Twitter | Search | |
Brian Grant
GKE co-TL, Kubernetes Steering Committee member, SIG Architecture co-Chair, CNCF TOC member
320
Tweets
171
Following
2,580
Followers
Tweets
Brian Grant Feb 1
Replying to @bgrant0607
It's already out of date! Here it is again, with LTS WG added.
Reply Retweet Like
Brian Grant Feb 1
I just updated the Kubernetes project group diagram. Will check it into the community repo somewhere.
Reply Retweet Like
Brian Grant Jan 29
Replying to @opsgoodness @tmTech
Looking forward to hearing more about your experience!
Reply Retweet Like
Brian Grant Jan 23
Replying to @thockin @vllry and 2 others
And the Omega paper:
Reply Retweet Like
Brian Grant Jan 23
Replying to @thockin @vllry and 2 others
Reply Retweet Like
Brian Grant Jan 19
Replying to @ahmetb
I have been waiting for this for a long time. The equivalent in Borg has been widely used in Google since before we started Kubernetes
Reply Retweet Like
Brian Grant Jan 17
Replying to @thockin
And respect semantic versioning
Reply Retweet Like
Brian Grant Jan 17
Replying to @jirv500 @castrojo and 5 others
Not at all. Thanks for providing feedback. That's how we get better. Thanks also for sticking with it. Most impactful changes take time
Reply Retweet Like
Brian Grant Jan 16
Thanks for pointing that out. That experience is not unique unfortunately. On my list of project management issues to think about
Reply Retweet Like
Brian Grant Jan 14
Making great progress towards completing the fundamental declarative primitives in Kubernetes.
Reply Retweet Like
Brian Grant Dec 20
Cool! Welcome!
Reply Retweet Like
Brian Grant Dec 20
Replying to @garethr @cnab_spec
I do understand the goal of deployment tool interoperability, though. CI tools/systems have a similar challenge. For instance, Jenkins plugins are coupled to Jenkins.
Reply Retweet Like
Brian Grant Dec 20
Replying to @garethr @cnab_spec
Doesn't seem especially cloud native. (1) Imperative. (2) Obscures but doesn't abstract the deployment environment. (3) Ditto for the deployment tools, which all have very different upgrade behaviors. (4) Lacks observability. (5) Not GitOps friendly.
Reply Retweet Like
Brian Grant Dec 20
Replying to @garethr @cnab_spec
Thanks for the writeup, Gareth. How does this compare with Docker Application Bundles, App Registry, Nulecule, and/or Elasticbox?
Reply Retweet Like
Brian Grant Dec 19
Definitely. Blast radius and availability are major issues at higher scales -- don't put all your eggs in one basket. That said, I believe we have room for significant scale improvements without radical changes.
Reply Retweet Like
Brian Grant Dec 19
From what I can tell, Terraform's all-or-nothing approach pushes users to adopt other tools to manage less static resource types, even for providers other than Kubernetes.
Reply Retweet Like
Brian Grant Dec 19
If you're interested in helping to improve K8s distributed systems semantics, there are definitely issues to sort out.
Reply Retweet Like
Brian Grant Dec 19
We should support optional immutability (independent of the authz mechanism).
Reply Retweet Like
Brian Grant Dec 19
IMO, apply is a significant innovation in K8s. There should be no hard line between human configuration and automation. Pushing merging into the ecosystem exacerbates tool fragmentation -- been there, done that. The API schema complexity is a bug, though.
Reply Retweet Like
Brian Grant Dec 19
Spec and status were split in Omega, and that has its own problems. We considered splitting at the storage layer, but were limited by etcd2 and 2-phase commit was too complicated to implement.
Reply Retweet Like