Millet Porridge

As a Python developer I basically never used jupyter, because IPython was good enough. Later I found colab can conveniently save debugging history and use local services. With a decent experience in some aspects, I've grown to love using it recently.

This post introduces a special optimization for Dockerfiles. Then, based on introducing such optimizations, I'll discuss existing tools' shortcomings and the use and extension of some emerging build tools. Finally, for these tools, I'll briefly discuss their use cases, hoping to help readers with image building.

In the previous blog, I have given the upterm capability to allow VSCode SSH connections. I'd like to give it more power to support VSCode Web.

Since I introduced `dashboard' in our PaaS system, my colleagues complained a lot that it was unstable, often disconnected and difficult to transfer files. It is good to know that the dashboard is widely used in our system and developers want it to have more features and stability.

In fact, we often consider the function expansion of API gateway for our PaaS platform. Recently, I started to look into API gateway products. In this blog, I will briefly introduce the function and give some considerations for future development. I hope to be able to attract more ideas. I would like readers to share their prospects for API gateways.

Earlier blogs introduced our use of K8s scheduled tasks and the source implementation of K8s cron jobs. But after actual use we found problems affecting user experience; I discuss solutions for each, hoping to help everyone, with usage suggestions appended.

Self-built K8s clusters have no shortage of pitfalls, especially as the Node count grows — problems gradually surface

Gitpod is a web-based VSCode that can open Github projects directly. When I first used it, only the web version existed — quite inconvenient. Retrying it recently, I found it can now open the local VSCode directly, similar to remote ssh. This tool feels full of potential; interested readers can explore it themselves. Here I briefly analyze the usage and principles; the second half explores the future direction of development modes and tools.

We needed to add Webshell functionality to our PaaS; reusing the native Dashboard directly seemed to reduce much development burden, so I planned secondary development. But Dashboard's own permission control is too weak, and its frontend pages are many and complex. To integrate it into our platform, the frontend needs trimming and the backend's authentication changed. This blog mainly introduces how to run and debug the program, and the concrete trimming process.

I personally think stream-like data structures exist mainly to postpone the moment of real computation — data is only produced when truly needed. Incidentally, their existence also solves how computers describe 'infinity': as long as you can always fetch data from the source, the source can be considered infinite. This blog shares a stream processing application — the prime sieve.