Configuring resource requests and limits for uipathctl-created workloads

How to adjust CPU and memory requests and limits when uipathctl-created workloads fail to schedule or complete.

Description

When you run a uipathctl command, a Kubernetes workload created by the command may fail because of its CPU or memory requirements.

You may encounter one of the following symptoms:

• The pod remains in the Pending state, and its events contain Insufficient cpu or Insufficient memory.
• The pod is terminated with the OOMKilled status.
• The workload is CPU-throttled or does not complete before the command times out.
• An admission policy rejects the pod because its resource requests or limits do not comply with a ResourceQuota or LimitRange configured in the namespace.

The issue can affect uipathctl-created workloads used for prerequisite checks, health checks, support bundle generation, migrations, Helm operations, and upgrades.

Solution

First, identify the affected pod and inspect its events:

kubectl get pods -A
kubectl describe pod <pod-name> -n <namespace>

Depending on the reported condition:

• For OOMKilled, increase the memory limit. Also increase the memory request if the pod must be scheduled on a node with more available memory.
• For a Pending pod with Insufficient cpu or Insufficient memory, reduce the request only if the workload can safely run with less guaranteed capacity. Otherwise, add cluster capacity.
• For quota or limit-range errors, specify requests and limits that comply with the policy configured in the target namespace.
• For slow or CPU-throttled workloads, increase the CPU limit.

You can change the resource requirements in input.json or override them for a single command.

Configuring resource requirements in input.json

uipathctl groups its workloads into the following resource classes:

Class	Operations	Default requests (CPU / memory)	Default limits (CPU / memory)
diagnostic	Health checks and read-only validation probes	50m / 64Mi	250m / 256Mi
operational	Prerequisite checks, support bundles, Helm operations, and identity data migration	1m / 1k	2 / 2Gi
workload	Object store, PVC, and MongoDB migrations, and the upgrade file server	250m / 512Mi	2 / 2Gi

To change the defaults for commands that use the Automation Suite configuration, add workload_resources to input.json. For example, to increase the resources available to data-heavy migration workloads, use the following configuration:

{
  "workload_resources": {
    "workload": {
      "requests": {
        "cpu": "500m",
        "memory": "1Gi"
      },
      "limits": {
        "cpu": "2",
        "memory": "4Gi"
      }
    }
  }
}

You can configure the diagnostic, operational, and workload classes independently. You can also specify only the values you want to change. Any omitted CPU or memory value retains its default.

Requests cannot exceed their corresponding limits. Use standard Kubernetes resource quantities, such as 500m or 1 for CPU and 256Mi or 1Gi for memory.

Overriding resource requirements for a single command

For commands that expose the --probe-cpu and --probe-memory options, you can override the resource requirements of the helper pod created by that command. For example:

uipathctl health check --probe-cpu 500m --probe-memory 512Mi

Each option sets both the request and the limit for the specified resource. If you specify only one option, the other resource retains its effective value.

The options are available for the following commands:

• uipathctl health check
• uipathctl health test
• uipathctl health diagnose
• uipathctl config add-host-admin
• uipathctl config enable-basic-auth
• uipathctl service aicenter sync-skills
• uipathctl service aicenter sync-skill-status
• uipathctl service orchestrator notifications-migration start

When both configuration methods apply, uipathctl resolves the values in the following order, where the later value takes precedence:

1. The built-in default.
2. The workload_resources value in input.json or the applied cluster configuration.
3. The --probe-cpu or --probe-memory command-line option.

After updating the resource requirements, run the affected command again and confirm that the uipathctl-created workload starts successfully.

Set the smallest requests and limits that allow the workload to complete. Requests that are too high can prevent scheduling, whereas limits that are too low can cause throttling or termination.