Product

Compute nodes

Websites run on compute nodes: servers that ProHostiva's control plane manages, deciding what runs where and watching each node's health.

Deploy your project Check your project for free

Compute nodes Healthy

node-1Healthychecked in 12 s ago
node-2Healthymost free capacity
PlacementNew site → node-2after the safety reserve
ExistingStays where it runsits DNS points there

example

01 / Architecture

One part decides, the other runs.

Keeping the two apart means we can add servers as ProHostiva grows, without changing how you deploy.

Control plane
Decides
Your dashboard, the API, the database and the worker. It decides which node runs each website, queues the work and keeps a record of every deployment.
Compute nodes
Run
Servers that build and run customer containers and the proxy in front of them. Each runs our node agent, which carries out the work it's given and reports back.

02 / Health

Every node checks in every 30 seconds.

Each check-in carries the node's CPU, memory and disk use. The control plane turns that into one of four states.

Healthy: Checking in on time and within its limits. The only state that takes new deployments.
Degraded: A check-in is late, or CPU, memory or disk use is above 85%. Sites keep running; no new deployments.
Offline: No check-in for five minutes. No new deployments until it's back.
Disabled: Taken out of service by us, for example for maintenance.

03 / Placement

Where your website runs.

You don't choose a server, and you don't have to think about one.

A new website goes to the healthy node with the most free capacity. A reserve of memory and disk on every node is never counted as free, so a busy node isn't filled to the brim.
An existing website stays on the node it runs on, because its domain's DNS points there.
If that node can't take a deployment right now, the deploy is refused with a clear reason and your live site keeps running. We never move a site quietly, which would break its domain.

04 / Security

Nodes do only what they're told.

How the control plane and the nodes talk to each other.

Nodes call the control plane over HTTPS; the control plane never connects in to them.
Each node has its own credential, so one can be shut out without touching the others.
Nodes receive specific, structured jobs, such as “build this deployment” or “restart this site”, never arbitrary commands.
No node exposes its Docker socket to the network.

05 / For you

What it means for your site.

Mostly nothing: it all happens behind your dashboard. One practical point: your domain points at the node your site runs on, so use the CNAME record for subdomains such as www when you connect a domain. It follows your site automatically.

There's no choice of region or server, and no automatic failover to another node.
The DNS records in your dashboard always point at the node your site actually runs on.