4 ms·
> A cool feature would be "This pod stopped, and didn't restart for 10 minutes" and the reason is "{the reason in plain English}". There is a little "bulb" ico
by suryao 3y ago
> A cool feature would be "This pod stopped, and didn't restart for 10 minutes" and the reason is "{the reason in plain English}".
There is a little "bulb" icon on the top. That enables a "highlight mode" where you can select some text, usually kubernetes error messages. That is sent to an OpenAI LLM to (1) convert the error into plain english (2) show potential fixes.
This is limited by what GPT4 can provide but it is still helpful in some cases. Maybe we should highlight this feature more :)
> Kubernetes feels like a leaky thing to throw abstractions over, so the pain with using it is you need to become an expert in it, [..]
We believe two things: (1) k8s features exist because they are genuine user needs and we don't want to hide them (2) things get very complex very quickly because the "beast" is complex.
Argonaut's approach is what I call "Progressive Discovery". There is a good happy path that works for 80% of use cases. When you need to deviate, we make it easy to work with the underlying primitives to enable your use case.
This enables users to move fast without hitting an abstraction ceiling and it keeps the complexity in check.
That said - Argonaut is not for everyone. If you have just a service or two, you're better off running it on a hosted runner instead of an aws or gcp or azure. It is going to be cheaper at small scale too.
- quickthrower2 3y agoThanks. I think using AI to help with translating error messages, and perhaps finding the logs that matter, and even being an "expert system" of sorts to give an opinion on what went wrong is useful. Sort of how Windows diagnosis of why something isn't working (like wifi) has got better over the years, probably by adding more and more edge cases!
- suryao 3y agoYeah, I think there is a long way to go before the AI gets good enough to be trustworthy for infra type tasks but I am super excited for that future.
- frakkingcylons 3y agoI hope your AI helper is named Jason :)
- suryao 3y agoHaha
- RadiozRadioz 3y ago> sent to an OpenAI LLM to (1) convert the error into plain english It is unfortunate that this is necessary. This feels like a bodge instead of a fix for the underlying issue. Are the error messages genuinely impenetrable / the wrong level of abstraction, or are they appropriate and the engineers simply lack the training? I should point out that I'm not speaking down on the service or the people who use it. I think this is a useful and pragmatic feature, but it feels like those other issues should be looked at before we start throwing everything at LLMs.