rasahq / helm-charts

Helm charts for Rasa products

Rasa Helm Charts

Official Helm charts for Rasa products:

• Rasa Open Source
• Rasa Action Server

Helm charts for components used by Rasa:

• Duckling

Prerequisites

• Helm 3 (>= 3.5 )
• Kubernetes 1.14+

How to use Rasa Helm repository?

You need to add this repository to your Helm repositories:

helm repo add rasa https://helm.rasa.com
helm repo update

Where to get help

• If you encounter bugs or have suggestions for this Helm chart, please create an issue in this repository.
• If you have general questions about usage, please create a thread in the Rasa Forum.

How to contribute

We are very happy to receive your contributions. You can find more information about how to contribute to Rasa (in lots of different ways!) here.

To contribute via pull request, follow these steps:

1. Create an issue describing the feature you want to work on
2. Create a pull request describing your changes

Remember that your PR should include:

1. Details about the changes introduced by a given chart version (use the artifacthub.io/changes annotation in the Chart.yaml file)
2. Update the chart version
3. Add additional annotations that indicate your changes (see: https://artifacthub.io/docs/topics/annotations/helm/#supported-annotations)

License

Licensed under the Apache License, Version 2.0. Copyright 2021 Rasa Technologies GmbH. LICENSE.

helm-charts's Issues

Network policy should match the rasa fullname and not just rasa

Hi, if I specify a different name, like "rasa-production" the network policies no longer apply.

See:

helm-charts/charts/rasa/templates/network-policy/rasa-bot-allow-egress-http-https.yaml

Line 9 in 2875622

app.kubernetes.io/name: rasa

It should be something like: {{ include "rasa-common.names.fullname" . }}

Seemingly randomly getting an error of function "lookup" not defined

I'm running into this strange error, when it actually previously worked a few days ago.

I initially got this error with helm upgrade ..., and then get it now with the normal install.

Whats confusing to me is I don't see this path in the repo?

helm install --namespace rasa-initial rasa-initial  rasa/rasa --debug            
install.go:149: [debug] Original chart version: ""
install.go:166: [debug] CHART PATH: /home/chris/.cache/helm/repository/rasa-1.17.7.tgz

Error: parse error at (rasa/charts/postgresql/charts/common/templates/_secrets.tpl:84): function "lookup" not defined
helm.go:75: [debug] parse error at (rasa/charts/postgresql/charts/common/templates/_secrets.tpl:84): function "lookup" not defined

ingressClassName missing from ingress

Hi,

The ingressClassName is missing from the helm chart and templates.

https://kubernetes.io/docs/concepts/services-networking/ingress/#the-ingress-resource

It is available in the Rasa-X chart: https://github.com/RasaHQ/rasa-x-helm/blob/87a311f92481a2f476a517fb9c80766b16696022/charts/rasa-x/values.yaml#L830

Thanks.

Cannot set rasaXToken, rasaToken and JWT_SECRET from custom secret

Hi,

The helm chart doesn't seem to allow setting custom secrets for these values outside of the values file like the Rasa X Helm chart allows.

It would also help to have the naming between the different charts aligned, as it is very confusing dealing with multiple names referring to the same thing, across the charts.

Thanks,
Jonathan

Duckling Network Policy missing

Hi,

This chart has all the other network policies required, but the duckling one is missing:

https://github.com/RasaHQ/rasa-x-helm/blob/87a311f92481a2f476a517fb9c80766b16696022/charts/rasa-x/templates/network-policy.yaml#L1020-L1061

Thanks,
Jonathan

404 error connecting to socket channel on K8S cluster

hello we used this HELM Chart to deploy our RASA BOT image with credentials.yaml configured for socket channel.
Of course we are able to get it all working before K8S deployment

Now we deployed the bot in K8S cluster using this helm chart and the bot is responding with "Hello from Rasa: 3.3.1" when we visit https://my-bot.domain.com.. But we are getting the below error when we are connecting to socket.io channel.

https://my-bot.domain.com./socket.io/?EIO=4&transport=polling&t=OUz5vHV 404
Error: xhr poll error
    at n.value (index.js:99:957)
    at n.<anonymous> (index.js:193:227277)
    at r.emit (index.js:84:263874)
    at n.value (index.js:193:225378)
    at index.js:193:224880

we followed the configuration here and tried below combinations and still the same above error

We didn't deploy RasaX /Enterprise HELM chart. SO its just RASA Opensource and Action server Helm Charts we deployed.

1. set applicationSettings.credentials=false ( the idea here is to let the RASA use the credentials.yaml from our image)
2. Then changed the rasa-values.yaml as follows...

credentials:
      # -- Enable credentials configuration for channel connectors
      enabled: true

      # -- Additional channel credentials which should be used by Rasa to connect to various
      # input channels
      ## See: https://rasa.com/docs/rasa/messaging-and-voice-channels
      additionalChannelCredentials: {}
      #rest:
      #  facebook:
      #    verify: "rasa"
      #    secret: "<SECRET>"
      #    page-access-token: "<PAGE-ACCESS-TOKEN>"
   
      socketio:
        user_message_evt: user_uttered 
        bot_message_evt: bot_uttered 
        session_persistence: true

can you help where / what we are missing

Documentation on how to use HTTPS between Duckling, ActionServer and RasaOss pods?

Hi,

Is there a way to activate HTTPS between all the components (Duckling, ActionServer and RasaOss pods?)

It is not clear whether this should be done via an internal ingress instead of a service.

Best,

Network policy for redis points to postgres port

Hi,

The network policy for redis points to the postgres port:

helm-charts/charts/rasa/templates/network-policy/rasa-bot-redis.yaml

Lines 13 to 25 in c93efed

	{{- if and (not .Values.redis.external.enabled) (eq (include "rasa-common.redis.available" .) "true") }}
	- to:
	- podSelector:
	matchLabels:
	app.kubernetes.io/name: redis
	ports:
	- protocol: TCP
	port: {{ template "rasa-common.psql.port" . }}
	{{- else }}
	- ports:
	- protocol: TCP
	port: {{ template "rasa-common.psql.port" . }}
	{{- end }}

Can you please fix this to use the redis port value from the values file.

Rasa Chart currently requires Bitnami Charts that are no longer available

Currently the Rasa Chart uses the following dependencies

Repository	Name	Version
https://charts.bitnami.com/bitnami	postgresql	~10.16.2
https://charts.bitnami.com/bitnami	rabbitmq	~8.32.2
https://charts.bitnami.com/bitnami	redis	~15.7.6

All those are no longer available, so a helm dependency update fails with unmatchable requirements.

Updating the Charts to the versions below works, but requires tweaking the postgresql configuration a bit, since the name and defaults for keys in the secrets have changed with the new bitnami charts:

postgresql:
  ...
  # used to be "postgresql-password" in older versions of the bitnami charts
  existingSecretKey: postgres-password
  auth:
    username: "postgres"
    database: "rasa"

Repository	Name	Version
https://charts.bitnami.com/bitnami	postgresql	~12.2.1
https://charts.bitnami.com/bitnami	rabbitmq	~11.10.0
https://charts.bitnami.com/bitnami	redis	~17.8.0

How to use external redis server for lock store in rasa oss

I follow redislockstore but getting error

redis.exceptions.ResponseError: MOVED 7981 <MY REDIS IP>:<MY REDIS PORT>

Does Rasa Oss support external Redis for the lock store?

Thanks

[CI] Remind to update the docs every time when dependencies are updated

The Rasa docs contain (WIP) versions of the subcharts used as dependencies. We should make sure that every time when one of the subcharts is updated the information about versions in the docs is updated as well.

Event broker config supports only Pika

I want the following event broker config in endpoints.yml but the helm chart doesn't allow this:

event_broker:
  type: file
  path: rasa_events.json

The template is configured to always emit the Rabbit related event broker values which will be rejected by the file broker and rasa will not start.

Kafka eventBroker setting still insert rabbit env vars

When the key data.endpoints.eventBroker.type is set to kafka, the following default Rabbit related values should not be set:

      password: ${RABBITMQ_PASSWORD}
      port: ${RABBITMQ_PORT}
      username: ${RABBITMQ_USERNAME}

Can't train an inital model with files downloaded from github

Hello ,
I am trying to deploy using helm and i want an inital model to be trained. i have downloaded my bot files and placed it into a directory under "app" name with the rasa-values.yml . i saw on the documentation that we can download and train an initial model
here is my configuration

initContainers:

• name: init
  image: "busybox"
  command: ["/bin/sh", "-c"]
  args:
  • "cd ./app && touch app.txt" #just for testing
    volumeMounts:
  • mountPath: /app
    name: models

-- Specify additional volumes to mount in the rasa-oss container

Ref: https://kubernetes.io/docs/concepts/storage/volumes/

volumes:
- name: models
emptyDir: {}

-- Specify additional volumes to mount in the rasa-oss container

volumeMounts:

• name: models
  mountPath: /app

readOnly: true

could you help me please ? the model is not trained despite i can reach my bot and i think its the starting project of rasa that is loaded (tested with curl)

am not using rasa x yet (they said it will be remove so..)

Deployment no longer works with OpenShift

Have been testing against Openshift 4.11 and the pods won't spin up citing

create Pod test-postgresql-0 in StatefulSet test-postgresql failed error: pods "test-postgresql-0" is forbidden: unable to validate against any security context constraint: [provider "anyuid": Forbidden: not usable by user or serviceaccount, spec.containers[0].securityContext.runAsUser: Invalid value: 1001: must be in the ranges: [1001300000, 1001309999], provider "nonroot-v2": Forbidden: not usable by user or serviceaccount, provider "nonroot": Forbidden: not usable by user or serviceaccount, provider "pcap-dedicated-admins": Forbidden: not usable by user or serviceaccount, provider "hostmount-anyuid": Forbidden: not usable by user or serviceaccount, provider "machine-api-termination-handler": Forbidden: not usable by user or serviceaccount, provider "hostnetwork-v2": Forbidden: not usable by user or serviceaccount, provider "hostnetwork": Forbidden: not usable by user or serviceaccount, provider "hostaccess": Forbidden: not usable by user or serviceaccount, provider "splunkforwarder": Forbidden: not usable by user or serviceaccount, provider "node-exporter": Forbidden: not usable by user or serviceaccount, provider "privileged": Forbidden: not usable by user or serviceaccount]

I had tried a values file looking like the following

applicationSettings:
  initialModel: "https://github.com/RasaHQ/rasa-x-demo/blob/master/models/model.tar.gz?raw=true"
postgresql:
  volumePermissions:
    securityContext:
      runAsUser: "auto"
  securityContext:
    enabled: false
  shmVolume:
    chmod:
      enabled: false
nginx:
  image:
    name: nginxinc/nginx-unprivileged
    port: 8080

Deleted my namespace and re-ran the installation - but the same error persists. Looking in the stateful set, it does appear as though uid 1001 is used for the user, which is not allowed in OpenShift.

In JSON schema `fullnameOverride` is defined as type: null - means it cannot be overwritten with string

When I try to deploy the Rasa Helm chart and set the fullnameOverride in my values file, I get an error:

helm -n mynamespace install --dry-run --generate-name rasa/rasa --set fullnameOverride=test1

gives

Error: INSTALLATION FAILED: values don't meet the specifications of the schema(s) in the following chart(s):
rasa:
- fullnameOverride: Invalid type. Expected: null, given: string

It seeps Helm 3 applies the JSON Schema to validate the values.yaml, and because this attribute is marked as type: null in the values.schema.json, it does not seem to be possible to override it with a string value.

Please let me know if you have any workaround

mongo Custom Tracker store

Hello,
can i request how to configure custom Tracker store ( we use Mongo DB) in Rasa helm chart

Before HELM, we have configured our custom tracker store in endpoints.yaml as follows

tracker_store: type: myTracker.MongoTrackerStore url: mongodb+srv://mongodb-URL db: admin username: admin password: mypassword

But not sure how can we make use of our custom tracker store when we deploy using K8S

We did give a try by configuring endpoints.trackerstore in values.yaml, but as per the comment i think it needs to enable postgresql.install = true, but we want to use MongoDB.

any suggestions pls?

nil pointer evaluating interface when upper level doesn't exist in .Values

I get the following error when enabling rabbitmq without having a .Values.rabbitmq.service section in values.yaml:

Error: UPGRADE FAILED: template: rasa/templates/deployment.yaml:107:14: executing "rasa/templates/deployment.yaml" at <include "rasa.containers.env" .>: error calling include: template: rasa/templates/_containers-env.yaml:16:3: executing "rasa.containers.env" at <include "rasa-common.rabbitmq.envs" .>: error calling include: template: rasa/charts/rasa-common/templates/_rabbitmq.tpl:30:24: executing "rasa-common.rabbitmq.port" at <.Values.rabbitmq.service.port>: nil pointer evaluating interface {}.port

This is a known helm issue.

The offending lines in this case:

helm-charts/charts/rasa-common/templates/_rabbitmq.tpl

Lines 26 to 31 in c792f05

	{{/*
	Return the rabbitmq port.
	*/}}
	{{- define "rasa-common.rabbitmq.port" -}}
	{{- coalesce .Values.rabbitmq.service.port 5672 -}}
	{{- end -}}

As it is implemented now, the coalesce is all but useless. This can be worked around in the user side, if an empty .Values.rabbitmq.service is defined, but that almost defeats the purpose of a coalesce/default.

On the side of the Chart, the following line would fix the problem:

{{- default 5672 ((.Values.rabbitmq).service).port -}}

Duckling container won't start (Helm Chart v4.5.6)

Context:

Openshift v4.9
Rasa-x Helm Chart v4.5.6
Duckling Image: rasa/duckling:0.2.0.2

Due security constraints of OpenShift is the global securityContext set to false

The duckling container throws the following error in the log:

duckling-example-exe: log: createDirectory: permission denied (Permission denied)

This has probably something to do with the context of openshift. With Rasa-x Helm Chart v4.5.4 (and Duckling image "0.1.6.3") we didn't had this issue.

Duckling endpoint not passed to pod when external is True and an external URL is provided

helm-charts/charts/rasa/templates/helpers/_duckling.tpl

Line 5 in e7e72f4

{{- if or .Values.duckling.install (and .Values.duckling.external.enabled (not .Values.duckling.external.url)) -}}

This results in the following:
if false or (true and (not true)) resulting in false. We need this to be true when external = true and external url = true.

{{- if or .Values.duckling.install (and .Values.duckling.external.enabled .Values.duckling.external.url) -}} would be better.

This means the duckling endpoint is not provided to the pod here:

helm-charts/charts/rasa/templates/_containers-env.yaml

Lines 21 to 23 in e7e72f4

	{{- if eq (include "rasa.endpoints.duckling" .) "true" }}
	{{ include "rasa.duckling.envs" . }}
	{{- end }}

Revisit structure of deployment docs

We need to revisit how we recommend deploying Rasa Open Source, Rasa X, and Rasa Action Server in the docs.

New production recommendation: deploy with new Rasa Open Source helm chart and forward events to Rasa X (using helm chart that is coming soon)

Open questions

• When do we nudge people to set up Rasa X?
• How do we structure the documentation with the separate deployments?
• What do we do with local mode?

@tmbo and I discussed removing the Rasa X installation from here, making the helm charts the default deployment method for the production use case. However, it's not clear where we nudge people to set up Rasa X.

For Rasa X / Enterprise, we also discussed structuring the documentation on deployment similar to gitlab with the two installation methods we'll have there:

• helm charts
• single line install

null type of nodePort field in JSON schema

'nodePort ' field in JSON schema has type of null which i think should be an integer.

Add envFrom to rasa deployement

Hi there,
I just make a proposal #112 to add an envFrom resource to the rasa deployment to easily add environament variables from configmaps or secrets data's