Infrastructure Pods and Resource Management
During the installation procedure you will be able to provide information on how to best "operationalize" your infrastructure. Through the configuration of
--operator-resources you will be able to drive the operator
Pods scheduling and to be able to assign resources.
The usage of these advanced properties assumes you’re familiar with the Kubernetes Scheduling concepts and configurations.
| The aforementioned flags setting will work both with |
The most basic operation we provide is to let you assign Camel K operator
Pods to a specific cluster
--node-selector option. The functionality is based on
NodeSelector Kubernetes feature. As an example, you can schedule Camel K infra
Pods to a specific
Node of your cluster:
kamel install --node-selector kubernetes.io/hostname=ip-172-20-114-199.ec2.internal ...
The option will accept any
key=value pair as specified and supported by Kubernetes. You can specify more than one
--toleration option will let you tolerate a Camel K infra
Pod to support any matching
Taint according the
Toleration Kubernetes feature. As an example, let’s suppose we have a node tainted as "dedicated=camel-k:NoSchedule". In order to allow the infra
Pods to be scheduled on that
Node we can provide the following option during installation procedure:
kamel install --toleration dedicated=camel-k:NoSchedule ...
The option accept a value in the following format
Key[=Value]:Effect[:Seconds] where the values in squared bracket are considered optional and
Effect must admit any of the
Taints accepted values:
NoExecute. You can specify more than one
While installing the Camel K operator, you can also specify the resources requests and limits to assign to the operator
--operator-resources option. The option will expect the configuration as required by Kubernetes Resource Management.
kamel install --operator-resources requests.cpu=256m --operator-resources limits.cpu=500m --operator-resources limits.memory=512Mi ...
The value expected by the option are in the form
requestType must be either
requestResource must be either
value expressed in the numeric value as expected by the resource. You can specify more than one
|if you specify a limit, but does not specify a request, Kubernetes automatically assigns a request that matches the limit.|
The main contributor to compute resources consumption is likely to be the Maven builds that are performed in the operator
Pod (assuming you use the default build strategy). So the resource requirements should be defined accordingly. The following requirements are sensible defaults that should work in most cases:
resources: limits: cpu: "1" memory: 1Gi requests: cpu: 500m memory: 512Mi
Note that if you plan to perform native builds, then the memory requirements must be increased significantly. Also, the CPU requirements are rather "soft", in the sense that it won’t break the operator, but it’ll perform slower in general.