Skip to main content

Placement Constraints

Placement constraints let you control which Docker Swarm nodes are eligible to run your application's workloads. This page explains the default behavior for cluster-mode apps in Appliku and how to adjust it.

Placement constraints apply only to cluster deployments. Standalone server deployments run on a single server and have no placement logic.

Default Behavior​

When Appliku generates the docker-compose.yml for a cluster deployment, it chooses a default placement automatically based on the topology of your cluster:

  • Cluster has worker nodes -- All application services are constrained to node.role == worker. Manager nodes are reserved for the Swarm control plane and are not used for workload placement.
  • Cluster has no worker nodes (single-node / manager-only clusters) -- No placement constraint is applied. Docker schedules workloads on the available manager node.

The same rule applies to one-off commands (for example, the release command): when the default worker constraint is in effect, the generated docker service create call adds --constraint 'node.role == worker'; otherwise the constraint is omitted.

Allow Placement on Manager Nodes​

Some teams want manager nodes to carry workload alongside workers -- for example, to get more out of a small cluster without maintaining a separate dedicated manager. Appliku exposes a per-application toggle for this.

  1. Open your application in the Appliku dashboard
  2. Go to Settings → Build
  3. Check Allow placement on manager nodes
  4. Save
  5. Redeploy the application for the change to take effect

When this option is enabled, Appliku skips the default node.role == worker constraint even when worker nodes are present. Docker Swarm then considers every node in the cluster -- managers and workers -- when scheduling the application's services.

note

This flag is only meaningful for cluster-mode applications. It has no effect on standalone server deployments.

Per-Process Placement Constraints​

For fine-grained control, each process can carry its own placement constraints that override the cluster-wide default. This is useful when some processes should be pinned to specific nodes (for example, a process that needs a node labeled disk=ssd).

  1. Open your application
  2. Go to the Processes tab
  3. Edit the process
  4. Set Placement constraints -- one constraint per line (for example, node.role == worker or node.labels.disk == ssd)
  5. Save and redeploy

Per-process constraints take priority over the application-wide default, so a process with explicit constraints is unaffected by the "Allow placement on manager nodes" toggle.

Priority Summary​

For each service in the generated docker-compose.yml, placement is decided in this order (highest wins):

  1. Per-process placement constraints -- used verbatim if set on the process.
  2. Default cluster constraint (node.role == worker) -- applied when the cluster has worker nodes and Allow placement on manager nodes is off.
  3. No constraint -- applied when the cluster has no worker nodes, or when Allow placement on manager nodes is on.

Next Steps​