AI for Kubernetes with ChatGPT and k8sgpt

Поділитися
Вставка

КОМЕНТАРІ • 37

  • @DevOpsToolkit
    @DevOpsToolkit  Рік тому +6

    What do you think about AI for Kubernetes? Is it a gimmick or something we should use today or something that will become useful in the future?

    • @gvoden
      @gvoden Рік тому +3

      If it reduces operational tickets and stupid questions about why is my pod in crashloopbackoff then yes :)

    • @kaokakelvin6202
      @kaokakelvin6202 Рік тому +3

      i have used k8sgpt to troubleshoot issues in our cluster, including one that was a blocker for several weeks, very amazing tooling.

    • @tiriyoncontinuum9519
      @tiriyoncontinuum9519 Рік тому +3

      Use it carefully, as it can persuade the operator that a mistake is correct. Furthermore, excessive use might numb the operators problem solving ability.

  • @Michael-wt5vl
    @Michael-wt5vl Рік тому +9

    You are a very effective communicator. Great video once again.

  • @THEROOT1111
    @THEROOT1111 Рік тому +1

    Very interested in all your vids as i'm not a devop myself to get an experienced view of already available solutions. Keep up the good work

  • @XesivKuSH
    @XesivKuSH Рік тому +2

    Great analysis. Can't wait to see what more it will have to offer in future as it grows.

  • @FlashMan3
    @FlashMan3 Рік тому +4

    Great video Victor, keep going! 👌

  • @marinowijay
    @marinowijay Рік тому +1

    Fantastic and insightful videos as always, Viktor!!!

  • @radiantmind1079
    @radiantmind1079 Рік тому +1

    The master of the whole field ... such a great tutorial like always

  • @aniketmhala7299
    @aniketmhala7299 6 місяців тому +1

    excellent explanation.💐

  • @nilesh-gule
    @nilesh-gule Рік тому +1

    This looks like an interesting project. Would give it a try. Hopefully it will improve over period of time and your dream will come true to fix issue automatically.

  • @mehmetdemir4781
    @mehmetdemir4781 Рік тому +1

    Btw thank you for your job, you explain it very well

  • @AnkitKumar-ky6bn
    @AnkitKumar-ky6bn Рік тому +1

    Very Nice video

  • @felipeozoski
    @felipeozoski Рік тому +2

    Viktor you are the man, I don´t care :)

  • @tonycaveirametal
    @tonycaveirametal Місяць тому +1

    Are the data from our troubleshooting being used to train open ai models, this concern comes if data protection might be applied?

    • @DevOpsToolkit
      @DevOpsToolkit  Місяць тому +1

      k8sgpt can be configured to use over 10 different models. If data leaving your system is a concern you can use one of the self-managed models like, for example, Ollama.

  • @Ruben-by4oy
    @Ruben-by4oy Рік тому +2

    I would say we as DevOps engineers are will be around for at least next 10-15 years even if demand of the other specialities like QA or Programming will be less and less.

    • @peterolofsson9871
      @peterolofsson9871 11 місяців тому +1

      Definitely, some has to get an ass chewing from management when something goes wrong

  • @Fayaz-Rehman
    @Fayaz-Rehman Рік тому +1

    Thank you Good job.💥

  • @CloudNativeJanitor
    @CloudNativeJanitor Рік тому +1

    Glad you did it. Can't be avoided. Please embrace 😊. But don't you mean GenAI! Kubeflow, Sheldon, spark, jupyternotes were already used in K8s. AI is not a strange workload. Using GenAI to develop manifests, test them, optimise them would be easy and nice

    • @DevOpsToolkit
      @DevOpsToolkit  Рік тому +1

      Yeah. This time I did not go through AI/ML workloads running in Kubernetes or helping write manifests but rather using AI to analyze the state of Kubernetes resources.

    • @CloudNativeJanitor
      @CloudNativeJanitor Рік тому +1

      @@DevOpsToolkit Still I am glad indeed you are exploring and sharing your insights into this area🤗
      The analysis would be better if it is customized with a good data source as you started in the beginning, the monitoring tools, the console, in addition to some uptodate FAQ, or learned data about the issues would have produced much more superior and informative results

  • @mcnairymichael
    @mcnairymichael Рік тому +1

    Tremendous potential here. It would be interesting to see if it can be integrated into Gitops. Say, an option to create a PR for a fix instead of doing it automatically and then submitting it to a CI/CD pipeline.

  • @fenarRH
    @fenarRH Рік тому +2

    k8sgpt looks like a dummy data pipeline (for now) from k8s cluster to a openai backends and presents the response from openai on a cli level.

  • @ziaurrehman4738
    @ziaurrehman4738 Рік тому +1

    Awesome thanks

  • @leandrolandmayer16
    @leandrolandmayer16 Рік тому +1

    Since we are talking about AI, is there any chance of a video on gpt engineer?
    Great video, BTW

    • @DevOpsToolkit
      @DevOpsToolkit  Рік тому

      If you mean an engineer working directly with language models, that's not one of my main areas of expertise 😔

  • @mehmetdemir4781
    @mehmetdemir4781 Рік тому +1

    Is it really Aİ behind? I think one good thing would be to take good and bad resources (those with noticed errors on it), train the AI model and then use that model on your cluster. But to do that, you must get old events from a lot of clusters through a lot of companies. Which is not that easy.

    • @DevOpsToolkit
      @DevOpsToolkit  Рік тому +1

      In the nutshell, all it does is send events to chatgpt and output the results.

  • @microst99
    @microst99 Рік тому +1

    I think the analysis is skewed because of the gpt-3.5 model, instead of 4.

  • @julianomoraisbarbosa
    @julianomoraisbarbosa Рік тому +2

    # til

  • @kisali777
    @kisali777 Рік тому +1

    Another one

  • @weirdstuffsforyou
    @weirdstuffsforyou Рік тому +2

    Your videos are awesome but for me this is not worth using. I made lots of custom controllers that do far better than this without using AI. AI is nowadays just unwanted hype. It's still not good. Even chatgpt used to repeat things at one point.

  • @RakeshKumar-eb9re
    @RakeshKumar-eb9re Рік тому +1

    First viewer

  • @PythonRadar
    @PythonRadar Рік тому +1

    No issue with mentioning AI , it’s no longer hype at this point 🎉