Questions and feedback from FOSDEM

[raw belfry]

I received quite a few questions both while on stage and after.

One thing that came up more than once is how do pipelines work that push images to a private registry and/or deploy to production environments when executed locally.

I believe this is a great question, one that I'm still trying to figure out. But my current answer is that pipelines should be parametrized in some way so that you can either change the target environment or it detects it automatically based on the environment.

Another question that came up more than once is secret management.

To my understanding there are ways to share secrets with pipelines safely, but a better solution is still being worked on.

I also got questions about running service dependencies (for example for integration tests). @pine canyon said that it's also in progress, so that's what I told the audience.

Got one question about logs and visualization. I pointed them to @simple prairie's demo from one of the earlier community calls.

Extensions came up a few times in various contexts: reusability, currently having to duplicate the same code across languages for higher level libraries.

I had a couple confusing conversations as well:

• Some people asked about the difference between Make/Taskfile/etc and Dagger.
• Someone asked about compressing the layers resulting from the container build (not sure I understood the question at all)

Oh, observability came up at one point. How and where should observability be integrated (metrics, traces). The talk after mine was about building observability into CI/CD, so that also made me think how that would work with Dagger: https://fosdem.org/2023/schedule/event/how_we_gained_observability_into_our_cicd_pipeline/

[raw belfry]

Well, that's what I remember. (Note to self: write the questions down next time)

Maybe @pine canyon can add a few more that I forgot.

[sick granite]

just saw the video, nice presentation. Biggest takeaway I got from it: observability stack is complex to operate, setup and maintain. I see a good value by making Dagger do all this automagically and offer a hosted service to allow users to visualize, troubleshoot and understand what's happening with their pipelines.

[raw belfry]

Personally, I'd rather expect Dagger to integrate with my observability stack of choice. OpenTelemetry is what the industry seems to consider a standard these days.

[sick granite]

well.. this as well. Thing is.. there's many teams that don't even have an observability stack for CI/CD and don't want to maintain it.

[raw belfry]

I'm sure that's true. I'm just saying, focusing on the right integrations first and then figuring out what to provide as a service is the way to go IMO.

[pine canyon]

You've done an amazing job. Don't beat yourself too hard 🙂
Doing a talk and noting down questions live is quite hard.

I didn't catch other questions. Also, the Go devroom was so packed I couldn't even extract myself from my place to be able to join you for Q&A. So you caught more questions than me 🙂

[raw belfry]

Thanks! BTW, do you want to post our selfie somewhere?

[pine canyon]

Hi Mark, sorry I've been sick right after FOSDEM. But yes, for sure. It's a great idea!