Node Operations

Nodes have their own day-2 operations beyond adding them.

Capacity#

Each node has separate build and rollout concurrency settings — how many builds and how many container cutovers it runs at once. The CLI adjusts them per node:

zensailor node list
zensailor node capacity <node-id>            # show limits
zensailor node capacity <node-id> --build 2 --rollout 4

Maintenance mode#

Putting a node on maintenance drains it: stateless services reschedule to healthy nodes, and deployments stop landing there. Core containers (the platform) are never drained — the main node keeps the platform running even while it drains workloads. Volume-backed services stay on their data node (draining cannot move them) — see Storage Across Nodes.

zensailor node maintenance-on <node-id>
zensailor node maintenance-off <node-id>

Node pressure#

The platform watches disk, memory and CPU on every node. A node under sustained pressure stops receiving new deployments — the scheduler refuses machines that are over threshold, not just individual runs — and fires notifications with recovery events and a disk-space forecast. You are told what is happening to capacity before it silently degrades.

Decommissioning#

Removing a node (zensailor node rm <node-id>) only works when it holds no managed data: either migrate volume-backed services away first, or accept that their data on that node is lost with it. Stateless services simply reschedule elsewhere.