# Configure the controller manager rate limit

# Configure the controller manager rate limit

This pages describes how to configure the overall rate limit (token bucket rate
limit) for the controller manager in Config Connector.

The rate limit controls how many requests Config Connector makes to the
Kubernetes API server. There are two settings:

- The `qps` (queries per second) setting determines
  the average rate at which Config Connector can make requests to the
  Kubernetes API server.

- The `burst` value allows for more requests than the `qps` limit
  for short periods. This helps Config Connector to reconcile
  quickly, even during times which require more queries to the Kubernetes
  API server, such as when applying a new batch of configuration using GitOps.

You can think of the "token bucket" algorithm for rate limiting like this:
imagine a bucket of capacity `burst` tokens, being refilled at the `qps`
rate. Each request requires one token from the bucket. If the bucket is empty,
requests are delayed until tokens become available.

## Configure the rate limit for namespaced controller manager

If Config Connector is configured to run in
[namespaced mode](https://docs.cloud.google.com/config-connector/docs/how-to/install-namespaced), you can
use the `NamespacedControllerReconciler` custom resource to configure the rate
limit of the `cnrm-controller-manager` controller's Kubernetes client
in your designated namespace. You can configure rate limit for this
controller in Config Connector version 1.119 and later.

1. Create a file named `configure-rate-limit.yaml` and copy the
   following YAML into it:

       apiVersion: customize.core.cnrm.cloud.google.com/v1beta1
       kind: NamespacedControllerReconciler
       metadata:
         name: cnrm-controller-manager # name must not contain the namespace ID suffix
         namespace: NAMESPACE
       spec:
         rateLimit:
           qps: 80 # the default value is 20
           burst: 40 # the default value is 30

   Replace `NAMESPACE` with the name of your namespace.
2. Use `kubectl apply` to apply the rate limit configuration to your cluster:

   ```
   kubectl apply -f configure-rate-limit.yaml
   ```
3. Verify the configuration is successful by running the following command:

   ```
   kubectl get namespacedcontrollerreconciler cnrm-controller-manager -n NAMESPACE -o jsonpath='{.status.healthy}'
   ```

   The output should display `status.healthy` field set to `true`.
4. Verify that the `--qps` and `--burst` flags are added to the
   `cnrm-controller-manager` controller's manager container by running the following command:

   ```
   kubectl describe statefulsets -n cnrm-system -l "cnrm.cloud.google.com/scoped-namespace=NAMESPACE"
   ```

   The output should contain the following:

       Containers:
       manager:
         Image:      gcr.io/gke-release/cnrm/controller:fd4c42c
         Port:       23232/TCP
         Host Port:  0/TCP
         Command:
           /configconnector/manager
         Args:
           --qps=80
           --burst=40
           --scoped-namespace=config-control
           --prometheus-scrape-endpoint=:8888

## Configure the rate limit in cluster mode

Starting in version 1.125, if Config Connector is configured to run in cluster
mode, you can configure the rate limit of the `cnrm-controller-manager`
controller's Kubernetes client using the `ControllerReconciler` custom resource.
The usage is similar to the namespaced mode, but in this case, you should
use the `ControllerReconciler` CR instead of `NamespacedControllerReconciler`,
as in the following example:

    apiVersion: customize.core.cnrm.cloud.google.com/v1alpha1
    kind: ControllerReconciler
    metadata:
      name: cnrm-controller-manager
    spec:
      rateLimit:
        qps: 80 # the default value is 20
        burst: 40 # the default value is 30