4 ms·

I think it is related to but not the same as the OP but to me, something that you don't think of much is that this was invented by Google for Google-scale issue

I think it is related to but not the same as the OP but to me, something that you don't think of much is that this was invented by Google for Google-scale issues. What does this mean? Lots of things that would be relatively easy for Google make it harder for much smaller companies. Logging? Google can build a log aggregator. Hardware? Google have 100s of 1000s of servers. Documentation? Originally, certain knowledge could be kept in small teams who worked on it, the documentation then had to be generated after-the-fact, which is perhaps why the docs are quite factual and not so good at introducing some of the concepts. You want GUI tools for setting up e.g. rbac and certs? Great but Google don't because they have people who can write the relevant automation scripts in their sleep so don't need what us mere mortals might prefer when new to it. I do like it though and use it for some of our systems.

motoboi 5y ago
Although people love to think of Kubernetes as “the thing google runs on that was open sourced”, it is actually a simplified version, written from scratch, based on the same principles of the tool google actually runs on. On used to run on, because it seems that there is no public information about the new platform after Borg. As a comparison, Mesos was built with the same goal (open source version inspired on Borg) and made some design choices Kubernetes haven’t. UPDATE: - I don’t really know why Google spent the effort on building Kubernetes, but I believe it was for Google Cloud. And it needed to be open sourced to form a community.
- ayewo 5y ago
  Your last sentence with respect to Apache Mesos led me to this previous HN discussion: Apache Mesos to be moved to Attic [1]. 1: https://news.ycombinator.com/item?id=26713082 https://news.ycombinator.com/item?id=26713082
- menage 5y ago
  > On used to run on, because it seems that there is no public information about the new platform after Borg. I left Google quite a few years ago, but my impression is that Borg is still going strong for many of Google's services. There was an attempt to replace it with a new system named Omega, but that ended up being more of an experiment - some of the findings of which were rolled back into future Borg development.
  - randomswede 5y ago
    My understanding is that it's still borg. All the cool features from omega were backported into borg, since they were needed to support (insert product here), and in the end some of the omega team got an OK to write and open-source kubernetes based on the findings from omega. When I left, the big ongoing project in that space was an effort to replace babysitter (most people in Google do not have to worry about babysitter, and for 98% of the time, I didn't, but the 2% where I needed to meant that I kept myself informed of changes in that space).
- kelp 5y ago
  I always thought Google built Kubernetes as a commoditize your complement strategy: https://www.gwern.net/Complement https://www.gwern.net/Complement
jaaron 5y ago
Maybe I shouldn't be surprised that so many don't understand where k8s came from. Google doesn't run on k8s. k8s was _not_ invented or designed to solve Google-scale problems. Since it's inception, the open source community and several corporate contributors have made k8s much closer to something that can solve google scale problems, but that's definitely not where it started, or arguably, where it is now.

motoboi 5y ago

Although people love to think of Kubernetes as “the thing google runs on that was open sourced”, it is actually a simplified version, written from scratch, based on the same principles of the tool google actually runs on. On used to run on, because it seems that there is no public information about the new platform after Borg. As a comparison, Mesos was built with the same goal (open source version inspired on Borg) and made some design choices Kubernetes haven’t. UPDATE: - I don’t really know why Google spent the effort on building Kubernetes, but I believe it was for Google Cloud. And it needed to be open sourced to form a community.

ayewo 5y ago
Your last sentence with respect to Apache Mesos led me to this previous HN discussion: Apache Mesos to be moved to Attic [1]. 1: https://news.ycombinator.com/item?id=26713082 https://news.ycombinator.com/item?id=26713082
menage 5y ago
> On used to run on, because it seems that there is no public information about the new platform after Borg. I left Google quite a few years ago, but my impression is that Borg is still going strong for many of Google's services. There was an attempt to replace it with a new system named Omega, but that ended up being more of an experiment - some of the findings of which were rolled back into future Borg development.
- randomswede 5y ago
  My understanding is that it's still borg. All the cool features from omega were backported into borg, since they were needed to support (insert product here), and in the end some of the omega team got an OK to write and open-source kubernetes based on the findings from omega. When I left, the big ongoing project in that space was an effort to replace babysitter (most people in Google do not have to worry about babysitter, and for 98% of the time, I didn't, but the 2% where I needed to meant that I kept myself informed of changes in that space).
kelp 5y ago
I always thought Google built Kubernetes as a commoditize your complement strategy: https://www.gwern.net/Complement https://www.gwern.net/Complement

jaaron 5y ago

Maybe I shouldn't be surprised that so many don't understand where k8s came from. Google doesn't run on k8s. k8s was _not_ invented or designed to solve Google-scale problems. Since it's inception, the open source community and several corporate contributors have made k8s much closer to something that can solve google scale problems, but that's definitely not where it started, or arguably, where it is now.