Original link: https://zdyxry.github.io/2022/08/14/Weekly-Issue-2022-08-14/
[[Databend]] Introduction to efficiency engineering, using SaaS to complete most of the work, Self Host Github Runner try to use [[AWS]] internal services to reduce traffic costs.
Efficiency engineering experience:
- Correct positioning
- incremental change
- Open and transparent
[[Lyft]] Improvements to [[envoy]] configuration, existing problems:
The original routing configuration is maintained using the [[Jinja]] template, which cannot perform good syntax checking and verification, and is now replaced by [[Golang]]
The original [[envoy]] API is referenced in both the configuration plane and the control plane. When [[envoy]] is upgraded, compatibility issues need to be considered, and now it is only referenced in the control plane
The original Route table check tool could not be distributed well (why would a binary not be distributed well?), it was changed to an internal check tool, and CI detection results were added.
Introduction to basic configuration parameters of CPU and Memory cgroup.
[[cgroup]] Use cpuset to avoid CPU throttling.
[[kubernetes]] How to remove namespace, usually because of [[Finalizers]].
Who should write Terraform? Based on the actual organizational situation, I understand that the ultimate expectation is that the Developer will write. Read together Developers Should Deploy Their Own Code
[[hashicorp]] Survey on Cloud. Multi-cloud is already a fact, mostly independent teams maintain multi-cloud resources. 94% of them believe that there is a waste of resources.
The author’s point of view is that using a top-tier server can take on a lot of services, and vertical scaling is far more convenient than horizontal scaling. From an availability point of view, you can use public cloud, but don’t use public cloud completely, the premium of public cloud is obvious. But this is all based on the large scale of the business scenario. If the business scale is small, then paying by volume is a good choice.
When should a [[microservice]] design be done? Don’t do this unless the reason is very clear.
When the existing program is still running well, a good module split may be a better choice than directly splitting into microservices, the former is itself a part of the latter. The premise of microservice design is that the infrastructure is sufficiently complete.
Explain why we need [[containers]] and [[kubernetes]] from an application deployment perspective. Before, from the perspective of application release (binary -> tar -> rpm/deb -> container) to explain why [[container]] is needed.
After all, we’re already in an age of “shit jobs”. And if this “shit” comes from a systemic force with a slightly irresistible color of the times, then we need to have more reflection on our work.
And finally, a bowl of “chicken soup” that Graeber hands over at the end of the book:
“Most people like to talk about freedom in the abstract, and even claim that freedom is the most important thing…but very few people think about what it means to live in freedom.”
So the purpose of this book is not to propose solutions, but to advocate for people to think and discuss true freedom and more reflection on “normal”.
“Shit job is a window”.
It not only looks at society from the outside, but also looks at life from the inside.
[[Shopee]]’s observation article
Little people’s love is always sweet.
This article is reprinted from: https://zdyxry.github.io/2022/08/14/Weekly-Issue-2022-08-14/
This site is for inclusion only, and the copyright belongs to the original author.