FAQ

This document describes FAQs when using OpenFunction.

Q: How to use private image repositories in OpenFunction?

A: OpenFunction uses Shipwright (which utilizes Tekton to integrate with Cloud Native Buildpacks) in the build phase to package the user function to the application image.

Users often choose to access a private image repository in an insecure way, which is not yet supported by the Cloud Native Buildpacks.

We offer a workaround as below to get around this limitation for now:

  1. Use IP address instead of hostname as access address for private image repository.
  2. You should skip tag-resolution when you run the Knative-runtime function.

For references:

buildpacks/lifecycle#524

buildpacks/tekton-integration#31

Q: How to access the Knative-runtime function without introducing a new ingress controller?

A: OpenFunction provides a unified entry point for function accessibility, which is based on the Ingress Nginx implementation. However, for some users, this is not necessary, and instead, introducing a new ingress controller may affect the current cluster.

In general, accessible addresses are for the sync(Knative-runtime) functions. Here are two ways to solve this problem:

  • Magic DNS

    You can follow this guide to config the DNS.

  • CoreDNS

    This is similar to using Magic DNS, with the difference that the configuration for DNS resolution is placed inside CoreDNS. Assume that the user has configured a domain named “openfunction.dev” in the ConfigMap config-domain under the knative-serving namespace (as shown below):

    1. $ kubectl -n knative-serving get cm config-domain -o yaml
    3. apiVersion: v1
    4. data:
    5.   openfunction.dev: ""
    6. kind: ConfigMap
    7. metadata:
    8.   annotations:
    9.     knative.dev/example-checksum: 81552d0b
    10.   labels:
    11.     app.kubernetes.io/part-of: knative-serving
    12.     app.kubernetes.io/version: 1.0.1
    13.     serving.knative.dev/release: v1.0.1
    14.   name: config-domain
    15.   namespace: knative-serving

    Next, let’s add an A record for this domain. OpenFunction uses Kourier as the default network layer for Knative Serving, which is where the domain should flow to.

    1. $ kubectl -n kourier-system get svc
    3. NAME               TYPE           CLUSTER-IP     EXTERNAL-IP   PORT(S)                      AGE
    4. kourier            LoadBalancer   10.233.7.202   <pending>     80:31655/TCP,443:30980/TCP   36m
    5. kourier-internal   ClusterIP      10.233.47.71   <none>        80/TCP                       36m

    Then the user only needs to configure this Wild-Card DNS resolution in CoreDNS to resolve the URL address of any Knative Service in the cluster.

    Where “10.233.47.71” is the address of the Service kourier-internal.

    1. $ kubectl -n kube-system get cm coredns -o yaml
    3. apiVersion: v1
    4. data:
    5.   Corefile: |
    6.     .:53 {
    7.         errors
    8.         health
    9.         ready
    10.         template IN A openfunction.dev {
    11.           match .*\.openfunction\.dev
    12.           answer "{{ .Name }} 60 IN A 10.233.47.71"
    13.           fallthrough
    14.         }
    15.         kubernetes cluster.local in-addr.arpa ip6.arpa {
    16.           pods insecure
    17.           fallthrough in-addr.arpa ip6.arpa
    18.         }
    19.         hosts /etc/coredns/NodeHosts {
    20.           ttl 60
    21.           reload 15s
    22.           fallthrough
    23.         }
    24.         prometheus :9153
    25.         forward . /etc/resolv.conf
    26.         cache 30
    27.         loop
    28.         reload
    29.         loadbalance
    30.     }
    31.     ...

    If the user cannot resolve the URL address for this function outside the cluster, configure the hosts file as follows:

    Where “serving-sr5v2-ksvc-sbtgr.default.openfunction.dev” is the URL address obtained from the command “kubectl get ksvc”.

    1. 10.233.47.71 serving-sr5v2-ksvc-sbtgr.default.openfunction.dev

After the above configuration is done, you can get the URL address of the function with the following command. Then you can trigger the function via curl or your browser.

  1. $ kubectl get ksvc
  3. NAME                       URL
  4. serving-sr5v2-ksvc-sbtgr   http://serving-sr5v2-ksvc-sbtgr.default.openfunction.dev

Q: How to enable and configure concurrency for functions?

A: OpenFunction categorizes function types into “sync runtime” and “async runtime” based on the type of request being handled. These two types of functions are driven by Knative Serving and Dapr + KEDA.

Therefore, to enable and configure the concurrency of functions, you need to refer to the specific implementation in the above components.

The following section describes how to enable and configure concurrency of functions in OpenFunction according to the “sync runtime” and “async runtime” sections.

Sync runtime

You can start by referring to this document in Knative Serving on enabling and configuring concurrency capabilities.

Knative Serving has Soft limit and Hard limit configurations for the concurrency feature.

Soft limit

You can refer to the Global(ConfigMap) and Global(Operator) sections of this document to configure global concurrency capabilities.

And for Per Revision you can configure it like this.

  1. apiVersion: core.openfunction.io/v1beta1
  2. kind: Function
  3. metadata:
  4.   name: function-sample
  5. spec:
  6.   serving:
  7.     scaleOptions:
  8.       knative:
  9.         autoscaling:
  10.           target: "200"

Hard limit

OpenFunction currently doesn’t support configuring Hard limit for Per Revision. You can refer to the Global(ConfigMap) and Global(Operator) sections of this document to configure global concurrency capabilities.

In summary

In a nutshell, you can configure Knative Serving’s Autoscaling-related configuration items for each function as follows, just as long as they can be passed in as Annotation, otherwise, you can only do global settings.

  1. # Configuration in Knative Serving
  2. apiVersion: serving.knative.dev/v1
  3. kind: Service
  4. metadata:
  5.   name: helloworld-go
  6.   namespace: default
  7. spec:
  8.   template:
  9.     metadata:
  10.       annotations:
  11.         autoscaling.knative.dev/<key>: "value"
  13. # Configuration in OpenFunction (recommended)
  14. apiVersion: core.openfunction.io/v1beta1
  15. kind: Function
  16. metadata:
  17.   name: function-sample
  18. spec:
  19.   serving:
  20.     scaleOptions:
  21.       knative:
  22.         autoscaling:
  23.           <key>: "value"
  25. # Alternative approach
  26. apiVersion: core.openfunction.io/v1beta1
  27. kind: Function
  28. metadata:
  29.   name: function-sample
  30. spec:
  31.   serving:
  32.     annotations:
  33.       autoscaling.knative.dev/<key>: "value"

Async runtime

You can start by referring to the document in Dapr on enabling and configuring concurrency capabilities.

Compared to the concurrency configuration of sync runtime, the concurrency configuration of async runtime is simpler.

  1. # Configuration in Dapr
  2. apiVersion: apps/v1
  3. kind: Deployment
  4. metadata:
  5.   name: nodesubscriber
  6.   namespace: default
  7. spec:
  8.   template:
  9.     metadata:
  10.       annotations:
  11.         dapr.io/app-max-concurrency: "value"
  13. # Configuration in OpenFunction (recommended)
  14. apiVersion: core.openfunction.io/v1beta1
  15. kind: Function
  16. metadata:
  17.   name: function-sample
  18. spec:
  19.   serving:
  20.     annotations:
  21.       dapr.io/app-max-concurrency: "value"

Last modified June 21, 2022: add FAQ doc (#118) (630a997)