Affinity Trait

Allows constraining which nodes the integration pod(s) are eligible to be scheduled on, based on labels on the node, or with inter-pod affinity and anti-affinity, based on labels on pods that are already running on the nodes.

It’s disabled by default.

This trait is available in the following profiles: Kubernetes, Knative, OpenShift.

Configuration

Trait properties can be specified when running any integration with the CLI:

$ kamel run --trait affinity.[key]=[value] --trait affinity.[key2]=[value2] integration.groovy

The following configuration options are available:

Property Type Description

Property	Type	Description
`affinity.enabled`	`bool`	Can be used to enable or disable a trait. All traits share this common property.
`affinity.pod-affinity`	`bool`	Always co-locates multiple replicas of the integration in the same node (default false).
`affinity.pod-anti-affinity`	`bool`	Never co-locates multiple replicas of the integration in the same node (default false).
`affinity.node-affinity-labels`	`[]string`	Defines a set of nodes the integration pod(s) are eligible to be scheduled on, based on labels on the node.
`affinity.pod-affinity-labels`	`[]string`	Defines a set of pods (namely those matching the label selector, relative to the given namespace) that the integration pod(s) should be co-located with.
`affinity.pod-anti-affinity-labels`	`[]string`	Defines a set of pods (namely those matching the label selector, relative to the given namespace) that the integration pod(s) should not be co-located with.

affinity.enabled

bool

Can be used to enable or disable a trait. All traits share this common property.

affinity.pod-affinity

bool

Always co-locates multiple replicas of the integration in the same node (default false).

affinity.pod-anti-affinity

bool

Never co-locates multiple replicas of the integration in the same node (default false).

affinity.node-affinity-labels

[]string

Defines a set of nodes the integration pod(s) are eligible to be scheduled on, based on labels on the node.

affinity.pod-affinity-labels

[]string

Defines a set of pods (namely those matching the label selector, relative to the given namespace) that the integration pod(s) should be co-located with.

affinity.pod-anti-affinity-labels

[]string

Defines a set of pods (namely those matching the label selector, relative to the given namespace) that the integration pod(s) should not be co-located with.

Examples

To schedule the integration pod(s) on a specific node using the built-in node label kubernetes.io/hostname:
```
$ kamel run -t affinity.node-affinity-labels="kubernetes.io/hostname in(node-66-50.hosted.k8s.tld)" ...
```

To schedule a single integration pod per node (using the Exists operator):

$ kamel run -t affinity.pod-anti-affinity-labels="camel.apache.org/integration" ...

To co-locate the integration pod(s) with other integration pod(s):

$ kamel run -t affinity.pod-affinity-labels="camel.apache.org/integration in(it1, it2)" ...

The *-labels options follow the requirements from Label selectors. They can be multi-valuated, then the requirements list is ANDed, e.g., to schedule a single integration pod per node AND not co-located with the Camel K operator pod(s):

$ kamel run -t affinity.pod-anti-affinity-labels="camel.apache.org/integration" -t affinity.pod-anti-affinity-labels="camel.apache.org/component=operator" ...

More information can be found in the official Kubernetes documentation about Assigning Pods to Nodes.