Analytics Solution for GitHub Actions. No Need to Run CI in the Dark
blog.trunk.ioTrunk employee here - we're super excited about this launch!
We've got a few companies using it, and had other users in our community tell us "It's a very helpful tool that I didn't even know I needed". Even we ourselves have made a number of surprising discoveries about how often some of our hourly/nightly workflows have been failing by dogfooding this.
Let us know what you think! :)
For cost and practical reasons, I'm seriously considering to adopt Dagger.io instead of GH Actions.
Have you considered integration with other CI/CD pipeline engines or it doesn't matter as long as you git-commit?
Thanks in advance.
Hey thanks for the Q: We will support other CI systems than GitHub Actions, but exact timeline of which and when is TBD.
With regards to cost, you can self-host GitHub Actions runners. Internally we use an autoscaling group of spot instances via k8s, which is much cheaper than GitHub-hosted runners. We actually didn't do it for the cost, we did it because we wanted a local cache to persist on machines during the day, for performance reasons. When cost is a factor, having an analytics system help you optimize your jobs is super important.
GH actions has been great for us for CI but incredibly opaque when you're trying to diagnose problems. Simple jobs running on the gh hosted runners are fine, but we have a large pool of self-hosted runners running more complex workflows, and whenever things go sideways the answer is usually "did you try restarting all of them?" Hopefully this is helpful, at least someone is investing in making gh actions work for larger organizations.
Interested to see how different case studies pan out with this product. A mentor once told me that data is only useful if you can tell a compelling story. Hopefully soon something like this will make it easier to tell stories that help both business and engineering.
"data is only useful if you can tell a compelling story"
I really like this - with this product, we're focusing on trying to tell the story and also make the data actionable. I'm hoping we can both show that there is a problem, as well as guide you to how to fix it. Otherwise, at the end of the day it's just colorful lines.
Datadog already does this, how does this compare? Genuinely interested.
I would say that broadly the ergonomics of the product are where it is most differentiated. Datadog's solution - feels very much like a datadog product - in that it's a giant dump of data you have to craft to tell a story. Not all pipelines are the same, their execution patterns are different and expectations for Pass/Fail/Cancel differ as well.
Tbh it looks the same as Datadog’s solution.
It's not just about the visuals - its about how the data can be sliced and diced. How your pull requests behave is a very different phenomenon than from how your main branch is behaving. You need to be able to tease this apart easily.
Can Datadog do this for $7/month? My understanding is that their pricing is quite a bit higher than that.
If you're already paying for Datadog the price probably isn't an issue for you though.
Pretty neat, does GitHub provide any of these analytics themselves?
shockingly it doesn't
Have you considered a Github buyout your exit strategy?