Void runs a single large global namespace, rather than segmenting our fleet into "datacenters". This is largely a result of the small number of machines we have in each region, but also reflects that Void's fleet is viewed as a single large pool of computing power, rather than segmented computing power.

Presently, the control plane is located in the US, and the larger build machines are located in Hetzner datacenters in the EU. Small service machines are colocated in the US with the control plane. Where possible traffic egresses at the nearest edge to the source data to avoid our traffic needing to transit multiple regions.

The Control Plane

The control plane is composed of the Consul, Nomad, and Vault servers which are colocated on a set of machines hosted in the DigitalOcean SFO3 region. These machines are the point from which the fleet is commanded, and they operate as a highly available trio to ensure fault tolerance.