Is this project still alive?

[mrclrchtr]

Is this project still alive? Is it actively maintained? I've been using ARC for a while now and keep running into problems. Many of them are known as issues or in discussions. But it seems to me that no maintainers ever comment on them. And the development progress is not really noticeable.

I am currently in doubt as to whether I should continue to use ARC or whether I should use something like RunsOn or Ubicloud.

This should not be misunderstood, I am very happy that this project has been developed. However, it feels like it's at a dead end.

What does the future look like? Will it continue more actively at some point or is the air out of here?

[nikola-jokic]

Hey @mrclrchtr,

Can you please clarify what you mean when you say "is this project still active?". This project consists of 2 different API groups. One is maintained by the community while the other is maintained by GitHub. And please, let us know why you feel like the project is not actively maintained.

[mrclrchtr]

Hi @nikola-jokic,

thank you for the reply.

This project consists of 2 different API groups.

For example, I was not aware that there are two actively developed products. Does this mean that “legacy ARC” is still being developed or is the focus not completely on “autoscaling runner scale sets”?

And please, let us know why you feel like the project is not actively maintained.

There are so many unanswered questions to which there is unfortunately no response. Like here, for example:

• Autoscaling Runner Scale Sets with state (persistent volume as cache) #2878
• Docker layer cache in the autoscaling runner scale sets mode #2869
• Defining resources per repo/workflow #2795

One feature that I absolutely need, for example, is the caching of docker images. And I realize that this is needed by many and there is no simple solution yet.

Now the question is: should I go back to ARC, where this was possible without any problems? Or wait for this feature, which may never come because there is no response?

Maybe it's also the release cycle, which at least feels slow. One patch release per month is not that fast, is it?

To summarize:
The software is used by many large companies and many more smaller companies or even private individuals. I have the feeling that users are not getting the support they need.
I could be wrong, but I had to express that.

I'll say it again, don't get me wrong, I'm very happy and grateful for what has been achieved so far and is currently being achieved.

[nikola-jokic]

I am sorry you feel that way. However, some questions don't have a definitive answers. The idea behind the new scale set mode is to provide as much flexibility to users to pick the template based on their cluster constraints. That is the decision behind providing a runner template being the PodSpec. You can define spec based on your constraints, and ARC will be perfectly happy to spin it up for you. The new scaling mode ensures that it can manage the scaling demand by spinning up pods that you described using a template field. That is also one of the reasons we do not provide a Prometheus monitors out of box. You decide how to collect metrics, and we provide an endpoint that you can scrape.

As for the level of support for the community, my personal opinion is that it shouldn't be expected from maintainers of any open source project to keep up and answer every single discussion post. We try our best to keep up with issues, deliver fixes and continue improving ARC. There are many users of ARC with exceptional Kubernetes knowledge who have similar constraints like you, and they can definitely provide a better answer than I can for that particular setup.

[ctrlok]

It's not like anyone expecting enterprise-level support (still, k8s operator is kinda valuable part of any ci/cd system and making it "community maintained" should make it better compared to others, not worse)

Anyway, currently, there is a lack of documentation, and existing documentation is poor quality or covers just the simplest examples. For example: I read announce post few times (found it on second or third page of releases) — half of the links points on folders with go code. There is readme page with some kind of explanation (still, very general) and I checked — no other page points to it

And I'm not here to blame, I want to prove a point that documentation is super fragmented and it's very hard to understand what is what and how we can move our flows to the new version and can we do this at all

[mrclrchtr]

@nikola-jokic thanks again for the reply. As @ctrlok almost put it together well. I really don't expect enterprise support, but I've also read a lot of documentation and searched through discussions and spent hours trying to find a little info that helped me get the software up and running.

The community tries to help each other, but when information is missing and nobody knows what to do, sometimes help is needed from the maintainers (even if it's just a hint that things are not possible and not planned to be realized).

But I'm glad that we've talked about it here. The quick and friendly communication gives me confidence that this project will continue to evolve.

[ctrlok]

@mrclrchtr here are the values.yaml file for new sets and looks like it's possible to mount disks and use them as cache, I'll try it and will write a result later

[mrclrchtr]

Thanks, that's exactly how I tried it and gave up after hours/days 😅 if you find a solution, I'm very happy.

[Nastaliss]

I followed this deep dive video which allowed me to understand how the stack works and spin up action runners in about an afternoon !
I agree that having to look both in this repo's doc and in the official github doc was a bit frustrating at the beginning.

[jrbushell]

I'm also thinking the same question recently. Looking at the issues list, nothing has been triaged for 3 months. Hardly any PRs merged in the same period.