zbrookle / avionix Goto Github PK

I am trying to deploy an HPA which uses CrossVersionObjectReference. I do not see the kind field in the api docs here. Let me know if I am missing something.

What i'm trying to do:
install chart with ChartBuilder, sample code

 if builder.is_installed:
        builder.upgrade_chart()
 else:
        builder.install_chart()

What i'd expect to happen:
either install or upgrade chart

What actually happens:
i get a weird KeyError on this object:

  File "/home/federico/.local/lib/python3.8/site-packages/avionix/chart/chart_builder.py", line 316, in upgrade_chart
    self.__check_if_installed()
  File "/home/federico/.local/lib/python3.8/site-packages/avionix/chart/chart_builder.py", line 245, in __check_if_installed
    if not self.is_installed:
  File "/home/federico/.local/lib/python3.8/site-packages/avionix/chart/chart_builder.py", line 336, in is_installed
    return self.chart_info.name in installations["NAME"]
KeyError: 'NAME'

Solution:
after debugging and keeping note of the object i found the problem and solved it: if the kube config file in /home//.kube/config is group or world readable, when you run helm list helm returns a warning "YOUR KUBE CONFIG IS WORLD READABLE; THIS IS INSECURE" and your parser mixes the error message in the dictionary which is supposed to contain the installed charts.
As a result anyone with wrong permissions on the config file won't be able to install/upgrade charts.

if you could raise an exception or something to let the user know they should fix the permissions on their file it could be useful for the next guy who has the same problem.

Thank you!

Hi @zbrookle , I came accross your post on stackoverflow while looking for a viable solution to deploy helm charts with Python 3.
What I need might be already baked into Avionix already but couldn't find a documentation regarding how to do this so just wanted to reach out to you.
I'm looking to simply deploy existing helm charts from existing repositories without having to build the specs from ground up. As an example, let's suppose I just want to deploy harbor chart and what I do basically is:

helm repo add harbor https://helm.goharbor.io
helm install my-release harbor/harbor

Is this possible with Avionix? If not, are there any plans for the implementation?
Thanks

Need to fix header in set up dev environment section

I need to run a job in AKS, was looking for an option to deploy Job using avionix lib, pls add some example

Is there a chance Helm 2 will be officially supported by avionix? So far it is unclear how it operates with Helm 2.

Create an implementation of values.yaml that will allow for it's use in a similar way to classic helm charts, allowing for backwards compatibility if a user wants to publish their chart in the Helm eco system

Could not deploy the Helm chart in a local directory on Windows, error with the arguments!
Followed the example on - https://github.com/zbrookle/avionix/blob/master/examples/external_helm_charts_example.py
The example also seems to be a little confusing. could you please provide a better example if the feature is enabled ?

Hi,
changing the output directory in chart builder lead to unwanted behaviors, indeed it is considered only when writing out new chart files, not when deleting them after installation:

def __delete_chart_directory(self):
    if os.path.exists(self.chart_info.name):
        shutil.rmtree(self.chart_info.name)

(chart_info.name doesn't include the set output_directory).
Especially during upgrades, this cause merging values from old installations, differently from the expected program behavior.
I think the fix is really easy to implement (I also noticed there is a pretty old issue request to resolve this bug).
Please let me know if you prefer I fix this in a new merge request.

is this repo still active?

are future updates to be expected or is it up for maintenance? i am asking for the fast paced nature of kubernetes and its api.

Are there any plans to implement this feature (or have does it already exist and I have overlooked)? https://helm.sh/docs/helm/helm_rollback/

Is it possible lo load and install charts from local repo without creating them? The same way it is done by pyhelm. It was hard to find any info in the documentation - it showed how to build the chart first from an external URL, but is there a simple command for a local repo?

Example from pyhelm:

chart = ChartBuilder(json)
tiller.install_release(chart=chart.get_helm_chart(), namespace='default', dry_run=False)

If I look into the source code, ChartDependency takes a URL and then checks out the local repo. But what if I already have a local repo? Can it be a new feature?

I'm struggling to figure out how to auth. Is there an easy example for using a token?

Avionix currently requires pyyaml >=5.3, <5.4, but PyYAML versions prior to 5.4 are susceptible to the critical security vulnerability described in CVE-2020-14343.

Note: There was originally CVE-2020-1747, but the fix in PyYAML 5.3.1 was incorrect/incomplete.

I see there are some classes to build CustomResourceDefinitions but I can't see anything in the code to build a CustomResource. Is there a way to use avionix to build a CustomResource and package it as a helm chart?

does this project support helm v2.14.1 and helm v3.0.3

Without taking an options dict, we can't pass kube-context to tell helm which cluster to get the installations from. (Well we can, because the string handling isn't secure, we can just use f"{namespace} --kube-context {context}" for the namespace name, but that seems abusive).

is_installed as a property also has no way to pass the context. But that seems like a bigger lift to improve.

Howdy,

So I have read over the documentation a little bit and noticed there does not appear to be a way to override the chart path for the class ChartBuilder. The class uses the chart_info name property when building the path. I made an update to deal with this in my local work but was curious if there is a better way to do this?

For context, I need to deploy multiple charts with similar configurations but with different chart names.

Hi,

Is it possible to use built-in objects such as?

metadata:
  name: {{.Release.Name}}-{{.Chart.Name}}

Thanks
Schmil

Is there an option to list all helm charts installed in cluster?

Ingress with the extensions/v1beta1 API was deprecated in 1.16. The following two API versions map to the corresponding k8s versions:

Furthermore, the current Ingress class has the _non_standard_version set to v1beta1, which makes it impossible to override the apiVersion with the parameter when instantiating it.

I'm fairly new to using this library (thanks, by the way!), so I'm unsure how you'd prefer to handle this. If it's relatively straightforward, feel free to provide some high-level guidance and I can work on a PR.

Hi there!

I saw your comment over on cdk8s-team/cdk8s#21 but didn't want to answer there as it goes off-topic quite fast. Just like you I am working on something similar to cdk8s. I started it after finding cdk8s and getting frustrated with it 😅 (or actually frustrated with myself as I tried patching cdk8s but was to lazy to understand all the typescript-binding magic) So I thought maybe we could have a discussion about some design decisions as we might have similar goals. At least the python and helm focus is there so far ;)

I looked through your repo but didn't find any examples of how avionix looks in action. In the issue you said you generate helm charts - just like me :) But my helm charts are just single yamls with all the definitions inside. I thought about implementing something to also support values but I decided against it as the generation of the yamls is already dynamic and wanted to keep it simple. What is your use case? ArgoCD as well? Why not have "values" for your python chart?

The use case of @cweidinger for different environments / regions within ArgoCD I have not yet looked into, will have to talk to a colleague who knows more about ArgoCD. But in our use cases so far where we use my pdk8s deployment is generation time of the chart. So I can just access env variables or even execute kubectl to get state ouf of my target.